A 3D quantitative imaging biomarker in pre-treatment MRI predicts overall survival after stereotactic radiation therapy of patients with a singular brain metastasis

Abstract

Background

Brain metastases (BM) are the most frequent intracranial malignant tumor. Various prognostic factors facilitate the prediction of survival; however, few have become tools for clinical use.

Purpose

To investigate the role of three-dimensional (3D) quantitative tissue enhancement in pre-treatment cranial magnetic resonance imaging (MRI) as a radiomic biomarker for survival (OS) in patients with singular BM treated with stereotactic radiation therapy (SRT).

Material and Methods

In this retrospective study, 48 patients (27 non-small cell lung cancer and 21 melanoma) with singular BM treated with SRT, were analyzed. Contrast-enhanced MRI scans of the neurocranium were used for quantitative image analyses. Segmentation-based 3D quantification was performed to measure the enhancing tumor volume. A cut-off value of 68.61% of enhancing volume was used to stratify the cohort into two groups (≤68.61% and > 68.61%). Univariable and multivariable cox regressions were used to analyze the prognostic factors of OS and intracranial progression-free survival (iPFS).

Results

The level of enhancing tumor volume achieved statistical significance in univariable and multivariable analysis for OS (univariable: P = 0.005, hazard ratio [HR] = 0.375, 95% confidence interval [CI] = 0.168–0.744; multivariable: P = 0.006, HR = 0.376, 95% CI = 0.186–0.757). Patients with high-level enhancement (>68.61% enhancing lesion volume) survived significantly longer (4.9 vs. 10.2 months) and showed significantly longer iPFS rates (univariable: P < 0.001, HR = 0.046, 95% CI = 0.009–0.245).

Conclusions

Patients with lesions that show a higher percentage of enhancement in pre-treatment MRI demonstrated improved iPFS and OS compared to those with mainly hypo-enhancing lesions. Lesion enhancement may be a radiomic marker, useful in prognostic indices for survival prediction, in patients with singular BM.

Keywords

Brain metastases radiomics prognostic indices

Background

Brain metastases (BM) are the most frequent intracranial malignant tumor; two-thirds of BM – occurring in adult patients – are secondary to lung cancer, melanoma, or breast cancer.

Modern systemic therapies available today, target predominantly extracranial disease, which leads to an increase of the incidence of brain metastases (1).

The most common cause of BM is lung cancer; this type of cancer is also the most common cause of cancer-related death in the world (2). During the course of the disease, approximately 40% of patients with stage IV non-small cell lung cancer (NSCLC) will develop BM leading to a drastic reduction of survival, as well as quality of life (3,4). The prognosis of patients who suffer from BM from NSCLC is generally poor; the median overall survival (OS) is <1 year.

Melanoma is the cancer with the highest rate of dissemination to the brain. Approximately 40% of patients develop clinically apparent BM and on autopsy, up to 70% of patients show BM (5,6).

Survival time in BM is related to a plethora of factors and, while generally poor, is hard to predict. The individual prognosis may depend on the type of primary cancer, systemic control, type of treatment, and response to treatment. Consequently, great efforts have been made to identify the prognostic parameters that would help in choosing the right treatment strategy for BM patients (7 –11).

Standard therapies for BM patients include whole-brain radiation therapy (WBRT) and focal therapies such as resection as well as stereotactic radiation therapy (SRT) (11 –13). While WBRT has, historically, been widely used, some of the intrinsic limitations of WBRT have recently led to the implementation of more focal and more aggressive radiotherapy techniques in order to improve local control and reduce toxicity, especially in patients that are considered to be in an oligometastatic state (14).

The term “oligometastasis” was introduced in 1995 by Hellman and Weichselbaum. The term describes a state between local and systematic disease and implicates potential curative options with local treatments (15).

The identification of oligometastatic patients who would potentially benefit from these aggressive focal therapies is critical, especially since the scientific community has not yet agreed on a common definition of the term. Several authors have recently tried to identify the prognostic factors of survival, others have proposed prognostic indices based on prognostic factors, in order to simplify the decision-making in this context (16).

Factors such as Karnofsky performance status (KPS), patient age, tumor volume, presence of extracranial metastases, control of primary disease, and number of BM are used and combined in the literature to determine prognostic indices that have been proposed for the prediction of survival in patients with BM. Well-known scores are the disease-specific graded prognostic index (DS-GPA), the NSCLC-specific index (NSCLC-RADES), and the Score Index for Radiosurgery (SIR) (7,17–19).

