Outsmarting Metastatic Prostate Cancer: Integration of Imaging,Liquid Biopsies and Biomarkers With Artificial Intelligence

Abstract

Management of metastatic prostate cancer (mPCa) poses significant challenges due to inherent tumor heterogeneity and therapeutic resistance. Advances in molecular imaging, liquid biopsies, and biomarkers are enabling precision oncology, while artificial intelligence (AI), including machine learning (ML) and deep learning (DL), integrates complex datasets to improve diagnostic accuracy, risk stratification, and treatment guidance. This review highlights AI’s applications in mPCa, focusing on imaging, cell-free nucleic acids, circulating tumor cells, and genomic classifiers. We emphasize AI’s role in enhancing diagnostics and personalizing treatments, with implications for improving clinical outcomes through better decision-making. Finally, we discuss opportunities and challenges in deploying AI systems, stressing multimodal integration and validation for real-world clinical impact.

Keywords

metastatic prostate cancer artificial intelligence deep learning medical imaging biomarker diagnosis prognosis

1. Introduction

Prostate cancer (PCa) remains one of the leading causes of cancer-related mortality,¹ and the most frequently diagnosed malignancy among men in the United States² and Europe.^3,4 Metastatic prostate cancer (mPCa), particularly in its castration-resistant form (CRPC), accounts for a large proportion of prostate cancer-related deaths. The primary risk factors for PCa include African ancestry,⁵ and advancing age.⁶ Additional contributors include genetic mutation,⁷ family history,⁸ smoking,⁹ obesity and diets high in fat.¹⁰ Although recent advances, such as androgen receptor (AR) targeted therapies, taxanes, radioligands, and PARP inhibitors have expanded therapeutic options, outcomes remain variable and unpredictable, largely due to the profound intra- and inter-tumor heterogeneity of the disease.

Typical treatment pathway for PCa involves several stages: initial evaluation, confirmatory diagnosis, treatment selection, treatment intervention, and monitoring and prognosis. During initial evaluation, patients with suspected PCa routinely undergo rapid screening tests, including a digital rectal exam (DRE) and prostate specific antigen (PSA) test.¹¹ If uncertainty remains, additional diagnostic procedures are undertaken to confirm the disease. Commonly used approaches include imaging, tissue biopsy, liquid biopsy, genetic testing and patient stratification.¹² Imaging modalities consist of magnetic resonance imaging (MRI),¹³ transrectal ultrasound (TRUS),¹⁴ and positron emission computed tomography (PET), among others. Biopsy refers to the histological analysis of patient-derived tissue samples,¹⁵ while liquid biopsy represents a non-invasive option that evaluates circulating biomarkers. Relevant biomarkers include molecular and cellular indicators as well as exosomes.^16-19 Patient monitoring enables real-time assessment of disease status, while prognosis ultimately enables evaluation of survival and mortality outcomes after treatment.

In clinical settings, implementation of treatments for metastatic prostate cancer (mPCa) has long relied on established protocols, but these often struggle with tumor heterogeneity and resistance, limiting personalized care. As source of solutions, artificial intelligence (AI) is increasingly being integrated into clinical workflows, offering data-driven insights that augment human decision-making and enable more adaptive strategies. This transition from conventional clinical implementation to AI-enhanced approaches marks a pivotal evolution in oncology, promising greater precision and efficiency.

AI refers to the creation of computational systems designed to carry out tasks that traditionally require human cognition, such as pattern recognition, learning, reasoning, and decision-making. Over the past decade, AI, particularly through machine learning (ML) and deep learning (DL) approaches, has demonstrated transformative potential in biomedical research by efficiently uncovering complex patterns within large-scale datasets, including imaging, genomics, and electronic health records. Within oncology, AI has markedly improved the precision and efficiency of cancer detection, prognosis, and treatment planning.^20-22 Specifically, for PCa, AI algorithms trained on multiparametric MRI and digitized pathology slides enhance tumor identification, automate Gleason scoring, and aid clinical decisions regarding active surveillance or therapeutic intervention.^23-25 Moreover, new AI applications are being used to forecast resistance to androgen receptor–targeted therapies and to delineate molecular profiles linked to aggressive disease states.²⁶ As AI technology continues to progress, its incorporation into clinical practice offers strong potential for more tailored and precise cancer management.

Current pathology assessment in PCa faces several limitations. Major challenges include diagnostic subjectivity,²⁷ the labor-intensive nature of evaluation and a shortage of trained pathologists.²⁸ These issues highlight the need for AI-assisted diagnostic tools. AI encompasses a broad range of computational approaches, with machine learning and its subfield deep learning forming core components. Increasingly, AI has already shown strong potential in medicine, particularly in the analysis of medical images.^20,22 In PCa, applications include auto-segmentation, image registration, distinguishing benign from malignant lesions, tumor grading, and early diagnostic support in regulatory contexts.^29-32 Furthermore, AI-driven biomarker analysis has contributed to the discovery of novel biomarkers and has been applied to both diagnosis and prognosis of PCa.^33-35 Despite these advances, AI-assisted tools are not without limitations. Issues such as data standardization,³⁶ interpretability,³⁷ bias and fairness³⁸ and validation³⁹ remain unresolved and require continued investigation.

In this review, we outline the characteristic features and overall pathology of PCa. In addition, we provide a discussion of current diagnostic approaches, including imaging and liquid biopsy techniques. We highlight the integration of AI in clinical decision-making, with particular emphasis on its applications in biomarker-based diagnosis, prognosis, and novel biomarker discovery. Lastly, we address the key challenges and future opportunities in this field while providing our perspectives on potential directions for advancing PCa research and management.

1.1. Search Strategy

Literature search was conducted in PubMed, Embase, Scopus, Web of Science, and Cochrane Library for studies published between January 2015 and December 2025. Search terms included “prostate cancer,” “artificial intelligence,” “medical imaging,” “liquid biopsy,” “biomarkers,” “multiparametric MRI,” “PET/CT,” “ultrasound,” “cell-free DNA,” and “circulating tumor cells.” Inclusion criteria were primary references of original research, systematic reviews, and meta-analyses reporting on AI integration with imaging or liquid biopsy in metastatic prostate cancer. Data extraction focused on sample size, validation status, and key limitations. Risk of bias was assessed by using QUADAS-AI and PROBAST tools.^40-43

2. Current State of PCa Management

2.1. Prevalence and Features of PCa

PCa is the most frequently diagnosed noncutaneous malignancy and the second leading cause of cancer-related death among men in the United States.⁴⁴ Most PCa-related mortality arises from metastatic disease. Prognosis for localized PCa is excellent, with a 5-year survival rate of 99%. However, this rate declines sharply to approximately 30% in patients with distant metastases.^8,45 Once becoming metastatic, PCa frequently progresses to a castration-resistant or androgen-independent state, creating significant therapeutic challenges. Although a definitive cure for metastatic castration-resistant prostate cancer (mCRPC) has yet to be achieved, advances in prognostic markers, novel technologies, and therapeutic strategies are rapidly reshaping the treatment landscape.⁴⁶

2.1.1. Heterogeneity

Heterogeneity is one of the most common features of PCa. This disease hallmark of PCa arises from both inherent biological differences among individuals and differences in datasets. Age, race and geographic factors can contribute to the inherent heterogeneity of PCa, while differences in race can cause differences in PCa development. For example, FOXA1 mutations occur more frequently in Asian populations compared with European and American populations.⁴⁷ However, most of the studies are limited to populations in Western countries.⁴⁸ Multiple factors can cause heterogeneity of the PCa dataset, including staining variation,³⁹ artifacts⁴⁹ and imaging differences between scanners.⁵⁰ In contrast to conventional approaches, AI promises to offer the following technological capabilities. AI-based radiomics and machine learning can extract and quantify subvisual features of intratumoral heterogeneity that exceed human cognitive capacity in pathology practice.⁵¹ Specifically, habitat-based imaging using AI can partition tumors into distinct subregions based on imaging characteristics, providing quantitative measures of heterogeneity that predict aggressive disease and treatment resistance.^52,53 These approaches generate information about cancer cell subpopulation heterogeneity and tissue microarchitecture using both explicit and implicit features.⁵¹ AI software can track individual metastatic lesions across serial imaging timepoints to identify mixed responses where some lesions progress while others respond—a common clinical challenge in metastatic castration-resistant prostate cancer.⁵⁴ This automated analysis can detect oligo-progression suitable for directed radiotherapy or identify early widespread progression indicating need for treatment change, even when overall disease appears stable.⁵⁴ The ability to concisely assess disease status across multiple heterogeneous lesions addresses a critical gap in managing high-volume disease.⁵⁴

AI algorithms applied to histopathology can predict molecular aberrations (BRCA, homologous recombination deficiency, mismatch repair deficiency) from H&E slides, typically achieving AUCs of 0.72-0.91⁵⁵. This allows pre-screening to identify which patients harbor targetable molecular alterations without requiring expensive molecular testing of all lesions, thereby addressing inter-tumoral molecular heterogeneity.⁵⁵ AI can also identify novel cancer-related genes and predict resistance patterns to antiandrogen therapy.⁵⁶ Advanced radiomic approaches are capable of modeling inter-lesion relationships within individual patients, creating network-based representations that capture how multiple heterogeneous lesions collectively define disease phenotype and prognosis.⁵⁷ This moves beyond single-lesion analysis to characterize the “entire” patient based on their constellation of heterogeneous “unit”.⁵⁷ AI models combining imaging, genomic, and clinical data enable tailored treatment strategies that account for heterogeneity by predicting which patients will benefit from specific therapies, optimal androgen deprivation therapy duration, and risk of progression.^58,59 Machine learning can identify cell line phenotypes resistant to therapy and detect novel targetable signaling pathways.⁵⁶

2.1.2. PCa Development

PCa development includes three stages: Initial stage, localized PCa and metastatic PCa. In the initial stage of PCa development, chronic inflammation is present, which later progresses to proliferative inflammatory atrophy (PIA), a benign precursor of prostatic intraepithelial neoplasia and PCa.⁶⁰ According to one study, chronic inflammation is a key feature in the development of many cancers, including PCa.⁶¹ This process can induce epithelial hyperplasia and oxidative DNA damage in prostate epithelial cells.⁶² PIA is hypothesized to result from regenerative proliferation of prostate epithelial cells following injury, a process mediated by inflammatory oxidants.⁶³

AR plays a crucial role in the localized stage of PCa. As a ligand-dependent transcription factor, AR drives PCa development and progression. Upon binding with ligands (dihydrotestosterone, testosterone, or other androgens), AR translocates to the nucleus, dimerizes, and regulates the expression of various genes that promote cell proliferation or inhibit apoptosis.^64,65 As PCa advances, patients often develop obstructive urinary symptoms including difficulty urinating, a weak urine stream, or pain during urination.

Metastatic PCa represents a more advanced and severe stage compared to localized disease, often characterized by a greater symptom burden and higher mortality. It commonly involves the axial skeleton but may also spread to the lungs, liver, pleura and adrenal glands.⁶⁶ Bone metastases, in particular, are frequently associated with significant pain.

2.2. Diagnosis and Treatment

Typical PCa treatment pathway includes initial evaluation, diagnosis, treatment planning, therapy administration, continued monitoring to guide prognosis.

2.2.1. Diagnosis

In initial evaluation, general practitioners (GPs) will obtain the medical history and perform relevant physical examinations on patients suspected of having PCa. The two most common assessments are DRE and PSA test. The DRE allows palpation of the prostate to detect abnormalities while PSA test measures the level of PSA, a protein produced by the prostate gland, in the blood. So far, PSA testing remains the most commonly used diagnostic test for PCa¹¹ and plays a crucial role in both diagnosis and prognosis for PCa.^67,68 In general, higher PSA levels are associated with an increased likelihood of PCa. The American Urological Association provides threshold guidelines for interpretation of PSA concentrations.⁶⁹

However, higher PSA levels can also rise with non-malignant conditions, such as benign prostate hyperplasia, which becomes common with increased age in male. PSA levels can also be elevated following prostate inflammation, leading to false-positive cases where men with normal PSA levels are diagnosed with PCa and false-adverse cases where men with normal PSA levels still harbor cancer.⁷⁰ Therefore, PSA is not considered a PCa-specific biomarker.⁴⁴ To improve diagnostic accuracy, clinicians may consider PSA density (calculated by dividing the PSA level by prostate volume)⁷⁰ or using other ancillary tests such as the free-to-total PSA (fPSA) percentage, prostate health index (PHI), 4K score, etc.⁷¹ Among these options, fPSA has demonstrated particular utility to increase diagnostic accuracy.⁷² Importantly, advanced imaging is required to confirm PCa diagnosis. Commonly used imaging modalities include multiparametric MRI (mpMRI), transrectal ultrasound (TRUS), positron emission computed tomography (PET), among others. Findings from these evaluations are then reviewed by a clinical team to tailor treatment strategies collaboratively with the patient.⁷³

2.2.2. Cancer Genetics

Genome-wide association studies (GWAS) refer to a research method for uncovering genetic factors associated with complex diseases by analyzing large-scale population DNA samples.⁷⁴ In recent years, GWAS have been successfully applied to resolve genetic components of many complex human diseases, including PCa.¹² For example, SPOP, a gene that encodes a substrate-binding unit of a Cullin-based E3 ubiquitin ligase, along with TP53 and PTEN, is reported to be among the most frequently mutated genes identified in several studies of localized PCa.⁷⁵ In addition, most PSA-screened PCa in White patients have recurrent gene fusions, typically fusing the 5ʹ untranslated region of an androgen-regulated gene to nearly the entire coding sequence of an ETS-family transcription factor.⁷⁶ Genetic analysis of key PCa-associated biomarkers can aid in patient stratification based on PCa risks. For example, by using biopsy tissues, a cell-cycle progression score based on 31 genes can predict clinical progression and PCa-related mortality.⁷⁷

2.2.3. Treatment

Multiple treatment options exist for PCa. Active surveillance (AS) is suitable when immediate treatment does not offer additional benefits. Radiotherapy (RT) aims to kill cancer cells by carefully balancing radiation dose distribution, target volume and organ-at-risk considerations. Brachytherapy, another RT-based treatment, uses implanted radioactive sources in the prostate to localize therapeutic radiation.⁷³ Surgery is another effective approach, with prostatectomy being the most preferred curative option.⁷⁸

Other treatment modalities are emerging at the clinic, which include cryotherapy, hormone therapy, chemotherapy and immunotherapy. Cryotherapy uses gas inserted through needles to freeze cancer cells, where AI shows potential in predicting optimal temperatures settings.⁷⁹ Continuous monitoring is also important for predicting prognosis with AI partaking the real-time monitoring and prognosis modeling. For postoperative prognosis following radical prostatectomy (RP), predicting the likelihood of biochemical recurrence (BCR) and selecting adjuvant therapies or other treatment options are critical considerations.^80,81

For localized PCa, common treatment strategies include active AS, RP and RT. RP is considered the preferred strategy for low-risk PCa⁸² with the primary goal of curing the disease while preserving continuance and sexual function. Among RT strategies, external beam radiotherapy (EBRT) is a standard curative treatment for men with low to intermediate-risk disease.⁸³ In a 5-year post-treatment follow-up study,⁸⁴ men treated with RP experienced worsened urinary incontinence compared to those who underwent EBRT, though both groups reported similar declines in sexual function due to erectile dysfunction.

