Development and validation of a clinical-radiomics model for prediction of prostate cancer: a multicenter study

PURPOSE

To develop an early diagnosis model of prostate cancer based on clinical-radiomics to improve the accuracy of imaging diagnosis of prostate cancer.

METHODS

The multicenter study enrolled a total of 449 patients with prostate cancer from December 2017 to January 2022. We retrospectively collected information from 342 patients who underwent prostate biopsy at Minhang Hospital. We extracted T2WI images through 3D-Slice, and used mask tools to mark the prostate area manually. The radiomics features were extracted by Python using the "Pyradiomics" module. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used for data dimensionality reduction and feature selection, and the radiomics score was calculated according to the correlation coefficients. Multivariate logistic regression analysis was used to develop predictive models. We incorporated the radiomics score, PI-RADS, and clinical features, and this was presented as a nomogram. The model was validated using a cohort of 107 patients from the Xuhui Hospital.

RESULTS

In total, 110 effective radiomics features were extracted. Finally, 9 features were significantly associated with the diagnosis of prostate cancer, from which we calculated the radiomics score. The predictors contained in the individualized prediction nomogram included age, fPSA/tPSA, PI-RADS, and radiomics score. The clinical-radiomics model showed good discrimination in the validation cohort (C-index = 0.88).

CONCLUSION

This study presents a clinical-radiomics model that incorporates age, fPSA/PSA, PI-RADS, and radiomics score, which can be conveniently used to facilitate individualized prediction of prostate cancer before prostate biopsy.

Diagnostic value of integrated 18F-PSMA-1007 PET/MRI Compared with that of Biparametric MRI for the detection of Prostate Cancer

Objective

Materials and methods

Results

Conclusion

Artificial Intelligence (AI) for Automated Cancer Detection on Prostate MRI: Opportunities and Ongoing Challenges, From the AJR Special Series on AI Applications

Prostate MRI utilization has increased greatly in the past decade, serving primarily to direct targeted prostate biopsies. However, prostate MRI interpretation remains prone to consideration interreader variation. Artificial intelligence (AI) has the potential to standardize the detection of lesions on MRI that are suspicious for prostate cancer (PCa). This review explores the current status of AI for the automated detection of PCa by MRI. Recent literature providing promising results for AI models for PCa detection by MRI is highlighted. Numerous limitations of the existing literature are also described, including biases in model validation, heterogeneity in reporting of performance metrics, and lack of sufficient evidence of clinical translation. Challenges related to AI ethics and data governance are also discussed. An outlook is provided for AI in lesion detection by prostate MRI in the coming years, emphasizing current research needs. Future investigations, incorporating large-scale diverse multi-institutional training and testing datasets, are anticipated to enable the development of more robust AI models for PCa detection by MRI, though prospective clinical trials will ultimately be required to establish benefit of AI in patient management.

Using Prostate Imaging-Reporting and Data System (PI-RADS) Scores to Select an Optimal Prostate Biopsy Method: A Secondary Analysis of the Trio Study

BACKGROUND

While magnetic resonance imaging (MRI)-targeted biopsy (TBx) results in better prostate cancer (PCa) detection relative to systematic biopsy (SBx), the combination of both methods increases clinically significant PCa detection relative to either Bx method alone. However, combined Bx subjects patients to higher number of Bx cores and greater detection of clinically insignificant PCa.

OBJECTIVE

To determine if prebiopsy prostate MRI can identify men who could forgo combined Bx without a substantial risk of missing clinically significant PCa (csPC).

DESIGN, SETTING, AND PARTICIPANTS

Men with MRI-visible prostate lesions underwent combined TBx plus SBx.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

The primary outcomes were detection rates for grade group (GG) ≥2 and GG ≥3 PCa by TBx and SBx, stratified by Prostate Imaging-Reporting and Data System (PI-RADS) score.

RESULTS AND LIMITATIONS

Among PI-RADS 5 cases, nearly all csPCs were detected by TBx, as adding SBx resulted in detection of only 2.5% more GG ≥2 cancers. Among PI-RADS 3-4 cases, however, SBx addition resulted in detection of substantially more csPCs than TBx alone (7.5% vs 8%). Conversely, TBx added little to detection of csPC among men with PI-RADS 2 lesions (2%) relative to SBx (7.8%).