Another example is the recursive partitioning analysis (RPA) classification that is commonly used in the evaluation of patients in clinical trials. None of these indexes, however, managed to get into use of daily clinical practice. In this context, the attention of researchers has focused on the field of quantitative analysis. These analyses are based on magnetic resonance imaging (MRI), which is the routine examination for diagnosis and follow-up. In fact, it has been shown that contrast-enhanced MRI is qualitatively superior to enhanced computed tomography (CT) and to non-enhanced MRI (20,21). In addition, imaging biomarkers have recently been proposed as predictors of treatment success after radiotherapy treatments (22). Using the same techniques as in this work, Schneider et al. were able to differentiate responders from non-responders on basis of tumor enhancement (viability) and total tumor volume in patients with vestibular schwannoma (23).

Therefore, it was our aim to further investigate the informative value of tumor enhancement and to evaluate a potential role for three-dimensional (3D) quantitative, enhancement-based MRI parameters, as prognostic radiomic markers for overall survival in patients with a singular BM from NSCLC or melanoma following SRT.

Methods

Patient population

Patients with BMs from NSCLC or melanoma who were treated with SRT between April 2004 and May 2014 at our institution were identified (n = 90 NSCLC, n = 112 melanoma). Patients were included in the final analysis if they had a therapy naïve, singular BM and had undergone contrast-enhanced T1-weighted (T1W) MRI within 2.5 months before SRT. Patients who had received BM-directed therapy before SRT were excluded to avoid confounding treatment effects. Furthermore, patients with missing or inadequate MR images and patients with severe motion artifacts that affected the tumor region were excluded. The remaining 27 NSCLC and 21 melanoma patients were included into the further analysis. A sub-analysis of the NSCLC and melanoma cohorts was performed.

Written informed consent was acquired from all patients with respect to RT treatment and clinical data management for research purposes. Our institutional review board approved this retrospective study.

SRT technique

We performed SRT using a Novalis® therapy linear accelerator (LINAC, BrainLab®, Munich, Germany) with beam-shaping capability, a built-in multi-leaf collimator (MLC), and image guidance. The Novalis ExacTrac® image-guided frameless system enabled imaging of the patient in any couch position using a frameless positioning array. MRI/CT fusion planning was performed. The 3D treatment planning system iplanRT® was used. Gross tumor volume (GTV) was defined as the area of contrast enhancement on T1W MRI images, the PTV included a 1–2 mm isotropic safety margin. If fusion images were considered to be of good quality, the PTV margin used was only 1 mm. If fusion images were not considered adequate, a safety margin of 2 mm was used. The dose was prescribed to the 80% isodose at the PTV margin.

MRI technique

MRI studies were obtained on a 1.5-T (Siemens, Philips or GE) scanner.

The standard MRI protocol included native axial T1W turbo-spin echo (TSE) sequences and axial T2-weighted (T2W) TSE. After gadolinium-injection, axial T1W TSE sequences were acquired. Slice thickness was 5 mm. Matrix sizes and repetition times/echo times varied for the different scanners and were not standardized.

Imaging data evaluation

Semi-automatic 3D quantitative analysis was performed using a dedicated software for quantitative analysis (IntelliSpace Portal V.8, software from Philips Healthcare, Hamburg, Germany) which was originally used for liver lesion investigations. Previous studies have demonstrated that semi-automatic segmentation analysis can accurately segment in 3D and has high reproducibility (24,25).

After manual delineation of the tumor boundaries on one slice of the T1W contrast-enhanced MRI, the software automatically expands and adjusts the selection to all slices. The images are than compared to the native pre-contrast enhancement slices for the subtraction of background enhancement. Viable enhancing tumor is then defined in comparison to the enhancement of healthy cerebral parenchyma. Therefore, a cubic reference region of interest (ROI) of 1 cm³ of healthy cerebral parenchyma in the contralateral hemisphere on the same level as the BM was chosen. The differentiation between contrast-enhanced and non-enhanced tissue was made by the software based on the mean brightness value (MBV). Brightness values above the MBV plus 2 standard deviations (SD) were categorized as contrast-enhanced. Values below MBV + (2*SD) were considered as non-enhanced. During manual ROI placement, the investigator tried to avoid gray matter, ventricles, and motion artifacts. Throughout the process, manual corrections are possible. The software quantifies tumor diameter and volume and enhanced tumor volume in absolute numbers as well as the percentage of enhanced tumor volume. The process described above lasts ∼1 min for every examination. For the analysis, the patients were separated into two groups based on the high-enhancement and low-enhancement ratios of their tumors. An example of the segmentation outline and 3D rendering produced is shown in Fig. 1.