For metastatic PCa, androgen deprivation therapy (ADT) remains the standard of care. ADT has been shown to improve survival rate but is also associated with some adverse effects, including osteoporosis, loss of libido, sexual dysfunction, metabolic disturbances, and fatigue.⁸⁵ A combination therapy of taxanes and ADT has also been employed. Taxanes stabilize microtubules, preventing cancer cells from dividing, thereby promoting apoptosis. Docetaxel, a taxane chemotherapeutic agent that binds and stabilizes beta tubulin within microtubules, prevents depolymerization during mitosis and induces apoptosis.⁸⁶ Clinical studies have demonstrated that docetaxel improves overall survival in metastatic PCa.⁴⁶ However, resistance often develops, leading to disease progression.⁸⁷

Apart from chemotherapy, androgen-signaling targeted inhibitors such as abiraterone acetate and enzalutamide have been utilized in metastatic PCa.^88-91 Immunotherapy and bone-targeted radiopharmaceutical agents provide additional options. Sipuleucel-T, for example, consists of autologous peripheral blood mononuclear cells activated ex vivo with a recombinant fusion protein comprising the prostate antigen, prostatic acid phosphatase and an immune cell activator.⁹²

Gene-directed therapies are also available. BRCA2 is the most commonly mutated DNA repair gene in PCa and this underscores the potential benefit of poly(ADP-ribose) polymerase (PARP) inhibition and platinum-based chemotherapy.⁹³ Olaparib, a PARP1/2 inhibitor, already approved for advanced ovarian and breast cancers with germline BRCA1/2 mutations, has been considered in PCa management.⁹⁴ Despite these therapeutic advances, questions remain regarding optimal sequencing and combination of available treatments. More importantly, effective strategies of personalizing therapy to maximize benefit for individual patients are still lacking.

2.2.4. Prognosis

In terms of prognosis in PCa treatment, extensive studies indicate that the mortality rate for men treated with active surveillance is approximately 1% at 10 years.^95,96 Furthermore, research has revealed that combining enzalutamide with ADT significantly reduced radiographic disease progression or death by 61% compared with placebo plus ADT.⁹⁷ In addition, a study demonstrated the benefit of adding docetaxel to ADT for patients with hormone-sensitive prostate cancer (mHSPC), especially those with high-volume disease.⁹⁸ As for RP, a trial that randomly assigned 695 men with localized prostate cancer showed an overall mortality rate of 56% at 18 years for those treated with RP.⁹⁹ Finally, regarding the prognosis of radiotherapy (RT), a study that compared mortality rates between patients receiving hormonal treatment alone (3 months of total androgen blockade followed by flutamide) and those receiving the same hormonal treatment combined with radiotherapy showed that radiotherapy significantly improved 10-year overall mortality.¹⁰⁰

ADT treatment, however, can lead to drug resistance. Metastatic PCa is divided into metastatic hormone-sensitive prostate cancer (mHSPC) and metastatic castration-resistant prostate cancer (mCRPC) based on their response to ADT.¹⁰¹ Nearly all patients with mHSPC treated with ADT will eventually progress to the lethal stage of mCRPC.^102,103 The median survival for men with mCRPC is only 10 to 12 months in the absence of effective treatment strategies.¹⁰⁴

2.3. Medical Imaging

Magnetic resonance imaging (MRI) is an imaging technique that utilizes a combination of magnetic fields and radio waves to capture comprehensive images of organs and tissues. MRI plays an important role in diagnosis, and is a significant alternative method to TRUS guided biopsy for accurately excising target tissues.^105-107 However, MRI is subject to certain drawbacks. The resolution of MRI is important, but higher resolution comes at a higher cost. The need for higher quality imaging has provided an impetus for innovation, with AI tools being developed to enhance the resolution of MRI.

To improve diagnostic accuracy, the application of multiparametric MRI (mpMRI) in diagnosis proves particularly relevant. Undergoing mpMRI before biopsy has been recommended by both the European Association of Urology (EAU) and the American College of Radiology (ACR).^73,108 The examination typically takes about 20 to 45 minutes, and afterwards, a radiologist assesses the prostate by performing anatomical measurements (dimensions and volumes), calculations of PSA density, and assessment of the MRI scans. The result is measured using the PI-RADS scale, the most used classification system for the interpretation and reporting of prostate MRI.¹³ mpMRI comprises different types of images, including: T1W&T2W, Diffusion-weighted, Magnetic resonance spectroscopy (MRS). And Dynamic contrast-enhanced (DCE) MRI. Typically, T1W images and T2W images are used to visualize soft tissue anatomy, particularly water and fat content.¹⁰⁹ T1W images suffer from low zonal anatomy discrimination and post-biopsy artifacts.¹¹⁰ In T2W images, PCa appears homogenous, hypointense, and ill-delineated^111,112 due to diminished water content.¹¹³ Diffusion-weighted image contrast is generated by the random microscopic motion of water protons. In diffusion-weighted images, bright lesions are visible due to the dense collagen-rich tumor microenvironment.^114,115 Magnetic resonance spectroscopy (MRS) provides metabolic information, with PCa lesions exhibiting higher quantities of choline and lower levels of citrate due to the higher cell membrane turnover and density.^116,117 Dynamic contrast-enhanced (DCE) MRI is a technique that monitors the distribution of a gadolinium-based contrast agent for a period of time after the intravenous injection.¹¹⁸ In DCE-MRI images, PCa appears to have earlier, faster, and more intensely enhancing areas due to the high density and increased permeability of angiogenic microvasculature.¹¹⁹

PET is primarily used for cancer staging, tracking the effects of radiotherapy, and the detection of metastases. The development of AI assisted tools focusing on PET scan assistance has been an unpopular option for PCa diagnosis, as PET is most often employed alongside CT images.⁷³ Transrectal Ultrasound (TRUS) was first introduced in 1968¹²⁰ and was primarily used for guiding systematic biopsy.¹⁴ TRUS takes advantage of echoes: high-frequency sound waves generated by the transducer travel through the body and reflect back when they encounter an object, with the returning echoes providing information about the internal structures. Typically, an ultrasound probe is inserted into the rectum to obtain US images, while biopsy needles are simultaneously inserted into the prostate to retrieve the tissue samples.

2.4. Artificial Intelligence (AI)’s Potential in PCa Applications

Traditional diagnostics, including PSA, digital rectal exam, and TRUS-guided biopsies, lack specificity and fail to capture the dynamic molecular landscape of PCa. In recent years, technologies such as mpMRI, Prostate-Specific Membrane Antigen Positron Emission Tomography (PSMA-PET), liquid biopsies (cfDNA, CTCs), and gene signature panels (e.g., Decipher, Oncotype Dx) have improved risk stratification. However, the full potential of these technologies is limited by the complexity and volume of data they generate. Artificial Intelligence (AI) offers the capacity to analyze and integrate heterogeneous data, identify hidden patterns, and facilitate real-time clinical decision-making. In this review, we explore the roles and implications of AI in transforming the diagnostic and therapeutic landscape of metastatic prostate cancer.

2.4.1. Theory

Artificial intelligence (AI) encompasses an extremely diverse field of computational methods. Machine learning and, more specifically, deep learning have become two of the most prevalent forms of AI in the healthcare space. Machine learning strategies can be categorized into supervised learning, unsupervised learning and semi-supervised learning, while deep learning is a subfield of machine learning. Machine learning models are developed from direct data input and are then capable of performing discriminative and generative tasks.^118,121,122 Data input to a supervised learning model is labeled, while in unsupervised learning models, unlabeled data is provided and the models find hidden patterns within the data. Semi-supervised learning is a combination of both supervised learning and unsupervised learning, wherein models are trained on small and labeled datasets, then tested on larger datasets. Within the field of machine learning, artificial neural networks (ANNs) represent one major approach, and deep learning is a specialized subfield of neural networks. ANNs are universal function approximator^123,124 that are capable of discriminative and generative tasks.¹²⁵ ANN systems consist of artificial neurons, also called hidden layers, with each layer transforming the input data into an output by multiplication with the weight matrix.¹¹⁸ Deep learning is a more complex ANN architecture connected by multiple layers for extracting features.¹²⁶ Deep learning includes convolutional neural networks (CNNs), generative adversarial networks (GANs), transformers and so on. CNNs are a series of filter layers for image processing,¹²⁶ which consists of a convolution layer, pooling layer, and fully connected layer.

2.4.2. Application

AI displays immense potential in facilitating the entire PCa diagnostic pathway. AI application within this pathway is shown in Figure 1. For now, human assessments in PCa treatment demand AI assisted tools due to inherent subjectivity, processing time and a shortage of human labor. Human assessment is inevitably susceptible to subjectivity, which causes low consistency of diagnostic and prognostic results.^127,128 It is vital to increase diagnostic accuracy as misdiagnosis can result in mistreatment.^27,129 Human assessment is time-consuming and laborious so there is a need of AI-assisted tools. Multiple AI-assisted analytic modalities have been explored in metastatic prostate cancer, ranging from mpMRI and PSMA PET/CT–based imaging tools to machine learning approaches applied to TRUS, cell-free DNA, and circulating tumor cells. However, these approaches vary substantially in validation maturity and face modality-specific limitations (Table 1). According to one study, each patient undergoes at least 12 needle biopsies, producing 15 million biopsy samples per year worldwide,¹³⁰ which places a significant burden on pathologists, worsening the already dwindling supply of pathologists available to evaluate these samples.²⁸ To aid human assessment, computer-aided design (CAD) algorithms have been developed since the 1990s.¹¹⁸ Presently, AI demonstrates promising results in the medical field, especially through the analysis of medical images.^20-22 Several AI tools have gained approval from the US Food and Drug Administration.¹³⁰ With the advent of whole slide imaging (WSI) scanners^131-134 and digitalization of pathology laboratories, the future of AI in PCa treatment is promising.¹³⁵

Figure 1.

Traditional treatments for prostate cancer (PCa) include hormonal therapy, radical prostatectomy, immunotherapy, management of bone metastases, chemotherapy, personalized therapy, radiotherapy (RT), and radioligand therapy (RLT). These approaches generate large volumes of clinical and imaging data. Artificial intelligence (AI) can efficiently process and integrate these data to support clinical decision-making and optimize treatment strategies, with the potential to improve patient outcomes

Table 1.