CONCLUSIONS

While combined Bx increases the detection of csPC among men with MRI-visible prostate lesions, this benefit was largely restricted to PI-RADS 3-4 lesions. Using a strategy of TBx only for PI-RADS 5 and combined Bx only for PI-RADS 3-4 would avoid excess biopsies for men with PI-RADS 5 lesions while resulting in a low risk of missing csPC (1%).

PATIENT SUMMARY

Our study investigated an optimized strategy to diagnose aggressive prostate cancer in men with an abnormal prostate MRI (magnetic resonance imaging) scan while minimizing the risk of excess biopsies. We used a scoring system for MRI scan images called PI-RADS. The results show that MRI-targeted biopsies alone could be used for men with a PI-RADS score of 5, while men with a PI-RADS score of 3 or 4 would benefit from a combination of MRI-targeted biopsy and systematic biopsy. This trial is registered at ClinicalTrials.gov as NCT00102544.

Deep-Learning-Based Artificial Intelligence for PI-RADS Classification to Assist Multiparametric Prostate MRI Interpretation: A Development Study

BACKGROUND

The Prostate Imaging Reporting and Data System (PI-RADS) provides guidelines for risk stratification of lesions detected on multiparametric MRI (mpMRI) of the prostate but suffers from high intra/interreader variability.

PURPOSE

To develop an artificial intelligence (AI) solution for PI-RADS classification and compare its performance with an expert radiologist using targeted biopsy results.

STUDY TYPE

Retrospective study including data from our institution and the publicly available ProstateX dataset.

POPULATION

In all, 687 patients who underwent mpMRI of the prostate and had one or more detectable lesions (PI-RADS score >1) according to PI-RADSv2.

FIELD STRENGTH/SEQUENCE

T2 -weighted, diffusion-weighted imaging (DWI; five evenly spaced b values between b = 0-750 s/mm2 ) for apparent diffusion coefficient (ADC) mapping, high b-value DWI (b = 1500 or 2000 s/mm2 ), and dynamic contrast-enhanced T1 -weighted series were obtained at 3.0T.

ASSESSMENT

PI-RADS lesions were segmented by a radiologist. Bounding boxes around the T2 /ADC/high-b value segmentations were stacked and saved as JPEGs. These images were used to train a convolutional neural network (CNN). The PI-RADS scores obtained by the CNN were compared with radiologist scores. The cancer detection rate was measured from a subset of patients who underwent biopsy.

STATISTICAL TESTS

Agreement between the AI and the radiologist-driven PI-RADS scores was assessed using a kappa score, and differences between categorical variables were assessed with a Wald test.

RESULTS

For the 1034 detection lesions, the kappa score for the AI system vs. the expert radiologist was moderate, at 0.40. However, there was no significant difference in the rates of detection of clinically significant cancer for any PI-RADS score in 86 patients undergoing targeted biopsy (P = 0.4-0.6).

DATA CONCLUSION

We developed an AI system for assignment of a PI-RADS score on segmented lesions on mpMRI with moderate agreement with an expert radiologist and a similar ability to detect clinically significant cancer.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY STAGE: 2.

Rapid perceptual processing in two- and three-dimensional prostate images

Radiologists can identify whether a radiograph is abnormal or normal at above chance levels in breast and lung images presented for half a second or less. This early perceptual processing has only been demonstrated in static two-dimensional images (e.g., mammograms). Can radiologists rapidly extract the "gestalt" from more complex imaging modalities? For example, prostate multiparametric magnetic resonance imaging (mpMRI) displays a series of images as a virtual stack and comprises multiple imaging sequences: anatomical information from the T2-weighted (T2W) sequence, functional information from diffusion-weighted imaging, and apparent diffusion coefficient sequences. We first tested rapid perceptual processing in static T2W images then among the two functional sequences. Finally, we examined whether this rapid radiological perception could be observed using T2W multislice imaging. Readers with experience in prostate mpMRI could detect and localize lesions in all sequences after viewing a 500-ms static image. Experienced prostate readers could also detect and localize lesions when viewing multislice image stacks presented as brief movies, with image slices presented at either 48, 96, or 144 ms. The ability to quickly extract the perceptual gestalt may be a general property of expert perception, even in complex imaging modalities.