Fig. 1.

(a) 3D quantitative image analysis. Representative contrast-enhanced T1W MR image demonstrates the semi-automatic tumor and ROI segmentation. (b) 3D quantitative image analysis. qEASL color map of tumor: red represents the enhancing region.

Statistical analysis

All statistical analyses were performed using GraphPad (Version 7, San Diego, CA, USA), SPSS (version 20.0, Armonk, NY, USA), and Cutoff Finder (26). The OS and iPFS were measured starting from the time of first treatment.

Survival was evaluated using Kaplan–Meier analysis and compared using the Log-rank test as well as proportional hazard ratios (HR). The Cox proportional hazard model was used for multivariable analysis. Multivariable Cox regression analysis was not performed for iPFS because the number of events was too low for statistically appropriate analysis.

The continuous variable “percentage of enhanced tumor volume” obtained from image data analysis was dichotomized in order to stratify patients into two groups. The cut-off was found using the “survival: significance” function of Cutoff Finder. This function fits Cox proportional hazard models to the dichotomized variable and the survival variable. Survival analysis is executed using the functions coxph and survfit from the R package “survival” (27). The optimal cut-off is defined as the point with the most significant log-rank test split. The NSCLC cohort was used to find the optimal cut-off; the melanoma cohort was then used for validation.

In this work, a P-value < 0.05 was considered statistically significant. A P-value < 0.1 was considered a trend and was the criterion for inclusion in multivariable analysis.

Results

Patients

Patient characteristics are shown in Table 1. The overall study cohort comprised 48 patients (27 NSCLC, 21 melanoma). Baseline characteristics such as age, gender, KPS, extracranial disease, histology, and duration between imaging and SRT were recorded.

Table 1.

Patient characteristics.

	n, mean ± SD	%
Age (years)	62 ± 11.6
Sex (M/F)	28/20	58.3/41.7
NSCLC	27	56.25
Melanoma	21	43.75
UICC IV (at diagnosis of the primary tumor)	27	56.25
Extracranial metastases at time of radiation	35	72.92

NSCLC, non-small cell lung carcinoma; UICC, Union for International Cancer Control.

Mean patient age was 62 years. The majority of the patients were male (58.3%) and adenocarcinoma was the most common histologic subtype among the NSCLC patients (59%). More than half of the patients (56.25%) were Union for International Cancer Control (UICC) stage IV at the time of diagnosis. Thirty-five patients (72.9%) had extracranial metastases at the time of radiation. The mean time between baseline imaging and the SRT was 29 days (±23 days). The mean axial diameter and the mean total volume of the brain metastases were 20.4 mm and 6.0 cm³, respectively. The mean enhancing volume was 3.9 cm³ and the mean percentage enhancement was 71.8%.

Definition of the cut-off for the percentage of enhanced tumor volume

In the NSCLC cohort, optimal cut-off for dichotomization of the percentage of enhanced tumor volume was 68.61%. Splitting the NSCLC cohort in two groups based on this cut-off point led to the most significant log-rank test split, P = 0.045 (HR = 0.41, 95% confidence interval [CI] = 0.17–1.01). Validation of this cut-off was performed in the melanoma cohort with similar significant results, P = 0.043 (HR = 0.33, 95% CI = 0.11–1.01).

Follow-up and survival analysis

Analysis of the predictors of survival is shown in Tables 2 –4. The median OS after SRT was 8.9 months (95% CI = 6.4–11.4). As shown in Figure 2, in univariable analysis, patients with >68.61% volume of high-enhancement tissue (viability) showed significantly longer median OS rates than patients with low-enhancement tumor tissue (P = 0.005, HR = 0.37, 95% CI = 0.19–0.74). There was a trend towards shorter OS survival rates in older patients (P = 0.07, HR = 1.832, 95% CI = 0.95–3.53). The following factors were not significant predictors of survival in univariable analysis: gender; KPS; UICC stage; extracranial disease status; tumor volume; and tumor entity (NSCLC versus melanoma). In univariable subgroup analysis, NSCLC patients with adenocarcinoma histology showed significantly longer OS rates than the NSCLC patients of other histology.