Summary on Major AI Applications and Their Characteristics in Cancer Research

Domain	AI tool/Model	Sample size	Validation status	Key limitations	References
mpMRI	Deep learning segmentation/classification (PI-CAI, radiomics)	7,398 (meta-analysis); 1,000–5,000 (single studies)	Multi-institutional, some FDA-cleared, external validation limited	Inter-reader variability, small datasets, lack of standardization	40-43,136,137
PET/CT (PSMA)	AI for lesion detection, burden estimation, response prediction	11 studies, 100–1,000 per study	Mostly retrospective, low risk of bias, limited clinical parameter integration	Heterogeneity, limited prospective validation	40,41
TRUS	ML for lesion localization, fusion biopsy guidance	500–2,000	Single-center, pilot studies	High false-negative rate, anatomical coverage	4,138,139
Cell-free DNA	ML for biomarker discovery, risk stratification	200–1,000	Retrospective, external validation rare	Data heterogeneity, cost, integration challenges	43,140
Circulating Tumor Cells	ML for detection, prognosis	100–500	Pilot studies, limited validation	Low sensitivity, technical variability	43,140

AI, such as machine learning/deep learning, has empowered substantial improvements in diagnostic accuracy and precision for CRPC through advanced image analysis and automated segmentation techniques. Deep learning algorithms applied to multiparametric MRI achieve area under the curve (AUC) values of 0.768-0.854 for predicting progression to CRPC, with multimodal fusion models combining radiomics and clinical features reaching accuracies up to 94.2%.^141-143 AI-powered automated segmentation of PSMA PET/CT imaging demonstrates 98% accuracy in identifying lymph node involvement and metastatic disease, with sensitivity ranging from 62-97%, while significantly reducing inter-reader variability and reporting time.^42,144 In digital pathology, AI systems provide automated Gleason grading with 97.7% sensitivity and 99.3% specificity on core biopsies, minimizing inter-observer variability and enabling consistent risk stratification.^144,145These AI-based diagnostic tools not only enhance lesion detection and characterization but also reduce nonessential biopsies through high negative predictive values (97.5-98.0%) when configured for high-sensitivity rule-out, thereby decreasing overdiagnosis and improving workflow efficiency.^144,146

Consistently, AI technologies are transforming treatment selection, therapeutic monitoring, and prognostic prediction in CRPC through integration of multimodal clinical, genomic, and imaging data. Machine learning-based decision support systems for optimal sequencing of CRPC agents achieve C-indices of 0.729-0.827 for predicting cancer-specific and overall mortality across first-, second-, and third-line treatments, enabling personalized visualization of survival outcomes for individual patients.¹⁴⁷ AI-enabled prognostic tools analyzing H tissue images successfully stratify patients by metastatic risk, with high-risk groups showing significantly shorter metastasis-free survival (hazard ratio 0.19-0.39 for treatment benefit), facilitating early treatment intensification decisions.^148,149 Automated volumetric tumor burden assessment using AI on PSMA PET/CT predicts survival after Lu-177-PSMA therapy with C-indices of 0.71, demonstrating significant differences in median overall survival between low-risk and high-risk groups (30.9 vs 7.9 months).¹⁵⁰ Furthermore, AI platforms like CURATE.AI enable dynamic dose optimization for combination therapies by continuously predicting optimal drug doses based on individual patient response profiles (R²=0.96 for dose-efficacy correlation), improving treatment tolerability while maintaining efficacy.¹⁵¹ These AI-driven approaches have facilitated precision medicine by identifying genomic biomarkers (BRCA mutations, HRR defects, MSI-H/dMMR status) that predict treatment response to PARP inhibitors and immunotherapy, guiding selection of targeted therapies and improving clinical outcomes while restricting treatment-related toxicity.^56,152

3. AI in Imaging for Metastatic Prostate Cancer

AI has a wide range of applications in medical imaging. As shown in Figure 2, AI demonstrates capabilities in assisting diagnosis with medical images.

Figure 2.

AI has been applied to medical imaging across the three main stages of PCa management: Initial assessment and diagnosis, treatment, and monitoring. In the diagnostic stage, AI can automatically segment regions of interest (ROIs), assist in differentiating benign from malignant lesions, and support tumor grading. During treatment, AI contributes to image registration and clinical decision support, enabling more precise and individualized interventions. In the monitoring stage, AI can detect subtle predict patient prognosis and longitudinal changes, thereby facilitating timely adjustments to treatment plans

3.1. Multiparametric MRI (mpMRI)

AI-enhanced analysis of mpMRI has significantly improved the detection of clinically significant prostate lesions. A study revealed the potential in applications of AI algorithms for prostate gland segmentation, lesion identification, and classification using mpMRI and TRUS-Bx images.¹⁵³ AI can assist in several tasks on MRI images, including auto-segmentation, longitudinal comparison, lesion detection and classification, and PI-RADS scoring. Segmentation is required for accurate volume assessment. Deep learning approaches have shown more promising results, with DICE scores (a measurement of the overlap between the algorithm’s output and a ground truth) ranging from 80% to 90%.^23,154 According to the results, the Dice similarity coefficient was 87.12% for the entire prostate and 76.48% for the transition zone. In the prostate mpMRI longitudinal comparison process, it is necessary to review prior examinations and compare them to recent results, and AI-assisted comparisons can help identify subtle and tiny changes. AI can highlight suspicious areas on the images for lesion detection and classification tasks.²⁴ The results demonstrated that this AI system improved prostate cancer detection performance, achieving an area under the curve (AUC) of 0.93. In terms of PI-RADS scoring, one approach is to use a model to provide several components for radiologists, with the final decision made by the radiologists; the other approach is to train an algorithm to output the PI-RADS score directly.²⁵ Typically, the second type of algorithm outputs a pathology-based Gleason score instead of a PI-RADS score. For the 1,034 detected lesions, the kappa score between the AI system and the expert radiologist was moderate at 0.40. However, in 86 patients undergoing targeted biopsy, there was no significant difference in the detection rates of clinically significant cancer across any PI-RADS scores (P = 0.4–0.6).

3.2. PET Imaging and Theranostics

AI models have been applied to PSMA-PET/CT imaging for lesion detection and quantification. DL algorithms can distinguish bone metastases from benign degenerative changes and identify patterns predictive of treatment response. Theranostic applications, such as 177Lu-PSMA-617, are being paired with AI-driven models to optimize patient selection and monitor therapeutic efficacy.

PET is mostly used for cancer staging, tracking the effects of radiotherapy, and the detection of metastases. The field of AI-assisted tools used alongside PET for the diagnostic approach to PCa has not become very relevant as PET is not the optimal choice for PCa diagnosis.⁷³ PET is often used alongside CT imaging and AI-assisted tools typically operate and train using PET/CT images. AI has been applied on PET/CT images for both auto-segmentation and prognosis prediction tasks. For example, CNNs have been applied to auto-segmentation tasks on 18F-choline (FCH) PET/CT scans obtained prior to radical prostatectomy (RP) in 45 patients with newly diagnosed PCa.¹⁵⁵ The mean weight (range) of the prostate specimens was 44 g (20–109 g), while the CNN-estimated volume was 62 mL (31–108 mL), with a mean difference of 13.5 g or mL (95% CI: 9.78–17.32). Additionally, researchers proposed a deep learning model for localization and identification of cancer sites by combining hand-crafted perfusion-based features with deep autoencoder-based features generated from a dynamic¹¹ C-choline-PET/CT image set.¹⁵⁶ This model achieved a detection performance AUC = 0.812 for the benchmark set of 12 cases. To predict disease progression, Aloni et al proposed a model based on the analysis of PET/CT images,¹⁵⁷ obtaining the best performance in discriminant analysis (DA) classification (Sensitivity 72%, Specificity 68%, and Accuracy 69%).

3.3. Transrectal Ultrasound (TRUS) Imaging

Transrectal Ultrasound (TRUS) provides detailed imaging of the prostate gland and the surrounding tissue.⁴ TRUS can also be used to assess prostatic volume¹⁵⁸ and in the detection and staging of lesions.¹⁵⁹ It is highly recommended to use TRUS in suspicious areas,¹⁶⁰ as sonogram images help determine the location of the biopsy needles and the origin of the tissue samples.¹⁶¹ Even though TRUS has achieved great success, it still has some drawbacks, such as limited results and a high false positive rate. About 30% of prostate malignancies are isoechoic^14,162 and benign hyperplasia and prostatitis also appear as hypoechoic lesions in greyscale TRUS.¹⁴ One study proposed a deep learning model for real time auto-segmentation using TRUS images.²⁹ The model performance results demonstrated significant improvements in prostate segmentation compared with conventional automated techniques, achieving a median accuracy of 98% (95% CI: 95–99%), a Jaccard index of 0.93 (0.80–0.96), and a Hausdorff distance of 3.0 mm (1.3–8.7 mm). In addition, Zhu et al employed a deep convolutional neural network for image registration of 3-dimensional transperineal ultrasound prostate images.³⁰ As the results showed, convolutional neural network registration errors were smaller than 5 mm in 81% of the cases for 83 image pairs from 5 patients.

Tumor hypoxia is linked with aggressive phenotypes and treatment resistance. By means of unsupervised clustering and curve fitting on DCE-MRI, AI can delineate tumor subregions (habitats) with distinct perfusion profiles. Interestingly, these subregions often correlate with genomic signatures of hypoxia (e.g., EPAS1, HIF3A).

4. AI in Liquid Biopsy Interpretation

Liquid biopsy technologies, such as the analysis of cell-free DNA (cf-DNA), circulating tumor cells (CTCs), and exosomes, have revolutionized the non-invasive monitoring of mPCa by providing real-time insights into tumor dynamics, genomic alterations, and therapeutic responses. However, the high-dimensional and noisy nature of these biomarker datasets often challenges traditional analytical methods, leading to limitations in sensitivity, specificity, and clinical interpretability. AI, through machine learning (ML) and deep learning (DL) algorithms, significantly enhances the interpretation of these biomarkers by automating feature extraction, identifying subtle patterns, and integrating multimodal data for more accurate predictions. The following sections discuss detailed specific AI applications in liquid biopsy components.

4.1. Cell-free DNA (cfDNA)

There is growing interest in the genomics and proteomics of PCa as gene mutations play a critical role in the development of PCa.¹⁶³ Numerous biomarkers have been identified^16-19; however, an ideal and comprehensive panel for PCa diagnosis and prognosis has yet to be established, necessitating further research. The discovery and validation of clinically significant novel biomarkers remain essential.⁴

Biomarkers play a central role in liquid biopsy, an approach for diagnosing tumors and monitoring tumor-associated molecules in biological fluids such as blood and urine.¹⁶⁴ For example, cell-free DNA (cfDNA), fragments that shed into the bloodstream, offers valuable insight into cancer-driver genes and mutation.⁶⁰ Clinical studies have shown that men with PCa exhibit significantly higher levels of cfDNA compared to healthy controls.^165,166 In addition, extracellular vesicles also offer diagnostic potential. Large extracellular vesicles (L-EVs) isolated from the plasma of PCa patients contain genes frequently mutated in metastatic PCa cancer cells including MYC, AKT1, PTK2, KLF10, and PTEN. Moreover, the DNA content of L-EVs is associated with the development of PCa.¹⁶⁷ According to a report in 2018, cfDNA concentrations decreased significantly in patients after paclitaxel chemotherapy, suggesting that cfDNA quantification could be leveraged as a non-invasive biomarker for monitoring therapeutic efficacy.¹⁶⁸ Such findings underscore the dual utility of cfDNA and EVs in both diagnosis and longitudinal disease management.

4.2 Cellular Biomarkers in Cancer Stem Cells (CSCs)

Cellular biomarkers attract growing interests as they provide insight into the pathogenesis and progression of PCa.¹⁶⁹ Tumor-associated cells in the blood through interactions with neutrophils, platelets, cancer-associated fibroblast (CAFs), and tumor-associated macrophages (TAMs) are particularly promising as biomarkers.¹⁷⁰ Among cellular biomarkers, the most widely studied are cancer stem cells (CSCs) and circulating tumor cells (CTCs).

CSCs are self-renewing, tumor-proliferating cells within the cancer mass.¹⁷¹ A recent study suggests that CTCs may contribute to resistance to conventional therapies, opening avenues for novel therapies for PCa.¹⁷² Reported PCa stem cell biomarkers include integrins, CD44, CD133, CD166, and CD117.⁷⁴ CD177 is a cell surface protein used to recognize hematopoietic progenitor cells in bone marrow and has been linked to self-renewal capacity and cancer progression.¹⁷³

CTCs show substantial potential for PCa diagnosis and prognosis. ACTC positivity has been identified as an adverse prognostic indicator and may aid early detection.¹⁷⁴ The CellSearch platform, which detects and enumerates CTCS via immunomagnetic capture followed by fluorescence imaging, has gained FDA approval.⁷⁴ With next-generation sequencing (NGS) and sensitive CTC assays, CTC analysis has become a sophisticated component of liquid biopsy workflows.⁶⁰

5. AI and Biomarker-Driven Risk Stratification

5.1. Tissue-Based Genomic Classifiers

AI shows strong promise for predicting metastasis and prognosis based on biomarkers. One study built a prediction model combining CNNs and short-term memory neural networks.¹⁷⁵ Enabling the handling of three biological sequence problems: Prediction of subcellular localization, protein secondary structure and the binding of peptides to MHC Class II molecules. Another study identified a novel biomarker for PCa from proteomics by integrating supervised machine learning-based biomarker discovery with Boolean algebra-based signature derivation.¹⁷⁶ In addition, machine learning models leveraging amino acid metabolism-related genes have been used to predict Gleason score and link it to prognosis in prostate cacner.¹⁷⁷ A separate study developed a pathway-based model for predicting biochemical recurrence of PCa, enabling earlier risk stratification in patients.¹⁷⁸ Finally, Qiao et al analyzed PCa genomic data and identified MYLK as a novel, independent biomarker for BCR.¹⁷⁹

5.2. Predictive Biomarker Discovery

Biological phenotypes reflect underlying genomic sequences making gene analyses crucial in PCa diagnosis and prognosis. As illustrated in Figure 3, AI models that leverage molecular, cellular, exosomal and genetic features for computation can support prognosis and facilitate biomarker discovery. Because biological sequence analysis is labor-intensive AI is well-suited for high-dimensional analysis of transcriptomic, proteomic, and metabolomic datasets. A review of machine learning models based on next-generation sequencing (NGS) data highlights both the scale of NGS data and the promise of AI for extracting clinically relevant signals.³³ For gene-sequence analysis one study used patient DNA sequences to classify cancer types with machine learning.¹⁸⁰ Another developed a transcriptome-wide gene-expression prediction model to identify genes137 associated with PCa.¹⁸¹ Genes associated with PCa. A cross-cancer (CC) learning strategy further utilized breast cancer datasets to identify biomarkers for PCa due to the similarity between PCa and breast cancer DNA repair pathways.¹⁸² This approach revealed that ADIRF, SLC2A5, C3orf86, and HSPA1B from breast cancer biomarkers could be applied as indicators for clinical diagnosis of PCa. The architecture of this cross-learning model is shown in Figure 4.

Figure 3.

AI in PCa management increasingly leverages biomarker data to improve diagnosis and prognosis. Key biomarkers include PSA, circulating DNA, cellular markers, exosomes, and genetic profiles. Genetic analysis is central to patient stratification for tailored therapies. Integrating these datasets, AI supports personalized medicine, novel biomarker discovery, and improved prognostic accuracy

Figure 4.

