The accuracy of prostate lesion localization in cognitive fusion

1 Introduction

Prostate cancer (PCa) is one of the most common cancers in elderly men worldwide, and the first leading cancer type for the estimated new male malignant tumor cases in the United States [1]. Systematic biopsy guided by transrectal ultrasound remains the gold standard for diagnosis of PCa. However, the false negative rate of the current standard 12-core systematic biopsy is 10–20% [2], and leading to the underdiagnosis of clinically significant cancers in up to 35% of men [3]. Blindly increasing biopsy cores fail to improve the detection rate of PCa, instead, the operation cost and patient’s distress are increased. Conventional transrectal ultrasonography (TRUS) is not ideal for detecting PCa and the new frontier of sonographic modalities, such as real-time elastosonography and contrast-enhanced ultrasonography [4], remain under investigation and are not recommended for routine use [5]. Advances in imaging techniques, particularly multiparametric magnetic resonance imaging (mpMRI), allow prostate lesion visualization with a more precise localization. Taking advantage of mpMRI can be beneficial for targeted biopsy in PCa.

Some unnecessary biopsies can be avoided and suspicious target cores can be obtained by MRI guided biopsy [6]. Furthermore, mpMRI has been shown to have a higher accuracy and higher specificity for the detection of clinically significant PCa [7–9]. The advantage of mpMRI-guided targeted biopsy is clear. However, some limitations could include the associated costs and the specialized equipment.

The physician can mentally relocate the most appropriate area detected on the prebiopsy mpMRI, based on its zonal topography and on anatomical landmarks (ejaculatory duct, cyst, calcification, etc.), which is called cognitive fusion [10]. However, studies related to the match accuracy in cognitive fusion are sparse.

The MRI-TRUS fusion using fusion software is precise in overcoming human error and enabling targeted biopsy of suspicious lesions by visualization in TRUS images. We concentrated on the accuracy of prostate lesion localization in cognitive fusion compared with MRI-TRUS fusion (taken as a gold standard) and explored the applied scope of cognitive fusion.

2 Materials and methods

2.2 Cognitive fusion

The TRUS and MRI-TRUS fusion procedure were performed using MyLab Twice US systems (Esaote, Genoa, Italy) equipped with a biplane endorectal probe (TRT33; Esaote) and an operating bandwidth of 3–9 MHz. TRUS was individually performed by two operators (a senior sonographer with 15 years of experience in urogenital ultrasound and a junior sonographer with only 2 years of work experience). The cognitive fusion was proceeed as follows: First, the mpMRI images and reports on the PACS were reviewed by the two sonographers in individual workstations before performing the TRUS examination. Furthermore, the distance from suspicious area to the apex, posterior surface of the prostate gland and urethra were measured and recorded (Fig. 1). Then the patients were placed in the lithotomy position. Two operators performed TRUS examination for a few minutes each. The suspicious areas were marked with a circle (5 mm diameter) in transection and sagittal TRUS images based on the measured distance on MRI and anatomical landmarks (ejaculatory duct, cyst, calcification, etc). Finally, the four marked TRUS images were recorded on a US workstation.

OPEN IN VIEWER

Fig.1

Schematic representation of prostate base, mid and apex section shown on sagittal view (A); The red lesion is a mpMRI target shown on the transverse and sagittal views (B, C). Cognitive fusion of a mpMRI target on a TRUS image can be measured by simple three-dimensional distance [distance from the apex (z), distance from the urethra (x); distance from the posterior surface of the gland (y)].

3 Results

The mean age of the 32 patient participants was 68±9.4 years old (range 46–85 years old). The mean PSA level was 13.5±6.7 ng/mL (range 5.2–29.6 ng/mL), and the mean diameter of lesions was 11.9 mm (range 5–21 mm). The distribution of mpMRI lesions based on the location, PI-RADS score and the maximum diameter of lesions are presented in Table 1.

Using cognitive fusion imaging, 86.1% (31/36) of the lesions were accurately located by the senior sonographer; however, only 69.4% (25/36) of the lesions were accurately located by the junior sonographer. The correlation coefficient was 0.605 (P = 0.000) for the two sonographers.

The maximum diameter and PI-RADS score of the lesions were important factors that affected the accuracy of cognitive fusion (P < 0.05). The results of the logistic regression analysis are summarized in Table 2.

The lesions with high PI-RADS scores and the lesions with large diameters were more accurately located using cognitive fusion (P < 0.05). We showed cognitive fusion was 100% accurate when performed by an experienced sonographer (Table 3).

The MRI-TRUS fusion guided biopsy is a category of MRI-guided targeted biopsy and some reports have confirmed a higher accuracy rate, higher specificity and better risk stratification of PCa using MRI-TRUS fusion biopsy [12–14]. Cognitive fusion is easier compared with MRI-TRUS fusion using special ultrasonic image fusion software. Currently, most studies mainly focus on the difference of the PCa detection rate between MRI-TRUS fusion and cognitive fusion [15–17]. However, the accuracy of tracking lesions using cognitive fusion is not insufficient to MRI-TRUS fusion in patients with great lesions in the MRI an patients with high PSA levels.

Studies have shown that cognitive fusion prostate biopsies demonstrated a higher detection of PCa than the TRUS guided systematic biopsy [6, 18]. Previously, a common viewpoint was that cognitive fusion required a perfectly experienced TRUS operator and a reliable registration of mpMRI and TRUS images. In our study, a senior sonographer was qualified to perform the cognitive fusion.

In our study, we compared the geometric accuracy of cognitive and MRI-TRUS fusion (taken as the gold standard) and confirmed that the accuracy of cognitive fusion was not impacted by age, PSA level of patients or lesion location. Suspicious MRI targets eligible for TRUS-guided biopsy are those with a maximum diameter of 7 mm or longer [19]. In 6 of 36 enrolled lesions with a diameter of 5-6 mm, only 1 (16.7%) lesion was described accurately in cognitive fusion by the senior sonographer. Cornud F [20] compared the precision of cognitive fusion guided biopsy with the MRI/3D TRUS fusion system and confirmed that MRI/3D TRUS fusion system achieved a higher precision than the cognitive fusion guided biopsy. In our study, if suspicious MRI targets were larger than 7 mm and showed a PI-RADS score of at least 4, the lesions accurately located with cognitive fusion were consistent with that of MRI-TRUS fusion.

There were some limitations in our study. First, the number of lesions were low and were not located according to the 27-region subdivision scheme, therefore a future large-scale study is needed to further clarify the applied scope of cognitive fusion. Second, the MRI-TRUS fusion procedure was performed by one operator without consensus testing.

5 Conclusion

In conclusion, cognitive fusion is a reliable technique with dependency on working experience, and its accuracy of locating suspicious lesions is consistent with MRI-TRUS fusion in patients with high PI-RADS score and large lesions.

Author contribution

GX and LHX: project development, data collection, data analysis, manuscript writing. GX and LHX contributed equally to this paper. JW: data collection. HDS: data collection and analysis. HL: data collection and analysis. SSD: data analysis. RW: manuscript writing, project development.

Compliance with ethical standards