Table 2.

Univariable and multivariable analysis of potential predictive factors of overall survival in the overall cohort.

	Univariable				Multivariable
	P	HR	95% CI		P	HR	95% CI
Sex (F vs. M)	0.555	0.827	0.44	1.554
Viability (≤ vs. > 68.61%)	0.005*	0.372	0.186	0.744	0.006*	0.376	0.186	0.757
KPS (≤ vs. > 70%)	0.429	0.777	0.416	1.452
UICC (I–III vs. IV)	0.156	1.584	0.839	2.991
ECM	0.134	1.777	0.837	3.772
Age	0.07	1.832	0.952	3.528	0.211	1.563	0.777	3.146
Volume (≤ vs. > median)	0.923	0.97	0.52	1.808
Entity (melanoma vs. NSCLC)	0.076	0.562	0.297	1.062	0.207	0.645	0.326	1.275

High enhancement (viability) is a positive predictive factor of OS in univariable and multivariable analysis.

*significant.

KPS, Kamofsky performance status; ECM, extracranial disease; OS, overall survival.

Table 3.

Univariable and multivariable analysis of potential predictive factors of OS in the NSCLC cohort.

	Univariable				Multivariable
	P	HR	95% CI		P	HR	95% CI
Sex (M vs. F)	0.319	0.636	0.261	1.548
Viability (≤ vs. >68.61%)	0.052	0.413	0.169	1.007	0.332	0.989	0.967	1.011
KPS (≤ vs. >70%)	0.158	0.538	0.228	1.271
UICC (I–III vs. IV)	0.055	2.902	0.976	8.633	0.319	1.861	0.548	6.317
ECM (none vs. present)	0.126	2.106	0.811	5.464
Age (≤ vs. >64)	0.475	1.342	0.598	3.009
Histology (adenocarcinoma vs. other)	0.006*	0.266	0.104	0.682	0.032*	0.349	0.133	0.915
Volume (≤ vs. >median)	0.073	2.172	0.931	5.066	0.47	1.444	0.533	3.917

Adenoarcinoma histology is a positive predictive factor of OS in univariable and multivariable analysis.

*significant.

KPS, Kamofsky performance status; ECM, extracranial disease; OS, overall survival.

Table 4.

Univariable and multivariable analysis of potential predictive factors of OS in the melanoma cohort.

	Univariable				Multivariable
	P	HR	95% CI		P	HR	95% CI
Sex (M vs. F)	0.334	1.689	0.584	4.885
Viability (≤ vs. >68.61%)	0.053	0.335	0.111	1.015	0.053	0.335	0.111	1.015
KPS (≤ vs. >70%)	0.729	0.832	0.294	2.357
UICC (I–III vs. IV)	0.194	0.465	0.146	1.478
ECM (none vs. present)	0.353	2.04	0.453	9.195
Age (≤ vs. >54)	0.498	1.421	0.514	3.932
Volume (≤ vs. >median)	0.284	0.565	0.199	1.604

KPS, Kamofsky performance status; ECM, extracranial disease; OS, overall survival.

Multivariable analysis confirmed a potential prognostic role of high enhancement tumor volume (P = 0.006, HR = 0.376, 95% CI = 0.19–0.76) as well as the prognostic role of adenocarcinoma histology in the NSCLC subgroup (P = 0.032, HR = 0.35, 95% CI = 0.13–0.92).

IPFS rates are shown in Table 5. Median iPFS was 4.6 months. Patients with a higher level of enhancing tumor volume showed significantly longer iPFS rates (univariable regression: P < 0.001, HR = 0.046, 95% CI = 0.009–0.245).

Table 5.

Univariable analysis of potential predictive factors of intracranial progression-free survival in the overall cohort.