A Cross-Learning Model leverages similarities in DNA repair mechanisms across ovarian, prostate, and breast cancers. It includes two parallel analysis paths: One AI-based and another statistical. In the AI path, a prediction model classifies tissue types and disease states, while an explanation module highlights each gene’s contribution. Model performance is assessed by comparing AI outcomes with statistical results based on area under the curve (AUC)

Beyond genes, downstream products such as RNA, protein and metabolites also show strong potential as biomarkers. Li et al developed a deep learning model that distinguished PCa from benign prostate hyperplasia using 6 protein variables.³⁴ Another deep learning model analyzed expressed prostatic secretion (EPS) from urine based on proteomics distinguish PCa from benign prostatic hyperplasia (BPH)¹⁸³ by sema7A and SPARC as high-performing markers. Another study highlighted the potential of metabolites as biomarkers, identifying 25 metabolites that could distinguish PCa tissues from normal tissues.¹⁸⁴ Extending this line of work, researchers have applied convolutional neural network (CNN) to urine metabolomic data to identify metabolite-based biomarkers for PCa.³⁵

6. Clinical Decision Support and Multimodal Integration

AI is increasingly embedded across the prostate cancer (PCa) management because clinicians must synthesize data from medical images that often contain clinically relevant patterns not readily visible to the human eye.¹⁸⁵ Deep learning in particular excels at handling large, high-dimensional datasets.¹⁵⁶ Thus, AI shows promise in assisting the diagnosis and prognosis process. In 2019, Strom et al reported the validation of an AI guided liquid biopsy approach underscoring AI’s potential in diagnostics.^186,187

Collectively, these advances position AI to support auto-segmentation, registration, identification of benign and malignant cancer, automated grading, early diagnosis, and prognosis prediction. Furthermore, these tasks are widely applied in the PCa diagnosis and treatment process, including radiotherapy, brachytherapy, active surveillance, surgery, and prognosis.

6.1. Typical AI-Assisted Tasks

Typical AI-assisted tasks in diagnosis encompass auto-segmentation, registration, identification and automatic grading system. Auto-segmentation is one the most common AI-assisted tasks, applied across diagnosis, treatment, monitoring and prognosis. Numerous automatic segmentation tools have been proposed over the years^188,189 but only recently have deep learning based CAD delivered stable, clinically relevant results.¹¹⁸ For example, AI has been used for prostate gland segmentation, lesion identification, and classification on mpMRI and TRUS-Bx.¹⁵³ Real-time TRUS auto-segmentation with deep learning has also been reported.²⁹ For PET imaging, a CNNs model was utilized by training on pre-operative¹⁸ F-choline (FCH) PET/CT scans from 45 newly diagnosed PCa patients prior to their RPa.¹⁵⁵ In a related work, Rubinstein et al combined hand-crafted perfusion-based features with deep autoencoder-based features generated from a dynamic¹¹ C-choline-PET/CT to perform both classification and localization.¹⁵⁶

Registration refers to aligning regions of interest (ROIs) between pretreatment and reference images. It offers utility, as tissues may shift between acquisitions.¹¹⁸ Deep convolutional neural networks have been deployed for registration on both TRUS images³⁰ and CT images.¹⁹⁰

AI also aids differentiation between benign and malignant tumors. In one study, a recurrent neural networks (RNNs) reported an area under the curve (AUC) of 0.99 on the test set and 0.93 on an independent external cohort of more than 12,000 slides.¹⁹¹ The STHLM3 framework further predicts the presence, extent and Gleason score of malignancy from population-based needle biopsy datasets.¹⁸⁶ Dov et al developed a model to identify high risk ROIs for pathologists’ further biopsy.³¹ This hybrid human-machine approach highlights top 20 ROIs with the highest malignancy probability in each biopsy, allowing pathologists to focus exclusively on these areas.

Grading systems are another common application of AI for diagnosis. As reported recently, AI grading system can perform at expert-level grading, surpassing non-uropathologists.¹⁹² Panda challenges have also demonstrated expert-level performance of AI in scoring.⁴⁸ Pantanowitz et al developed a grading model that showed promising performance in distinguishing between high- and low-grade PCa on an external validation set.³² Bulten et al created a deep-learning system to grade prostate biopsies following the Gleason grading standard, which can delineate individual glands, assign Gleason growth patterns, and determine biopsy-level grades.¹⁹³

6.2. Typical AI-Assisted Tasks in Treatment

Typical AI-assisted tasks in treatment include radiotherapy, brachytherapy, active surveillance, monitoring and prognosis. Radiotherapy needs to optimize dose distribution in consideration of both location and time, as well as target volume and organ at risk. AI can facilitate segmentation and prediction of dose distribution, actual delivered dose and treatment outcomes. Determining the optimal dose distribution across location and time is essential in radiotherapy. An important set of data to support clinical decision-making is the identification of target volume and organ at risk. Hence, AI-assisted auto-segmentation can provide valuable support.¹⁹⁴ Apart from segmentation tasks, AI has been widely applied to predict dose distribution in head and neck cancers as well as lung cancer.¹⁹⁵ For PCa treatment, one study used a deep learning approach to automate the calculation of dose distribution for PCa radiotherapy.¹⁹⁶ In terms of predicting the actual delivered dose, results from the study showed that the deep neural network achieved excellent agreement with treatment planning dose maps.¹⁹⁷ AI can also be applied to predict treatment outcomes, such as genitourinary toxicity¹⁹⁸ and incontinence.¹⁹⁹

Brachytherapy refers to a treatment in which radioactive sources are implanted into the prostate, allowing for localized radiation dosage.⁷³ Therefore, brachytherapy is a delicate procedure that requires consideration of many factors, in which context AI offers a valuable opportunity for assistance. AI can similarly be applied to the auto-segmentation task in brachytherapy.²⁰⁰ In addition, it can be used for the prediction of dosage plans. A machine learning algorithm developed for predicting source patterns has achieved clinical success, with performance comparable to expert-level quality and delivery within the required time.²⁰¹

Patients stratified as a low-risk cohort are suitable for active surveillance. AI has wide applications in active surveillance. For example, AI can help detect subtle changes in a patient’s condition. In addition, AI can assist in the acquisition and interpretation of images.²⁰² Currently, there are examples of non-AI platform being used to track and guide patients, such as the PASS Risk Calculator. Thus, AI shows great potential in building AI-assisted tracking platforms.

As for surgery, it is generally unrealistic to have AI perform the surgery, but it is reasonable to train AI to assist doctors. AI systems that display real-time visual information for lesion location, needle positioning, and measurements significantly improve procedural accuracy and efficiency. A mixed reality biopsy navigation system demonstrated a 53% reduction in procedure time and successful first-attempt completion in 70% of cases compared to only 20% without the system.²⁰³ Similarly, an AI navigation system for vacuum-assisted breast biopsy achieved superior real-time tracking performance, with mean average precision of 0.829 for tumor detection and 0.765 for needle tip tracking, substantially outperforming junior surgeons.²⁰⁴ Furthermore, AI can learn from real-time data and help adjust the procedure automatically.²⁰⁵

AI can standardize medical reports by automating documentation processes, which improves consistency, reduces documentation burden, and enhances efficiency in patient monitoring.^206,207 It can reduce reporting time, thereby saving time and labor. Moreover, AI can increase the consistency of reports, avoiding any inherent subjectivity of human assessment. With a higher level of standardization, it would be easier to build a complete and consistent database.

In terms of prognosis prediction, researchers have proposed several AI-assisted tools. Wulczyn et al developed an AI-based Gleason grading system to predict prostate cancer-specific mortality.²⁰⁸ However, as the Gleason score itself is inherently biased, the use of long-term follow-up data from AI is preferred. Another model combined clinical data and digital histopathology data from prostate biopsies to computationally predict patient specific outcomes and guide treatment decisions.²⁰⁹ Beyond biopsy analysis, AI models have also been developed to predict disease progression by using PET/CT images.¹⁵⁷

6.3. AI’s Potentials in Early Diagnosis

Apart from AI assistance in existing PCa treatment pathways, there is a potential for early diagnosis using routine imaging such as CT, though this area requires further research. While CT is invaluable for assessing bone and joint pathology, it is generally not recommended for cancer screening because of high radiation exposure. Nevertheless, incidental CT scans obtained for other indications may enable earlier PCa detection.⁴ An AI-based model analyzing such CT scans has proved promising for early detection of PCa.²¹⁰ In the accidental discovery of PCa, Patients with critical conditions who undergo CT scans may be simultaneously screened for prostate cancer, and early identification using this technology could significantly improve prognosis. Early diagnosis is therefore a crucial factor for patient outcomes. In cases of false positive predictions from CT-based machine learning models, patients may be referred for additional confirmatory examinations, such as laboratory tests or biopsies, to verify the diagnosis. In contrast, false negative predictions pose a greater clinical risk, as they may delay diagnosis and lead to disease progression. Consequently, these algorithms should be trained on diverse and representative CT datasets to minimize false negative rates and improve diagnostic reliability.⁴

6.4. AI Integration in Clinical Practice

IBM Watson for Oncology, introduced in the mid-2010s, was one of the pioneering AI-driven clinical decision support systems designed to assist oncologists in treatment planning by analyzing patient data against vast medical literature and guidelines.²¹¹ Despite initial promise, its adoption has been limited due to challenges such as poor integration with electronic health records, inconsistent recommendations, and a user interface that disrupted clinical workflows.²¹² Recent evaluations highlight ongoing issues, including the system’s inability to fully replace human expertise and the need for continuous learning from real-world data, as evidenced in reports showing variable concordance rates with expert opinions and instances of unsafe recommendations.^213,214 These hurdles have led to scaled-back implementations and questions about its long-term viability in oncology.

Building on early systems like Watson, more recent AI decision support tools have emerged to address oncology challenges with improved capabilities. For instance, ArteraAI, a multimodal AI platform for prostate cancer, uses machine learning to predict treatment responses and personalize therapy based on pathology and clinical data, showing strong validation in clinical trials as of 2025.²¹⁵ ASCO’s Guidelines Assistant, launched in recent years and powered by Google Cloud’s Gemini, provides real-time evidence-based recommendations to oncologists, integrating guidelines and patient-specific factors to enhance decision-making.²¹⁶ Additionally, an autonomous AI agent leveraging GPT-4, developed in 2025, combines multimodal precision oncology tools for treatment support, demonstrating high accuracy in simulating clinical workflows and predicting outcomes.²¹⁷ These alternatives focus on better data integration and user-centric design, offering a more balanced approach to clinical deployment.

Clinical decision support systems (CDSS) in oncology offer significant merits, including enhanced precision in treatment recommendations, reduced decision fatigue for clinicians, and improved patient outcomes through personalized care based on large-scale data analysis.²¹⁸ However, demerits persist, such as barriers to clinical adoption stemming from data privacy concerns, validation challenges in diverse populations, and integration issues with existing healthcare infrastructure, which can lead to skepticism and underutilization.²¹⁹ Overall, while these systems promise transformative impact, rigorous regulatory oversight and ongoing refinements are essential to balance their benefits against these limitations.

7. Limitations and Future Directions

7.1. Data Standardization

Two major hurdles limit standardization: (1) the scarcity of high-quality, open-source dataset and (2) difficulty establishing ground truth. The lack of datasets covering a wide range of scenarios hinders robust assessment by the AI model.³⁶ Additionally, inter-observer variability further introduces inconsistent expert annotations. Mitigation approaches include training on multi-expert-annotated datasets such as PANDA and resolving label disagreements with majority voting system, the STAPLE algorithm or automated label cleaning method.^{48,192,220,221} Furthermore, pre-segmentation has also shown to improve diagnostic accuracy.³⁶

7.2. Interpretability

Improving interpretability has attracted growing interest in AI-assisted PCa pathology. Techniques such as gradient-weighted class activation mapping introduced by Arvaniti et al visualize model attention and highlight ROI.³⁷ Experts review of AI outputs provides an additional validation layer. In fact, experts have found AI errors during review.²²² Automatic concept explanations (ACEs) method has also demonstrated clinical significance by elucidating the features for decision making.²²³ Moreover, AI models may help discover novel features strongly associated with disease development.¹³⁰ Lastly, it is considered beneficial to summarize common errors and advantages of AI models.²²⁴ One research even provided a quantification of AI uncertainty.²²⁵

7.3. Bias and Fairness

Expanding digital pathology infrastructure is essential to address bias and improve fairness. Regional medical centers should prioritize slide digitization and the adoption of interoperable platforms. In resources limited settings, AI-augmented workflows can help triage rare or challenging cases and support clinicians.¹³⁰ In the meantime, computationally enabled microscopes can bridge the gap until whole-slide imaging becomes broadly available. A microscope overlays AI information has shown promising results in AI-assisted real-time computational imaging tools.³⁸

7.4. Validation

Validation of AI for PCa pathology remains challenging. Key issues include intrinsic model limitations of AI and its influence on human decision-making, limited generalizability, and performance degradation in rare cases, tumor heterogeneity, and genetic variability. AI systems can nudge human judgements, sometimes even when the AI is inaccurate, raising concerns about automation bias.²²⁶ False-positive alerts are often treated as “useful warnings,” but they can increase cognitive load and time on tasks.

In addition, generalization is hindered by heterogeneity at multiple levels, ranging from biologic variation in PCa, to variability in datasets. To improve robustness, training on large, diverse datasets is recommended.¹³⁰ Complementary methods include data augmentation²²⁰ and color enhancement,²²⁷ deliberately inclusion of artificial artifacts to harden models²²⁸ or samples from external dataset,²²⁹ color modification²³⁰ and style normalization²³¹ during training and validation.