	Univariable
	P	HR		95% CI
Sex (F vs. M)	0.525	1.335	0.548	3.250
Viability (≤ vs. > 68.61%)	<0.001*	0.046	0.009	0.245
KPS (≤ vs. > 70%)	0.869	1.080	0.434	2.686
UICC (I–III vs. IV)	0.169	0.523	0.207	1.318
ECM	0.636	1.282	0.458	3.587
Age	0.905	0.945	0.376	2.374
Volume (≤ vs. > median)	0.570	0.766	0.305	1.923
Entity (melanoma vs. NSCLC)	0.781	0.875	0.342	2.242

Note that multivariable Cox regression was not performed because the number of events was too low for statistically appropriate analysis.*significant.

Discussion

BM are the most common among intracranial malignancies and the most common cause of cancer-related neurological complications (28). Lung cancer is the most prevalent cause of BM and melanoma is the type of cancer with the highest incidence of BM. The presence of BM is associated with a decrease in OS rates, however, the individual response rates differ significantly from patient to patient. Therefore, researchers have recently made great efforts to identify prognostic factors of OS, particularly in oligometastatic patients who are candidates for aggressive focal therapies like SRT or surgical resection (8). Various prognostic factors and indices facilitate the prediction of survival of BM patients; however, few of them have become tools for clinical everyday use. In this context, the field of radiomics may offer new automated approaches in the future.

Fig. 2.

Overall survival stratified by enhanced tumor volume (P = 0.005, HR = 0.372, 95% CI = 0.186–0.744).

Several prognostic indices for the survival prediction of BM patients have been proposed. The recursive partitioning analysis (RPA) classification is commonly used in the evaluation of patients in clinical trials. The disease-specific graded prognostic index (DS-GPA), the NSCLC-specific index (NSCLC-RADES), and the Score Index for Radiosurgery (SIR) are other examples of well-known scores (7,17–19). These indices are calculated using different prognostic factors such as age, gender, KPS, the presence of extracranial disease (ECM), the number of BMs, the interval between primary tumor diagnosis and radiotherapy, as well as the control of primary disease and the total intracranial tumor volume. All of the mentioned indices have shown encouraging results in clinical studies, but their clinical use is limited by complexity, time intensity, and by the presence of variables that are time-consuming to obtain or that can only be calculated in the process of the disease (e.g. control of extracranial disease or BM volume) (29). In this context, researchers have pinned great hopes on the field of radiomics, the study of imaging biomarkers that might be used to predict treatment response and survival. Quantitative 3D imaging biomarkers have been described as potential predictors of treatment response as well as survival, in patients who have received focal therapies for extracranial as well as for intracranial malignancies (23,24,30 –35). In this work, we performed volumetric measurements of tumor enhancement in baseline MRIs in order to predict survival following SRT of singular BM. Quantitative measurement of the percentage of high-enhancing tumor volume, as well as of the total tumor volume, were performed using semi-automatic segmentation, a method that has been shown as precise, reader-independent, reproducible, and time-efficient (24,25).

In this context, the percentage of high enhancement is a factor that could be used in the prediction of iPFS and OS after SRT. The factor might be used in the calculation of future prognostic indices for patients with melanoma or NSCLC BM. Patients with hyper-enhancing lesions might benefit from receiving focal and aggressive treatments, such as SRT.

A potential biological explanation for the prognostic role of hyper-enhancement is that hypervascularization leads to an increased oxygenation in the tumor. Therefore, radiotherapy should be more effective in treating these lesions, since oxygenated cells are considerably more radiosensitive than hypoxic ones (36). In addition, metastases with higher proliferation rates are more likely to develop central necrosis suggesting that the volume of hypo-enhancement could be a surrogate marker for tumor proliferation rate.

The most important limitation of this study obviously stems from the retrospective nature of the analysis, which is prone to bias. Secondly, the relatively small number of patients poses a potential limitation. The overall cohort also consisted of patients suffering from two distinct tumor entities. We have chosen to use these pooled data in order to achieve a number of patients high enough for multivariable analysis, assuming that the biological basis of a potential predictive role of percentage of enhanced tumor volume is the same in both cohorts. This may confound multivariable analysis, since included covariates in multivariable analysis may have different impacts in the two sub-cohorts. However, it should be mentioned, that a cohort size of 48 patients is comparable to other works on 3D quantitative MRI analyses (24,25,30,33,37 –39). Third, technical limitations arose: due to the retrospective character of the study, the images were acquired using different scanners; slices were therefore standardized using a slice thickness of 5 mm. The relatively small sample size might also explain why “BM volume,” histology, UICC, and KPS did not show a significant association with iPFS and OS. Still, prospective validation of our results is necessary.

In conclusion, in this work, we evaluated a potential role for 3D imaging of radiomic biomarkers based on MRI contrast enhancement. The main finding of our work is that high enhancement of singular NSCLC and melanoma BM on baseline MRI is a significant predictor of OS and iPFS after SRT. Lesion enhancement might, therefore, play a significant role as a non-invasive imaging radiomic biomarker in future treatment decisions.

Footnotes

Declaration of conflicting interests

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

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Nayak

Lee

Wen

PY.

Epidemiology of brain metastases

Curr Oncol Rep 2012; 14:48–54.

Torre

Bray

Siegel

, et al. Global cancer statistics, 2012. CA Cancer J Clin 2015; 65:87–108.

Schouten

Rutten

Huveneers

, et al. Incidence of brain metastases in a cohort of patients with carcinoma of the breast, colon, kidney, and lung and melanoma. Cancer 2002; 94:2698–2705.

Newman

Hansen

HH.

Proceedings: Frequency, diagnosis, and treatment of brain metastases in 247 consecutive patients with bronchogenic carcinoma.

Cancer 1974; 33:492–496.

Flanigan

Jilaveanu

Faries

, et al. Melanoma brain metastases: is it time to reassess the bias? Curr Prob Cancer 2011; 35:200–210.

Fife

Colman

Stevens

, et al. Determinants of outcome in melanoma patients with cerebral metastases. J Clin Oncol 2004; 22:1293–1300.

Sperduto

Chao

Sneed

, et al. Diagnosis-specific prognostic factors, indexes, and treatment outcomes for patients with newly diagnosed brain metastases: a multi-institutional analysis of 4,259 patients. Int J Radiat Oncol Biol Phys 2010; 77:655–661.

Jenkinson

Haylock

Shenoy

, et al. Management of cerebral metastasis: evidence-based approach for surgery, stereotactic radiosurgery and radiotherapy. Eur J Cancer 2011; 47:649–655.

Nieder

Two new prognostic indices for brain metastases–where do we go from here?

Strahlenther Onkol 2008; 184:488–489.

10.

Rades

Gerdan

Segedin

, et al. Brain metastasis. Prognostic value of the number of involved extracranial organs. Strahlenther Onkol 2013; 189:996–1000.

11.

Khademalhosseini

Khademolhosseini

Kummer

, et al. [Stereotactic radiosurgery for brainstem metastases: In experienced hands a good treatment option with excellent risk-benefit ratio]. Strahlenther Onkol 2017; 193:347–349.

12.

Arrieta

Villarreal-Garza

Zamora

, et al. Long-term survival in patients with non-small cell lung cancer and synchronous brain metastasis treated with whole-brain radiotherapy and thoracic chemoradiation. Radiat Oncol 2011; 6:166.

13.

Deutsch

Parsons

Mercado

Jr.

Radiotherapy for intracranial metastases.

Cancer 1974; 34:1607–1611.

14.

Linskey

Andrews

Asher

, et al. The role of stereotactic radiosurgery in the management of patients with newly diagnosed brain metastases: a systematic review and evidence-based clinical practice guideline. J Neurooncol 2010; 96:45–68.

15.

Hellman

Weichselbaum

RR.

Oligometastases.

J Clin Oncol 1995; 13:8–10.

16.

Kaul D, Angelidis A, Budach V, et al. Prognostic indices in stereotactic radiotherapy of brain metastases of non-small cell lung cancer. Radiat Oncol 2015;10:244.

17.

Golden

Lamborn

McDermott

, et al. Prognostic factors and grading systems for overall survival in patients treated with radiosurgery for brain metastases: variation by primary site. J Neurosurg 2008; 109 Suppl:77–86.

18.

Rades

Dziggel

Segedin

, et al. A new survival score for patients with brain metastases from non-small cell lung cancer. Strahlenther Onkol 2013; 189:777–781.

19.

Weltman

Salvajoli

Brandt

, et al. Radiosurgery for brain metastases: a score index for predicting prognosis. Int J Radiat Oncol Biol Phys 2000; 46:1155–1161.

20.