7.5. AI Regulations

Integration of imaging, liquid biopsy, and biomarker data via AI enables more precise risk stratification, early detection, and personalized therapy in metastatic prostate cancer. Multimodal models combining radiomics, genomics, and clinical data have demonstrated improved diagnostic accuracy, reduced inter-reader variability, and enhanced prediction of treatment response.^232,233 AI-driven data integration supports non-invasive monitoring and may facilitate earlier intervention, potentially improving survival and quality of life.^{41,58,140,234,235} However, challenges remain in harmonizing data sources, ensuring generalizability, and validating clinical utility in prospective trials.^43,137,236

AI adoption in prostate cancer care is subject to evolving regulatory frameworks, including FDA clearance for imaging tools and compliance with the General Data Protection Regulation (GDPR) for data privacy.^136,236 Ethical governance requires transparency, explainability, and robust data stewardship to address the “black box” nature of some AI models and ensure patient autonomy.^233,235,237 Multicenter collaborations and standardized reporting are essential for equitable, accountable clinical integration.^140,233,235

7.6. Rare Cases

There are numerous rare cases to consider in real-world applications. These include: inflammation, atrophy, atypical small acinar proliferation, atypical intraductal proliferation, and some diagnostically challenging histological subtypes (i.e.ductal and intraductal carcinoma of the prostate, prostatic adenocarcinoma with neuroendocrine differentiation), and treatment-related changes.^{26,31,48,224,238-241} These cases pose significant challenges to the validation of AI applications in PCa. To deal with complex real-world situations, researchers have introduced generative adversarial network (GAN) to synthesize high-fidelity pathological images.^205,242 For instance, Falahkheirkhah et al combined synthetic images generated by GAN with real images to train a model that classifies prostate tissues.²⁴³ In most cases, experts only select representative slides per patient for training. However, this approach fails to capture the full spectrum of tumor heterogeneity. Therefore, it is preferable to incorporate data from all available slides of the patient and compare them with the most representative slides.¹³⁰ Regarding genetic diversity among different populations, some groups exhibit susceptibility to PCa. For example, FOXA1 mutations are more often observed in Asian populations compared to European and American populations.⁴⁷ Thus, it is prudent to validate the model’s applicability in various populations. However, most of the studies are still limited to populations from Western countries.⁴⁸

8. Perspectives

There are many remaining problems in AI-assisted PCa pathology. AI has potential in PSA ancillary task such as fPSA percentage, PHI and PCA3. For now, AI-assisted tools in PET images have not yet attracted much interest. In addition, most studies on AI-assisted tools on biomarkers have focused on genetic biomarkers. Alternative biomarkers, such as cellular biomarkers, still warrant further research. Multimodal models that integrate imaging, pathology, genomics, and real-world data by using deep neural networks or graph-based learning in PCa also merit attention. Finally, it is highly advisable to establish a standardized and efficient protocol for digitalizing patient data. Furthermore, developing validated pre-segmentation AI models before human assessment has clinical significance.

Current limitations include a lack of ground truth dataset and limited regulatory approval of AI-assisted tools. A shortage in ground truth datasets hinders the training and evaluation of AI model. Only AI tools that have received official medical approval can be applied in clinical settings. Even though several AI models have gained FDA approval, the number of officially approved models remains inadequate to meet the demands of medical application.

Current challenges include the validation and interpretability of AI models. The validation of AI models is a key issue, as misdiagnosis can lead to mistreatment and potentially avoidable deaths. AI models are often regarded as “black box”, which are inherently associated with low interpretability.

9. Conclusion

AI represents a transformative force in metastatic PCa care. Advances in AI-driven imaging tools can enhance early detection, monitor disease progression, and assess therapeutic response with greater accuracy and consistency. Similarly, AI-assisted biomarker discovery and interpretation facilitate the identification of novel prognostic and predictive indicators, supporting more individualized treatment strategies. By augmenting imaging interpretation, refining biomarker analysis, and enabling integrated clinical decision-making, AI supports precision medicine. However, robust validation, interpretability, bias and fairness, and data standardization are essential for safe and equitable implementation.

Footnotes

Acknowledgements

The authors would like to thank Dr. Frieda Law for her support during the project.

ORCID iDs

Luyuan Li

R. Daniel Bonfil

Wensi Tao

Ethical Considerations

This study does not involve work that entails any ethics approval.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was graciously supported by grants from the National Natural Science Foundation of China (32073002, 32471193), Li Ka-Shing Foundation at Shantou University Medical College (510858044), and Li Ka-Shing Foundation STU-GTIIT Joint-research Grant (2024LKSFG08).

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Sung

Ferlay

Siegel

, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: A Cancer Journal for Clinicians. 2021;71(3):209-249.

Siegel

Miller

Jemal

. Cancer statistics, 2018. CA: A Cancer Journal for Clinicians. 2018;68(1):7-30.

Ferlay

Colombet

Soerjomataram

, et al. Cancer incidence and mortality patterns in Europe: Estimates for 40 countries and 25 major cancers in 2018. European Journal of Cancer. 2018;103:356-387.

Rabaan

Bakhrebah

AlSaihati

, et al. Artificial intelligence for clinical diagnosis and treatment of prostate cancer. Cancers. 2022;14(22):5595.

Dess

Hartman

Mahal

, et al. Association of black race with prostate cancer–specific and other-cause mortality. JAMA Oncology. 2019;5(7):975-983.

Lorenzoni

Chaturvedi

Vignat

Laversanne

Bray

Vaccarella

. The current burden of oropharyngeal cancer: a global assessment based on GLOBOCAN 2020. Cancer Epidemiology, Biomarkers & Prevention. 2022;31(11):2054-2062.

Gallagher

Gaudet

Pal

, et al. Germline BRCA mutations denote a clinicopathologic subset of prostate cancer. Clinical Cancer Research. 2010;16(7):2115-2121.

Bruner

Moore

Parlanti

Dorgan

Engstrom

. Relative risk of prostate cancer for men with affected relatives: systematic review and meta‐analysis. International Journal of Cancer. 2003;107(5):797-803.

Plaskon

Penson

Vaughan

Stanford

. Cigarette smoking and risk of prostate cancer in middle-aged men. Cancer Epidemiology Biomarkers & Prevention. 2003;12(7):604-609.

10.

Allott

Masko

Freedland

. Obesity and prostate cancer: weighing the evidence. European Urology. 2013;63(5):800-809.

11.

Moura

Sampaio

Kobayashi

, et al. Structural and ultrastructural morphological evaluation of Giant anteater (Myrmecophaga tridactyla) prostate gland. Biology. 2021;10(3):231.

12.

Farashi

Kryza

Clements

Batra

. Post-GWAS in prostate cancer: from genetic association to biological contribution. Nature Reviews Cancer. 2019;19(1):46-59.

13.

Van Loenhout

Zijta

Smithuis

Schoots

Radiology Department of the Haaglanden MC . Prostate Cancer—PI-RADS v2. The Netherland: Alrijne Hospital and Erasmus MC; 2018.

14.

Clements

. The role of transrectal ultrasound in diagnosing prostate cancer. Current Urology Reports. 2002;3(3):194-200.

15.

Kartasalo

Bulten

Delahunt

, et al. Artificial intelligence for diagnosis and Gleason grading of prostate cancer in biopsies—current status and next steps. European Urology Focus. 2021;7(4):687-691.

16.

Jin

Fei

Wang

Song

Chen

. Detection and Prognosis of Prostate Cancer Using Blood‐Based Biomarkers. Mediators of Inflammation. 2020;2020(1):8730608.

17.

Martignano

Rossi

Maugeri

, et al. Urinary RNA-based biomarkers for prostate cancer detection. Clinica Chimica Acta. 2017;473:96-105.

18.

Filella

Foj

. Prostate cancer detection and prognosis: from prostate specific antigen (PSA) to exosomal biomarkers. International Journal of Molecular Sciences. 2016;17(11):1784.

19.

Stephan

Cammann

Meyer

H-A

Lein

Jung

. PSA and new biomarkers within multivariate models to improve early detection of prostate cancer. Cancer Letters. 2007;249(1):18-29.

20.

Hosny

Parmar

Quackenbush

Schwartz

Aerts

. Artificial intelligence in radiology. Nature Reviews Cancer. 2018;18(8):500-510.

21.

Acs

Rantalainen

Hartman

. Artificial intelligence as the next step towards precision pathology. Journal of Internal Medicine. 2020;288(1):62-81.

22.

Zhu

Mou

Chen

. Can the ChatGPT and other large language models with internet-connected database solve the questions and concerns of patient with prostate cancer and help democratize medical knowledge? Journal of Translational Medicine. 2023;21(1):269.

23.

Lee

Sung

Kim

C-S

, et al. Three-dimensional convolutional neural network for prostate MRI segmentation and comparison of prostate volume measurements by use of artificial neural network and ellipsoid formula. American Journal of Roentgenology. 2020;214(6):1229-1238.

24.

Lay

Tsehay

Greer

, et al. Detection of prostate cancer in multiparametric MRI using random forest with instance weighting. Journal of Medical Imaging. 2017;4(2):024506.

25.

Sanford

Harmon

Turkbey

, et al. Deep‐learning‐based artificial intelligence for PI‐RADS classification to assist multiparametric prostate MRI interpretation: a development study. Journal of Magnetic Resonance Imaging. 2020;52(5):1499-1507.

26.

Egevad

Delahunt

Samaratunga

, et al. The emerging role of artificial intelligence in the reporting of prostate pathology. Pathology. 2021;53(5):565-567.

27.

Carroll

Parsons

Andriole

, et al. NCCN guidelines insights: prostate cancer early detection, version 2.2016. Journal of the National Comprehensive Cancer Network. 2016;14(5):509-519.

28.

Cheng

Bostwick

. Urologic surgical pathology E-book; 2008.

29.

Van Sloun

RJG

Wildeboer

Mannaerts

, et al. Deep learning for real-time, automatic, and scanner-adapted prostate (zone) segmentation of transrectal ultrasound, for example, magnetic resonance imaging–transrectal ultrasound fusion prostate biopsy. European Urology Focus. 2021;7(1):78-85.

30.

Zhu

Najafi

Han

Hancock

Hristov

. Feasibility of image registration for ultrasound-guided prostate radiotherapy based on similarity measurement by a convolutional neural network. Technology in Cancer Research & Treatment. 2019;18:1533033818821964.

31.

Dov

Assaad

Syedibrahim

, et al. A hybrid human–machine learning approach for screening prostate biopsies can improve clinical efficiency without compromising diagnostic accuracy. Archives of Pathology & Laboratory Medicine. 2022;146(6):727-734.

32.

Pantanowitz

Quiroga-Garza

Bien

, et al. An artificial intelligence algorithm for prostate cancer diagnosis in whole slide images of core needle biopsies: a blinded clinical validation and deployment study. The Lancet Digital Health. 2020;2(8):e407-e416.

33.

Park

Heider

Hauschild

A-C

. Integrative analysis of next-generation sequencing for next-generation cancer research toward artificial intelligence. Cancers. 2021;13(13):3148.

34.

Gabriele

Aracri

Prestagiacomo

, et al. Development of a predictive model to distinguish prostate cancer from benign prostatic hyperplasia by integrating serum glycoproteomics and clinical variables. Clinical Proteomics. 2023;20(1):52.

35.

Chen

Zhang

Yang

, et al. Raman spectroscopy reveals abnormal changes in the urine composition of prostate cancer: an application of an intelligent diagnostic model with a deep learning algorithm. Advanced Intelligent Systems. 2021;3(4):2000090.

36.

van Leenders

GJLH

van der Kwast

Grignon

, et al. The 2019 International Society of Urological Pathology (ISUP) consensus conference on grading of prostatic carcinoma. The American Journal of Surgical Pathology. 2020;44(8):e87-e99.

37.

Arvaniti

Fricker

Moret

, et al. Automated Gleason grading of prostate cancer tissue microarrays via deep learning. Scientific reports. 2018;8(1):12054.

38.

Chen

P-HC

Gadepalli

MacDonald

, et al. An augmented reality microscope with real-time artificial intelligence integration for cancer diagnosis. Nature Medicine. 2019;25(9):1453-1457.

39.

Thomas

. Automated systems comparable to expert pathologists for prostate cancer Gleason grading. Nature Reviews Urology. 2020;17(3):131.

40.

Xie

X-R

Hou

Shan

, et al. AI-aided diagnostic performance for prostate MRI: Systematic review and meta-analysis. Prostate Cancer and Prostatic Diseases. 2025; 1-10.

41.

Ciccone

Garofano

Del Sorbo

, et al. Improving early prostate cancer detection through artificial intelligence: Evidence from a systematic review. Cancers. 2025;17(21):3503.

42.

Liu

Cundy

Woon

DTS

Lawrentschuk

. A systematic review on artificial intelligence evaluating metastatic prostatic cancer and lymph nodes on PSMA pet scans. Cancers. 2024;16(3):486.

43.

Kendrick

Francis

Hassan

, et al. Radiomics for identification and prediction in metastatic prostate cancer: a review of studies. Frontiers in oncology. 2021;11:771787.

44.

Barsouk

Padala

Vakiti

, et al. Epidemiology, staging and management of prostate cancer. Medical sciences. 2020;8(3):28.

45.

Punnen

Cooperberg

D’Amico

, et al. Management of biochemical recurrence after primary treatment of prostate cancer: a systematic review of the literature. Eur Urol. 2013;64(6):905-915. doi:10.1016/j.eururo.2013.05.025.

46.

Hotte

Saad

. Current management of castrate-resistant prostate cancer. Current Oncology. 2010;17(Suppl 2):S72-S79.

47.

Lee

, et al. A genomic and epigenomic atlas of prostate cancer in Asian populations. Nature. 2020;580(7801):93-99.

48.

Bulten

Kartasalo

Chen

P-HC

, et al. Artificial intelligence for diagnosis and Gleason grading of prostate cancer: the PANDA challenge. Nature Medicine. 2022;28(1):154-163.

49.

Wong

ANN

Leung

, et al. Current developments of artificial intelligence in digital pathology and its future clinical applications in gastrointestinal cancers. Cancers. 2022;14(15):3780.

50.

Pohjonen

Stürenberg

Rannikko

Mirtti

Pitkänen

. Spectral decoupling for training transferable neural networks in medical imaging. Iscience. 2022;25(2):103767.

51.

Laurinavicius

Rasmusson

Plancoulaine

Shribak

Levenson

. Machine-learning-based evaluation of intratumoral heterogeneity and tumor-stroma interface for clinical guidance. Am J Pathol. 2021;191(10):1724-1731. doi:10.1016/j.ajpath.2021.04.008.

52.

Zhai

Yang

. Quantification of intratumoral heterogeneity using habitat-based MRI radiomics for predicting high-Gleason scores and castration-resistant PCa: Retrospective study. BMC Cancer. 2025;25(1):1927. doi:10.1186/s12885-025-15300-8.

53.

Yang

Zheng

Zou

, et al. MRI radiomics and automated habitat analysis enhance machine learning prediction of bone metastasis and high-grade gleason scores in postate cancer. Acad Radiol. 2025;32(9):5303-5316. doi:10.1016/j.acra.2025.05.059.

54.

Barnett

Woo

Perk

, et al. Automated analysis of FDG-PET/CT imaging to monitor heterogeneous disease response in metastatic castration-resistant prostate cancer. American Society of Clinical Oncology; 2023.

55.

van Hees

Vlaming

Flach

, et al. AI-based prediction of molecular aberrations in prostate cancer using digital pathology: A systematic review. Crit Rev Oncol Hematol. 2026;217:105011. doi:10.1016/j.critrevonc.2025.105011.

56.

Bazarkin

Taratkin

Vovdenko

, et al.

Artificial intelligence in diagnostic, prognostic, and predictive genomic biomarkers for prostate cancer: Ready for prime time?

Urol Oncol. 2026;44(3):110965. doi:10.1016/j.urolonc.2025.12.001.

57.

Cavinato

Sollini

Ragni

, et al. Radiomics-based inter-lesion relation network to describe [(18)F]FMCH PET/CT imaging phenotypes in prostate cancer. Cancers (Basel). 2023;15(3):823. doi:10.3390/cancers15030823.

58.

Vidiyala

Parupathi

Sunkishala

, et al. Artificial intelligence: a new era in prostate cancer diagnosis and treatment. International Journal of Pharmaceutics. 2025;683:126024.

59.

Zamboglou

Doncker

Christoforou

, et al. oDigital pathology biomarkers for guiding radiotherapy-based treatment concepts in prostate cancer - a systematic review and expert consensus. Radiother Oncol. 2025;210:111039. doi:10.1016/j.radonc.2025.111039.

60.

Gallo

Stoyanova

, et al. Diagnosis and treatment of metastatic prostate cancer: the role of imaging, liquid biopsies, and biomarkers for deciphering tumor heterogeneity. In: Unraveling the Complexities of Metastasis. Elsevier; 2022:23-47.

61.

Bjurlin

Carter

Schellhammer

, et al. Optimization of initial prostate biopsy in clinical practice: sampling, labeling and specimen processing. The Journal of Urology. 2013;189(6):2039-2046.

62.

Bott

SRJ

Young

MPA

Kellett

Parkinson

Contributors to the UCL Hospitals' Trust Radical Prostatectomy Database . Anterior prostate cancer: is it more difficult to diagnose? BJU International. 2002;89(9):886-889.

63.

Boudadi

Antonarakis

. Resistance to novel antiandrogen therapies in metastatic castration-resistant prostate cancer. Clinical Medicine Insights: Oncology. 2016;10:CMO. Ss34534.

64.

Chang

Y-CC

Ackerstaff

Tschudi

, et al. Delineation of tumor habitats based on dynamic contrast enhanced MRI. Scientific Reports. 2017;7(1):9746.

65.

Cheever

Higano

. PROVENGE (Sipuleucel-T) in prostate cancer: the first FDA-approved therapeutic cancer vaccine. Clinical Cancer Research. 2011;17(11):3520-3526.

66.

De Marzo

Marchi

Epstein

Nelson

. Proliferative inflammatory atrophy of the prostate: implications for prostatic carcinogenesis. The American Journal of Pathology. 1999;155(6):1985-1992.

67.

Reiter

Potocki

Chien

Gribskov

Bier

. A systematic analysis of human disease-associated gene sequences in Drosophila melanogaster. Genome Research. 2001;11(6):1114-1125.

68.

Choyke

Loeb

. Active surveillance of prostate cancer. Oncology (Williston Park, NY). 2017;31(1):67-70.

69.

Sanda

Cadeddu

Kirkby

, et al. Clinically localized prostate cancer: AUA/ASTRO/SUO guideline. Part I: risk stratification, shared decision making, and care options. The Journal of Urology. 2018;199(3):683-690.

70.

Duffy

. Biomarkers for prostate cancer: prostate-specific antigen and beyond. Clinical Chemistry and Laboratory Medicine (CCLM). 2020;58(3):326-339.

71.

Tikkinen

KAO

Dahm

Lytvyn

, et al. Prostate cancer screening with prostate-specific antigen (PSA) test: a clinical practice guideline. The BMJ. 2018;362:k3581.

72.

Catalona

Richie

Ahmann

, et al. Comparison of digital rectal examination and serum prostate specific antigen in the early detection of prostate cancer: results of a multicenter clinical trial of 6,630 men. The Journal of Urology. 1994;151(5):1283-1290.

73.

Six

Veldhuis

Akin

. The ultimate guide to AI in prostate cancer.

74.

Wang

Tang

, et al. Advances in prostate cancer biomarkers and probes. Cyborg and Bionic Systems. 2024;5:0129.

75.

Barbieri

Baca

Lawrence

, et al. Exome sequencing identifies recurrent SPOP, FOXA1 and MED12 mutations in prostate cancer. Nature genetics. 2012;44(6):685-689.

76.

Tomlins

Rhodes

Perner

, et al. Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science. 2005;310(5748):644-648.

77.

Sommariva

Tarricone

Lazzeri

Ricciardi

Montorsi

. Prognostic value of the cell cycle progression score in patients with prostate cancer: a systematic review and meta-analysis. European Urology. 2016;69(1):107-115.

78.

Lowrance

Eastham

Savage

, et al. Contemporary open and robotic radical prostatectomy practice patterns among urologists in the United States. The Journal of Urology. 2012;187(6):2087-2093.

79.

Rashkovska

Kocev

Trobec

. Non-invasive real-time prediction of inner knee temperatures during therapeutic cooling. Computer Methods and Programs in Biomedicine. 2015;122(2):136-148.

80.

Han

Johnson

Gaed

, et al. Histologic tissue components provide major cues for machine learning-based prostate cancer detection and grading on prostatectomy specimens. Scientific Reports. 2020;10(1):9911.

81.

Eichler

Hempel

Wilby

Myers

Bachmann

Kleijnen

. Diagnostic value of systematic biopsy methods in the investigation of prostate cancer: a systematic review. The Journal of Urology. 2006;175(5):1605-1612.

82.

Cooper

Chay

Pienta

. The role of αvβ3 in prostate cancer progression. Neoplasia. 2002;4(3):191-194.

83.

Danila

Heller

Gignac

, et al. Circulating tumor cell number and prognosis in progressive castration-resistant prostate cancer. Clinical Cancer Research. 2007;13(23):7053-7058.

84.

Potosky

Davis

Hoffman

, et al. Five-year outcomes after prostatectomy or radiotherapy for prostate cancer: the prostate cancer outcomes study. J Natl Cancer Inst. 2004;96(18):1358-1367. doi:10.1093/jnci/djh259.

85.

Gravis

Boher

J-M

Joly

, et al. Androgen deprivation therapy (ADT) plus docetaxel versus ADT alone in metastatic non castrate prostate cancer: impact of metastatic burden and long-term survival analysis of the randomized phase 3 GETUG-AFU15 trial. European Urology. 2016;70(2):256-262.

86.

Hofman

Emmett

Violet

, et al. TheraP: a randomized phase 2 trial of 177Lu‐PSMA‐617 theranostic treatment vs cabazitaxel in progressive metastatic castration‐resistant prostate cancer (Clinical Trial Protocol ANZUP 1603). BJU International. 2019;124:5-13.

87.

Klein

Cooperberg

Magi-Galluzzi

, et al. A 17-gene assay to predict prostate cancer aggressiveness in the context of Gleason grade heterogeneity, tumor multifocality, and biopsy undersampling. European Urology. 2014;66(3):550-560.

88.

Klein

Yousefi

Haddad

, et al. A genomic classifier improves prediction of metastatic disease within 5 years after surgery in node-negative high-risk prostate cancer patients managed by radical prostatectomy without adjuvant therapy. European Urology. 2015;67(4):778-786.

89.

Lavery

Cooperberg

. Clinically localized prostate cancer in 2017: A review of comparative effectiveness. Elsevier; 2017:40-41.

90.

Loeb

Bjurlin

Nicholson

, et al. Overdiagnosis and overtreatment of prostate cancer. European Urology. 2014;65(6):1046-1055.

91.

Lu-Yao

Albertsen

Moore

, et al. Fifteen-year survival outcomes following primary androgen-deprivation therapy for localized prostate cancer. JAMA Internal Medicine. 2014;174(9):1460-1467.

92.

Mateo

Carreira

Sandhu

, et al. DNA-repair defects and olaparib in metastatic prostate cancer. New England Journal of Medicine. 2015;373(18):1697-1708.

93.

Nguyen

Haddad

Ross

, et al. Ability of a genomic classifier to predict metastasis and prostate cancer-specific mortality after radiation or surgery based on needle biopsy specimens. European Urology. 2017;72(5):845-852.

94.

Parker

Nilsson

Heinrich

, et al. Alpha emitter radium-223 and survival in metastatic prostate cancer. New England Journal of Medicine. 2013;369(3):213-223.

95.

Klotz

Zhang

Lam

Nam

Mamedov

Loblaw

. Clinical results of long-term follow-up of a large, active surveillance cohort with localized prostate cancer. Journal of Clinical Oncology. 2010;28(1):126-131.

96.

Selvadurai

Singhera

Thomas

, et al. Medium-term outcomes of active surveillance for localised prostate cancer. European Urology. 2013;64(6):981-987.

97.

Armstrong

Szmulewitz

Petrylak

, et al. ARCHES: a randomized, phase III study of androgen deprivation therapy with enzalutamide or placebo in men with metastatic hormone-sensitive prostate cancer. Journal of Clinical Oncology. 2019;37(32):2974-2986.

98.

Kyriakopoulos

Chen

Y-H

Carducci

, et al. Chemohormonal therapy in metastatic hormone-sensitive prostate cancer: long-term survival analysis of the randomized phase III E3805 CHAARTED trial. Journal of Clinical Oncology. 2018;36(11):1080-1087.

99.

Bill-Axelson

Holmberg

Garmo

, et al. Radical prostatectomy or watchful waiting in early prostate cancer. New England Journal of Medicine. 2014;370(10):932-942.

100.

Widmark

Klepp

Solberg

, et al. Endocrine treatment, with or without radiotherapy, in locally advanced prostate cancer (SPCG-7/SFUO-3): an open randomised phase III trial. The Lancet. 2009;373(9660):301-308.

101.

Yang

, et al. Molecular classifications of prostate cancer: basis for individualized risk stratification and precision therapy. Annals of Medicine. 2023;55(2):2279235.

102.

Chen

Y-x

Tan

L-m

Gong

J-p

, et al. Response prediction biomarkers and drug combinations of PARP inhibitors in prostate cancer. Acta Pharmacologica Sinica. 2021;42(12):1970-1980.

103.

Wang

Dehm

Hillman

, et al. A prospective genome-wide study of prostate cancer metastases reveals association of wnt pathway activation and increased cell cycle proliferation with primary resistance to abiraterone acetate–prednisone. Annals of Oncology. 2018;29(2):352-360.

104.

Shih

T-C

Liu

C-T

, et al. Targeting galectin-1 impairs castration-resistant prostate cancer progression and invasion. Clinical Cancer Research. 2018;24(17):4319-4331.

105.

Kasivisvanathan

Rannikko

Borghi

, et al. MRI-targeted or standard biopsy for prostate-cancer diagnosis. New England Journal of Medicine. 2018;378(19):1767-1777.

106.

Rouvière

Puech

Renard-Penna

, et al. Use of prostate systematic and targeted biopsy on the basis of multiparametric MRI in biopsy-naive patients (MRI-FIRST): a prospective, multicentre, paired diagnostic study. The Lancet Oncology. 2019;20(1):100-109.

107.

van der Leest

Cornel

Israël

, et al. Head-to-head comparison of transrectal ultrasound-guided prostate biopsy versus multiparametric prostate resonance imaging with subsequent magnetic resonance-guided biopsy in biopsy-naïve men with elevated prostate-specific antigen: a large prospective multicenter clinical study. European Urology. 2019;75(4):570-578.

108.

Mottet

Bellmunt

Bolla

, et al. EAU-ESTRO-SIOG guidelines on prostate cancer. Part 1: screening, diagnosis, and local treatment with curative intent. European Urology. 2017;71(4):618-629.

109.

Grover

VPB

Tognarelli

Crossey

MME

Cox

Taylor-Robinson

McPhail

MJW

. Magnetic resonance imaging: principles and techniques: lessons for clinicians. Journal of Clinical and Experimental Hepatology. 2015;5(3):246-255.

110.

Carroll

Coakley

Kurhanewicz

. Magnetic resonance imaging and spectroscopy of prostate cancer. Reviews in Urology. 2006;8(Suppl 1):S4.

111.

Akin

Sala

Moskowitz

, et al. Transition zone prostate cancers: features, detection, localization, and staging at endorectal MR imaging. Radiology. 2006;239(3):784-792.