Davis

Hudgins

Peterman

, et al. Diagnosis of cerebral metastases: double-dose delayed CT vs contrast-enhanced MR imaging. Am J Neuroradiol 1991; 12:293–300.

21.

Schellinger

Meinck

Thron

Diagnostic accuracy of MRI compared to CCT in patients with brain metastases.

J Neurooncol 1999; 44:275–281.

22.

Peeken

Nusslin

Combs

SE.

“Radio-oncomics”: The potential of radiomics in radiation oncology.

Strahlenther Onkol 2017; 193:767–779.

23.

Schneider T, Chapiro J, Lin M, et al. 3D quantitative assessment of response to fractionated stereotactic radiotherapy and single-session stereotactic radiosurgery of vestibular schwannoma. Eur Radiol 2016;26:849–857.

24.

Chapiro J, Duran R, Lin M, et al. Early survival prediction after intra-arterial therapies: a 3D quantitative MRI assessment of tumour response after TACE or radioembolization of colorectal cancer metastases to the liver. Eur Radiol 2015;25:1993–2003.

25.

Tacher

Lin

Chao

, et al. Semiautomatic volumetric tumor segmentation for hepatocellular carcinoma: comparison between C-arm cone beam computed tomography and MRI. Acad Radiol 2013; 20:446–452.

26.

Budczies

Klauschen

Sinn

, et al. Cutoff Finder: a comprehensive and straightforward Web application enabling rapid biomarker cutoff optimization. PLoS One 2012; 7:e51862.

27.

RC Team. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing, 2008.

28.

Chamberlain

MC.

Brain metastases: a medical neuro-oncology perspective.

Expert Rev Neurother 2010; 10:563–573.

29.

Sperduto

Berkey

Gaspar

, et al. A new prognostic index and comparison to three other indices for patients with brain metastases: an analysis of 1,960 patients in the RTOG database. Int J Radiat Oncol Biol Phys 2008; 70:510–514.

30.

Chapiro J, Lin M, Duran R, et al. Assessing tumor response after loco-regional liver cancer therapies: the role of 3D MRI. Expert Rev Anticancer Ther 2015;15:199–205.

31.

Fleckenstein

Schernthaner

Duran

, et al. 3D Quantitative tumour burden analysis in patients with hepatocellular carcinoma before TACE: comparing single-lesion vs. multi-lesion imaging biomarkers as predictors of patient survival. Eur Radiol 2016; 26:3243–3252.

32.

Donaldson

Buckley

O’Connor

, et al. Enhancing fraction measured using dynamic contrast-enhanced MRI predicts disease-free survival in patients with carcinoma of the cervix. Br J Cancer 2010; 102:23–26.

33.

Duran R, Chapiro J, Frangakis C, et al. Uveal Melanoma Metastatic to the Liver: The Role of Quantitative Volumetric Contrast-Enhanced MR Imaging in the Assessment of Early Tumor Response after Transarterial Chemoembolization. Transl Oncol 2014;7:447–455.

34.

Tacher

Lin

Duran

, et al. Comparison of existing response criteria in patients with hepatocellular carcinoma treated with transarterial chemoembolization using a 3D quantitative approach. Radiology 2016; 278:275–284.

35.

Chapiro J, Duran R, Lin M, et al. Transarterial chemoembolization in soft-tissue sarcoma metastases to the liver - the use of imaging biomarkers as predictors of patient survival. Eur J Radiol 2015;84:424–430.

36.

Conger

AD.

The effect of oxygen on the radiosensitivity of mammalian cells.

Radiology 1956; 66:63–69.

37.

Chapiro J, Duran R, Lin M, et al. Three-Dimensional Quantitative Assessment of Uterine Fibroid Response after Uterine Artery Embolization Using Contrast-Enhanced MR Imaging. J Vasc Interv Radiol 2015;26:670–678 e672.

38.

Pellerin

Lin

Bhagat

, et al. Comparison of semi-automatic volumetric VX2 hepatic tumor segmentation from cone beam CT and multi-detector CT with histology in rabbit models. Acad Radiol 2013; 20:115–121.

39.

Chapiro J, Wood LD, Lin M, et al. Radiologic-pathologic analysis of contrast-enhanced and diffusion-weighted MR imaging in patients with HCC after TACE: diagnostic accuracy of 3D quantitative image analysis. Radiology 2014;273:746–758.