112.

Jung

Donati

Vargas

Goldman

Hricak

Akin

. Transition zone prostate cancer: incremental value of diffusion-weighted endorectal MR imaging in tumor detection and assessment of aggressiveness. Radiology. 2013;269(2):493-503.

113.

Yoo

Kim

Jeong

. Multiparametric magnetic resonance imaging for prostate cancer: A review and update for urologists. Korean Journal of Urology. 2015;56(7):487-497.

114.

Maurer

Heverhagen

. Diffusion weighted imaging of the prostate—principles, application, and advances. Translational Andrology and Urology. 2017;6(3):490-498.

115.

Metens

Miranda

Absil

Matos

. What is the optimal b value in diffusion-weighted MR imaging to depict prostate cancer at 3T? European Radiology. 2012;22:703-709.

116.

Kurhanewicz

Vigneron

Hricak

Narayan

Carroll

Nelson

. Three-dimensional H-1 MR spectroscopic imaging of the in situ human prostate with high (0.24-0.7-cm3) spatial resolution. Radiology. 1996;198(3):795-805.

117.

Mazaheri

Shukla-Dave

Hricak

, et al. Prostate Cancer: Identification with Combined Diffusion-weighted MR Imaging and 3D 1H MR Spectroscopic Imaging—Correlation with Pathologic Findings1. Radiology. 2008;246(2):480-488.

118.

Wildeboer

van Sloun

Wijkstra

Mischi

. Artificial intelligence in multiparametric prostate cancer imaging with focus on deep-learning methods. Computer Methods and Programs in Biomedicine. 2020;189:105316.

119.

Verma

Turkbey

Muradyan

, et al. Overview of dynamic contrast-enhanced MRI in prostate cancer diagnosis and management. American Journal of Roentgenology. 2012;198(6):1277-1288.

120.

Watanabe

Kato

Morita

Tanaka

Terasawa

. Diagnostic application of the ultrasonotomography for the prostate. The Japanese Journal of Urology. 1968;59(4):273-279.

121.

Cortes

Vapnik

. Support-vector networks. Machine Learning. 1995;20:273-297.

122.

Altman

. An introduction to kernel and nearest-neighbor nonparametric regression. The American Statistician. 1992;46(3):175-185.

123.

Cybenko

. Approximation by superpositions of a sigmoidal function. Mathematics of Control, Signals and Systems. 1989;2(4):303-314.

124.

Hornik

. Approximation capabilities of multilayer feedforward networks. Neural Networks. 1991;4(2):251-257.

125.

Schmidhuber

. Deep learning in neural networks: An overview. Neural Networks. 2015;61:85-117.

126.

Litjens

Kooi

Bejnordi

, et al. A survey on deep learning in medical image analysis. Medical Image Analysis. 2017;42:60-88.

127.

Ozkan

Eruyar

Cebeci

Memik

Ozcan

Kuskonmaz

. Interobserver variability in Gleason histological grading of prostate cancer. Scandinavian Journal of Urology. 2016;50(6):420-424.

128.

Goodman

Ward

Osunkoya

, et al. Frequency and determinants of disagreement and error in Gleason scores: a population‐based study of prostate cancer. The Prostate. 2012;72(13):1389-1398.

129.

Liu

JTC

Glaser

Bera

, et al. Harnessing non-destructive 3D pathology. Nature Biomedical Engineering. 2021;5(3):203-218.

130.

Zhu

Pan

Mou

, et al. Harnessing artificial intelligence for prostate cancer management. Cell Reports Medicine. 2024.

131.

Pietzsch

Saalfeld

Preibisch

Tomancak

. BigDataViewer: visualization and processing for large image data sets. Nature Methods. 2015;12(6):481-483.

132.

Rueden

Schindelin

Hiner

, et al. ImageJ2: ImageJ for the next generation of scientific image data. BMC Bioinformatics. 2017;18:1-26.

133.

Reder

Glaser

McCarty

Chen

True

Liu

JTC

. Open-top light-sheet microscopy image atlas of prostate core needle biopsies. Archives of Pathology & Laboratory Medicine. 2019;143(9):1069-1075.

134.

Edelstein

Amodaj

Hoover

Vale

Stuurman

. Computer control of microscopes using µManager. Current Protocols in Molecular Biology. 2010;92(1):1-17.

135.

Bashashati

Goldenberg

. AI for prostate cancer diagnosis—hype or today’s reality? Nature Reviews Urology. 2022;19(5):261-262.

136.

Tiwari

Mishra

Kuo

T-R

. Current AI technologies in cancer diagnostics and treatment. Molecular Cancer. 2025;24(1):159.

137.

Chaddad

Tan

Liang

, et al. Advancements in MRI-based radiomics and artificial intelligence for prostate cancer: A comprehensive review and future prospects. Cancers. 2023;15(15):3839.

138.

Valizadeh

Morafegh

Fatemi

Ghafoori

Saligheh Rad

. Enhancing postate cancer classification: A comprehensive review of multiparametric MRI and deep learning integration. Journal of Magnetic Resonance Imaging. 2025.

139.

Zhang

, et al. Artificial intelligence in prostate cancer. Chin Med J (Engl). 2025;138(15):1769-1782. doi:10.1097/cm9.0000000000003689.

140.

Baydoun

Jia

Zaorsky

, et al. Artificial intelligence applications in prostate cancer. Prostate Cancer and Prostatic Diseases. 2024;27(1):37-45.

141.

Zhuang

, et al. A multiparametric MRI and baseline-clinical-feature-based dense multimodal fusion artificial intelligence (MFAI) model to predict castration-resistant prostate cancer progression. Cancers (Basel). 2025;17(9):1556. doi:10.3390/cancers17091556.

142.

Zhou

Zhang

Guo

, et al. Multimodal data integration for predicting progression risk in castration-resistant prostate cancer using deep learning: A multicenter retrospective study. Front Oncol. 2024;14:1287995. doi:10.3389/fonc.2024.1287995.

143.

Tun

Naing

Malik

Rahman

. Artificial Intelligence (AI)-based tools in the diagnosis and management of prostate cancer: A systematic review and meta-analysis. Prostate Cancer Prostatic Dis. 2025. doi:10.1038/s41391-025-01060-w.

144.

Zhang

Xiao

Liang

, et al. Artificial intelligence-driven prostate cancer diagnosis: Enhancing accuracy and personalizing patient care. Urol Oncol. 2026;44(3):110959. doi:10.1016/j.urolonc.2025.11.013.

145.

Rajih

Bakhsh

Borhan

Alqahtani

SAM

. Utilization of artificial intelligence in prostate cancer detection: A comprehensive review of innovations in screening and diagnosis. Front Immunol. 2025;16:1670671. doi:10.3389/fimmu.2025.1670671.

146.

Wang

Zhong

Fang

Huang

. Application and prospect of artificial intelligence in diagnostic imaging of prostate cancer. NPJ Digit Med. 2026;9(1):168. doi:10.1038/s41746-026-02354-6.

147.

Lim

Yoo

Lee

, et al. Toward precision medicine: Development and validation of a machine learning based decision support system for optimal sequencing in castration-resistant prostate cancer. Clin Genitourin Cancer. 2023;21(4):e211-e218.e4. doi:10.1016/j.clgc.2023.03.012.

148.

Mobadersany

Tian

Yip

, et al. AI-enabled analysis of H&E-stained prostate cancer tissue images: Assessing risk for metastasis prior to apalutamide (APA) treatment of patients with non-metastatic castration-resistant prostate cancer (nmCRPC). Presented at: 2023 ASCO Annual Meeting; 2023.

149.

Fay

Liao

Lone

, et al. Independent blinded validation of an AI-based digital histology classifier for prostate cancer recurrence and metastasis risk prediction. American Society of Clinical Oncology; 2024.

150.

Zang

Meng

, et al. Fully automated volumetric assessment of tumor burden using artificial intelligence on (68)Ga-PSMA-11 PET predicts survival after (177)Lu-PSMA therapy in metastatic castration-resistant prostate cancer. Eur J Nucl Med Mol Imaging. 2026;53(3):1913-1926. doi:10.1007/s00259-025-07628-x.

151.

Ngiam

Khor

. Big data and machine learning algorithms for health-care delivery. Lancet Oncol. 2019;20(5):e262-e273. doi:10.1016/s1470-2045(19)30149-4.

152.

Nuhn

De Bono

Fizazi

, et al. Update on systemic prostate cancer therapies: Management of metastatic castration-resistant prostate cancer in the era of precision oncology. Eur Urol. 2019;75(1):88-99. doi:10.1016/j.eururo.2018.03.028.

153.

Mata

Retamero

Gupta

García Figueras

Luna

. Artificial intelligence–assisted prostate cancer diagnosis: Radiologic-pathologic correlation. Radiographics. 2021;41(6):1676-1697.

154.

Karimi

Samei

Kesch

Nir

Salcudean

. Prostate segmentation in MRI using a convolutional neural network architecture and training strategy based on statistical shape models. International Journal of Computer Assisted Radiology and Surgery. 2018;13(8):1211-1219.

155.

Mortensen

Borrelli

Poulsen

, et al. Artificial intelligence‐based versus manual assessment of prostate cancer in the prostate gland: a method comparison study. Clinical Physiology and Functional Imaging. 2019;39(6):399-406.

156.

Rubinstein

Salhov

Nidam-Leshem

, et al. Unsupervised tumor detection in Dynamic PET/CT imaging of the prostate. Medical Image Analysis. 2019;55:27-40.

157.

Alongi

Stefano

Comelli

Laudicella

Barone

Russo

. New artificial intelligence model for 18F-choline PET/CT in evaluation of high-risk prostate cancer outcome: texture analysis and radiomics features classification for prediction of disease progression. Society of Nuclear Medicine. 2020.

158.

Aprikian

Luz

Brimo

, et al. Improving ultrasound-based prostate volume estimation. BMC Urology. 2019;19:1-8.

159.

Carter

Hamper

Sheth

Sanders

Epstein

Walsh

. Evaluation of transrectal ultrasound in the early detection of prostate cancer. The Journal of Urology. 1989;142(4):1008-1010.

160.

Ukimura

de Castro Abreu

Gill

Shoji

Hung

Bahn

. Image visibility of cancer to enhance targeting precision and spatial mapping biopsy for focal therapy of prostate cancer. BJU International. 2013;111(8):E354-E364.

161.

Harvey

Pilcher

Richenberg

Patel

Frauscher

. Applications of transrectal ultrasound in prostate cancer. The British Journal of Radiology. 2012;85(special_issue_1):S3-S17.

162.

Heijmink

SWTPJ

van Moerkerk

Kiemeney

LALM

Witjes

Frauscher

Barentsz

. A comparison of the diagnostic performance of systematic versus ultrasound-guided biopsies of prostate cancer. European Radiology. 2006;16:927-938.

163.

Checcucci

Autorino

Cacciamani

, et al. Artificial intelligence and neural networks in urology: current clinical applications. Minerva Urologica e Nefrologica. 2020;72(1):49-57.

164.

Sisodiya

Kasherwal

Khan

, et al. Liquid Biopsies: Emerging role and clinical applications in solid tumours. Translational Oncology. 2023;35:101716.

165.

Thompson

Van Leeuwen

Moses

, et al. The diagnostic performance of multiparametric magnetic resonance imaging to detect significant prostate cancer. The Journal of Urology. 2016;195(5):1428-1435.

166.

Tolkach

Kristiansen

. The heterogeneity of prostate cancer: a practical approach. Pathobiology. 2018;85(1-2):108-116.

167.

Vagner

Spinelli

Minciacchi

, et al. Large extracellular vesicles carry most of the tumour DNA circulating in prostate cancer patient plasma. Journal of Extracellular Vesicles. 2018;7(1):1505403.

168.

Mehra

Dolling

Sumanasuriya

, et al. Plasma cell-free DNA concentration and outcomes from taxane therapy in metastatic castration-resistant prostate cancer from two phase III trials (FIRSTANA and PROSELICA). European Urology. 2018;74(3):283-291.

169.

Bahmad

Jalloul

Azar

, et al. Tumor microenvironment in prostate cancer: toward identification of novel molecular biomarkers for diagnosis, prognosis, and therapy development. Frontiers in Genetics. 2021;12:652747.

170.

Lin

Shen

Luo

, et al. Circulating tumor cells: biology and clinical significance. Signal Transduction and Targeted Therapy. 2021;6(1):404.

171.

Chen

Rycaj

Liu

Tang

. New insights into prostate cancer stem cells. Cell Cycle. 2013;12(4):579-586.

172.

Zhang

Zhou

Wang

, et al. Current stem cell biomarkers and their functional mechanisms in prostate cancer. International Journal of Molecular Sciences. 2016;17(7):1163.

173.

Matthews

Jordan

Wiegand

Pardoll

Lemischka

. A receptor tyrosine kinase specific to hematopoietic stem and progenitor cell-enriched populations. Cell. 1991;65(7):1143-1152.

174.

Tian

Lei

Gong

Sun

Piao

. Clinical implication of prognostic and predictive biomarkers for castration-resistant prostate cancer: a systematic review. Cancer Cell International. 2020;20:1-15.

175.

Jurtz

Johansen

Nielsen

, et al. An introduction to deep learning on biological sequence data: examples and solutions. Bioinformatics. 2017;33(22):3685-3690.

176.

Shin

S-Y

Centenera

Hodgson

, et al. A Boolean-based machine learning framework identifies predictive biomarkers of HSP90-targeted therapy response in prostate cancer. Frontiers in Molecular Biosciences. 2023;10:1094321.

177.

Samaržija

Trošelj

Konjevoda

. Prognostic significance of amino acid metabolism-related genes in prostate cancer retrieved by machine learning. Cancers. 2023;15(4):1309.

178.

Wei

Han

Zhang

, et al. Deep learning-based multi-omics integration robustly predicts relapse in prostate cancer. Frontiers in Oncology. 2022;12:893424.

179.

Qiao

Zhang

Zeng

Wang

. Using machine learning method to identify MYLK as a novel marker to predict biochemical recurrence in prostate cancer. Biomarkers in Medicine. 2021;15(1):29-41.

180.

Hussain

Saeed

Muhammad

Islam

Sheikh

. Classifying cancer patients based on DNA sequences using machine learning. Journal of Medical Imaging and Health Informatics. 2019;9(3):436-443.

181.

Wang

Cai

, et al. Identification of novel susceptibility loci and genes for prostate cancer risk: a transcriptome-wide association study in over 140,000 European descendants. Cancer Research. 2019;79(13):3192-3204.

182.

Zhou

Arslanturk

Craig

Heath

Draghici

. Discovery of primary prostate cancer biomarkers using cross cancer learning. Scientific Reports. 2021;11(1):10433.

183.

Prestagiacomo

Tradigo

Aracri

, et al. Data-independent acquisition mass spectrometry of EPS-urine coupled to machine learning: a predictive model for prostate cancer. ACS Omega. 2023;8(7):6244-6252.

184.

Penney

Tyekucheva

Rosenthal

, et al. Metabolomics of prostate cancer gleason score in tumor tissue and serum. Molecular Cancer Research. 2021;19(3):475-484.

185.

Gillies

Kinahan

Hricak

. Radiomics: images are more than pictures, they are data. Radiology. 2016;278(2):563-577.

186.

Ström

Kartasalo

Olsson

, et al. Artificial intelligence for diagnosis and grading of prostate cancer in biopsies: a population-based, diagnostic study. The Lancet Oncology. 2020;21(2):222-232.

187.

Lundervold

. An overview of deep learning in medical imaging focusing on MRI. Zeitschrift fuer Medizinische Physik. 2019;29(2):102-127.

188.

Ghose

Oliver

Martí

, et al. A survey of prostate segmentation methodologies in ultrasound, magnetic resonance and computed tomography images. Computer Methods and Programs in Biomedicine. 2012;108(1):262-287.

189.

Singh

Gupta

Acharya

. Segmentation of prostate contours for automated diagnosis using ultrasound images: A survey. Journal of Computational Science. 2017;21:223-231.

190.

Elmahdy

Jagt

Zinkstok

, et al. Robust contour propagation using deep learning and image registration for online adaptive proton therapy of prostate cancer. Medical Physics. 2019;46(8):3329-3343.

191.

Campanella

Hanna

Geneslaw

, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole slide images. Nature Medicine. 2019;25(8):1301-1309.

192.

Jung

Jin

M-S

Kim

, et al. Artificial intelligence system shows performance at the level of uropathologists for the detection and grading of prostate cancer in core needle biopsy: an independent external validation study. Modern Pathology. 2022;35(10):1449-1457.

193.

Bulten

Pinckaers

Van Boven

, et al. Automated deep-learning system for Gleason grading of prostate cancer using biopsies: a diagnostic study. The Lancet Oncology. 2020;21(2):233-241.

194.

Almeida

Tavares

JMRS

. Deep learning in radiation oncology treatment planning for prostate cancer: a systematic review. Journal of Medical Systems. 2020;44(10):179.

195.

Francolini

Desideri

Stocchi

, et al. Artificial Intelligence in radiotherapy: state of the art and future directions. Medical Oncology. 2020;37(6):50.

196.

Kovalchuk

Buyyounouski

Xing

Yang

. Dosimetric features‐driven machine learning model for DVH prediction in VMAT treatment planning. Medical Physics. 2019;46(2):857-867.

197.

Mahdavi

Tavakol

Sanei

, et al. Use of artificial neural network for pretreatment verification of intensity modulation radiation therapy fields. The British Journal of Radiology. 2019;92(1102):20190355.

198.

Lee

Kerns

Ostrer

Rosenstein

Deasy

. Machine learning on a genome-wide association study to predict late genitourinary toxicity after prostate radiation therapy. International Journal of Radiation Oncology* Biology* Physics. 2018;101(1):128-135.

199.

Carrara

Massari

Cicchetti

, et al. Development of a ready-to-use graphical tool based on artificial neural network classification: application for the prediction of late fecal incontinence after prostate cancer radiation therapy. International Journal of Radiation Oncology* Biology* Physics. 2018;102(5):1533-1542.

200.

Nouranian

Ramezani

Spadinger

Morris

Salcudean

Abolmaesumi

. Learning-based multi-label segmentation of transrectal ultrasound images for prostate brachytherapy. IEEE Transactions on Medical Imaging. 2015;35(3):921-932.

201.

Nicolae

Morton

Chung

, et al. Evaluation of a machine-learning algorithm for treatment planning in prostate low-dose-rate brachytherapy. International Journal of Radiation Oncology* Biology* Physics. 2017;97(4):822-829.

202.

Porreca

Colicchia

Busetto

Ferro

. MRI and Active Surveillance for Prostate Cancer. MDPI; 2020:590.

203.

Trojak

Stanuch

Kurzyna

Darocha

Skalski

. Mixed reality biopsy navigation system utilizing markerless needle tracking and imaging data superimposition. Cancers. 2024;16(10):1894.

204.

Shao

Shen

Sun

, et al. Real-time deep learning for tumor segmentation and tool tracking: development and validation of an AI navigation system in vacuum-assisted breast biopsy. World Journal of Surgical Oncology. 2025;23(1):1-12.

205.

Goldenberg

Nir

Salcudean

. A new era: artificial intelligence and machine learning in prostate cancer. Nature Reviews Urology. 2019;16(7):391-403.

206.

Fahim

Hasani

Kabba

Ragab

. Artificial intelligence in healthcare and medicine: clinical applications, therapeutic advances, and future perspectives. Eur J Med Res. 2025;30(1):848. doi:10.1186/s40001-025-03196-w.

207.

Guo

Wang

, et al. Evaluating ambient artificial intelligence documentation: effects on work efficiency, documentation burden, and patient-centered care. J Am Med Inform Assoc. 2025;33:273-282. doi:10.1093/jamia/ocaf180.

208.

Wulczyn

Nagpal

Symonds

, et al. Predicting prostate cancer specific-mortality with artificial intelligence-based Gleason grading. Communications Medicine. 2021;1(1):10.

209.

Esteva

Feng

van der Wal

, et al. Prostate cancer therapy personalization via multi-modal deep learning on randomized phase III clinical trials. NPJ Digital Medicine. 2022;5(1):71.

210.

Korevaar

Tennakoon

Page

, et al. Incidental detection of prostate cancer with computed tomography scans. Scientific Reports. 2021;11(1):7956.

211.

Liu

Xie

Feng

. Using artificial intelligence (Watson for Oncology) for treatment recommendations amongst Chinese patients with lung cancer: feasibility study. Journal of Medical Internet Research. 2018;20(9):e11087.

212.

Dolfing

. Case Study 20: The $4 Billion AI Failure of IBM Watson for Oncology. https://www.henricodolfing.ch/case-study-20-the-4-billion-ai-failure-of-ibm-watson-for-oncology/. Accessed December 23, 2025.

213.

Ross

Swetlitz

. IBM’s Watson supercomputer recommended ‘unsafe and incorrect’ cancer treatments, internal documents show. Stat. 2018;25:1-10.

214.

Zhou

Zhang

, et al. Concordance study between IBM Watson for oncology and clinical practice for patients with cancer in China. The Oncologist. 2019;24(6):812-819.

215.

Artera . U.S. FDA Grants Artera Breakthrough Device Designation for AI-Powered Software Transforming Prostate Cancer Care. https://artera.ai/news/u-s-fda-grants-artera-breakthrough-device-designation-for-ai-powered-software-transforming-prostate-cancer-care. Accessed Dec, 23, 2025.

216.

Oncology ASoC . ASCO and Google Cloud Launch AI-Powered ASCO Guidelines Assistant. https://dailynews.ascopubs.org/do/asco-and-google-cloud-launch-ai-powered-asco-guidelines-assistant. Accessed Dec, 23, 2025.

217.

Ferber

El Nahhas

Wölflein

, et al. Development and validation of an autonomous artificial intelligence agent for clinical decision-making in oncology. Nature Cancer. 2025;6:1-13.

218.

Sutton

Pincock

Baumgart

Sadowski

Fedorak

Kroeker

. An overview of clinical decision support systems: benefits, risks, and strategies for success. NPJ Digital Medicine. 2020;3(1):17.

219.

Kočo

Siebers

CCN

Schlooz

, et al. The Facilitators and Barriers of the Implementation of a Clinical Decision Support System for Breast Cancer Multidisciplinary Team Meetings—An Interview Study. Cancers. 2024;16(2):401.

220.

Karimi

Nir

Fazli

Black

Goldenberg

Salcudean

. Deep learning-based gleason grading of prostate cancer from histopathology images—role of multiscale decision aggregation and data augmentation. IEEE Journal of Biomedical and Health Informatics. 2019;24(5):1413-1426.

221.

Bulten

Bándi

Hoven

, et al. Epithelium segmentation using deep learning in H&E-stained prostate specimens with immunohistochemistry as reference standard. Scientific Reports. 2019;9(1):864.

222.

Yamamoto

Tsuzuki

Akatsuka

, et al. Automated acquisition of explainable knowledge from unannotated histopathology images. Nature Communications. 2019;10(1):5642.

223.

Pinckaers

van Ipenburg

Melamed

, et al. Predicting biochemical recurrence of prostate cancer with artificial intelligence. Communications Medicine. 2022;2(1):64.

224.

Raciti

Sue

Retamero

, et al. Clinical validation of artificial intelligence–augmented pathology diagnosis demonstrates significant gains in diagnostic accuracy in prostate cancer detection. Archives of Pathology & Laboratory Medicine. 2023;147(10):1178-1185.

225.

Toledo-Cortés

Useche

Müller

González

. Grading diabetic retinopathy and prostate cancer diagnostic images with deep quantum ordinal regression. Computers in Biology and Medicine. 2022;145:105472.

226.

Meyer

Khademi

Têtu

Han

Nippak

Remisch

. Impact of artificial intelligence on pathologists’ decisions: an experiment. Journal of the American Medical Informatics Association. 2022;29(10):1688-1695.

227.

Otálora

Marini

Müller

Atzori

. Combining weakly and strongly supervised learning improves strong supervision in Gleason pattern classification. BMC Medical Imaging. 2021;21(1):77.

228.

Schömig-Markiefka

Pryalukhin

Hulla

, et al. Quality control stress test for deep learning-based diagnostic model in digital pathology. Modern Pathology. 2021;34(12):2098-2108.

229.

Harmon

Patel

Sanford

, et al. High throughput assessment of biomarkers in tissue microarrays using artificial intelligence: PTEN loss as a proof-of-principle in multi-center prostate cancer cohorts. Modern Pathology. 2021;34(2):478-489.

230.

Macenko

Niethammer

Marron

, et al. A method for normalizing histology slides for quantitative analysis. IEEE. 2009:1107-1110.

231.

Swiderska-Chadaj

de Bel

Blanchet

, et al. Impact of rescanning and normalization on convolutional neural network performance in multi-center, whole-slide classification of prostate cancer. Scientific Reports. 2020;10(1):14398.

232.

Camps

Jiménez-Franco

García-Pablo

Joven

Arenas

. Artificial intelligence-driven integration of multi-omics and radiomics: A new hope for precision cancer diagnosis and prognosis. Biochim Biophys Acta Mol Basis Disease. 2025;1871:167841.

233.

Khosravi

Fuchs

. Artificial Intelligence–driven cancer dagnostics: Enhancing radiology and pathology through reproducibility, explainability, and multimodality. Cancer Research. 2025;85(13):2356-2367.

234.

Vakili

Beheshti

Barzegar Behrooz

Łos

Vitorino

Ghavami

. Transforming prostate cancer care: Innovations in diagnosis, treatment, and future directions. International Journal of Molecular Sciences. 2025;26(11):5386.

235.

Feretzakis

Juliebø-Jones

Tsaturyan

, et al. Emerging trends in AI and radiomics for bladder, kidney, and prostate cancer: a critical review. Cancers. 2024;16(4):810.

236.

Johnson

Umapathy

Gigax

, et al. Artificial Intelligence in Prostate MRI: Addressing Current Limitations Through Emerging Technologies. Journal of Magnetic Resonance Imaging. 2025.

237.

Turkbey

Haider

. Artificial intelligence for automated cancer detection on prostate MRI: Opportunities and ongoing challenges, from the AJR special series on AI applications. American Journal of Roentgenology. 2021;219(2):188-194.

238.

Perincheri

Levi

Celli

, et al. An independent assessment of an artificial intelligence system for prostate cancer detection shows strong diagnostic accuracy. Modern Pathology. 2021;34(8):1588-1595.

239.

Sisk

, et al. A multi-resolution model for histopathology image classification and localization with multiple instance learning. Computers in Biology and Medicine. 2021;131:104253.

240.

Egevad

Swanberg

Delahunt

, et al. Identification of areas of grading difficulties in prostate cancer and comparison with artificial intelligence assisted grading. Virchows Archiv. 2020;477:777-786.

241.

Hammouda

Khalifa

El-Melegy

, et al. A deep learning pipeline for grade groups classification using digitized prostate biopsy specimens. Sensors. 2021;21(20):6708.

242.

Brendel

Getseva

, et al. Weakly-supervised tumor purity prediction from frozen H&E stained slides. EBioMedicine. 2022;80:104067.

243.

Falahkheirkhah

Tiwari

Yeh

, et al. Deepfake histologic images for enhancing digital pathology. Laboratory Investigation. 2023;103(1):100006.