Correlation of SUVmax and Apparent Diffusion Coefficient Values Detected by Ga-68 PSMA PET/MRI in Primary Prostate Lesions and Their Significance in Lymph Node Metastasis: Preliminary Results of an On-going Study

Objectives:

Gallium-68 (Ga-68) prostate specific membrane antigen (PSMA) positron emission tomography (PET) has been shown to be more accurate than multiparametric prostate magnetic resonance imaging (MRI) in detection of primary prostate lesions. Using hybrid PET/MRI we aim to detect the correlation between SUV_max and apparent diffusion coefficient (ADC) in primary prostate lesions and to assess their prognostic value in detection of lymph node (LN) metastasis.

Methods:

Twenty-six patients, who were diagnosed as having prostate cancer with biopsy and underwent Ga-68 PSMA PET/MRI together with biparametric prostate MRI (bpMRI) were included. SUV_max, SUV_mean and ADC were recorded for index lesions drawing a region of interest (ROI) of 1 cm² around the pixel with the highest SUV_max (ROI-1) and another ROI following borders of prostate tumor detected by bpMRI (ROI-2). Presence of LN metastasis was recorded according to PSMA PET/MRI.

Results:

SUV_max was inversely correlated with ADC (ROI-1: p=0.010; ROI-2: p=0.017 for b=800). SUV_max and SUV_means were both higher in patients with LN metastasis and ADC was lower in patients with LN metastasis for ROI-1. SUV_max cut-off value of 19.8 for ROI-1 and 20.9 for ROI-2 had sensitivity and specificity of 77.8% and 76.5%, respectively for detection of LN metastasis, whereas ADC (b=800) cut-off value of 0.92x10^-3 mm²/s had sensitivity and specificity of 87.5% and 76.5%, respectively. SUV_max/ADC (b=800) ratio increased the sensitivity and specificity to 100% and 82.4%, respectively.

Conclusion:

SUV and ADC values are inversely correlated in primary prostate lesions and the combined use of both values increases the diagnostic accuracy of hybrid PET/MRI in the detection of primary prostate lesions.

The Problems with the Kappa Statistic as a Metric of Interobserver Agreement on Lesion Detection Using a Third-reader Approach When Locations Are Not Prespecified

RATIONALE AND OBJECTIVES

To point out the problems with Cohen kappa statistic and to explore alternative metrics to determine interobserver agreement on lesion detection when locations are not prespecified.

MATERIALS AND METHODS

Use of kappa and two alternative methods, namely index of specific agreement (ISA) and modified kappa, for measuring interobserver agreement on the location of detected lesions are presented. These indices of agreement are illustrated by application to a retrospective multireader study in which nine readers detected and scored prostate cancer lesions in 163 consecutive patients (n = 110 cases, n = 53 controls) using the guideline of Prostate Imaging Reporting and Data System version 2 on multiparametric magnetic resonance imaging.

RESULTS

The proposed modified kappa, which properly corrects for the amount of agreement by chance, is shown to be approximately equivalent to the ISA. In the prostate cancer data, average kappa, modified kappa, and ISA equaled 30%, 55%, and 57%, respectively, for all lesions and 20%, 87%, and 87%, respectively, for index lesions.

CONCLUSIONS

The application of kappa could result in a substantial downward bias in reader agreement on lesion detection when locations are not prespecified. ISA is recommended for assessment of reader agreement on lesion detection.

Radiologists' preferences regarding content of prostate MRI reports: a survey of the Society of Abdominal Radiology

PURPOSE

To evaluate radiologist preferences regarding specific content that warrants inclusion in prostate MRI reports.

METHODS

Sixty-one members of the Society of Abdominal Radiology responded to a 74-item survey regarding specific content warranted in prostate MRI reports, conducted in August 2016.

RESULTS

General items deemed essential report content by ≥ 50% of respondents were prostate volume (80%), extent of prostate hemorrhage (74%), TURP defects (69%), coil type (64%), BPH (61%), contrast dose (61%), contrast agent (59%), medications administered (59%), and magnet strength (54%). Details regarding lesion description deemed essential by ≥ 50% were overall PI-RADS category (88%), DCE (±) (82%), subjective degree of diffusion restriction (72%), T2WI intensity (72%), T2WI margins (65%), T2WI shape (52%), DWI 1-5 score (50%), and T2WI 1-5 score (50%). Details deemed essential to include in the report Impression by ≥ 50% of respondents were lymphadenopathy and metastases (100%), EPE (98%), SVI (98%), neurovascular bundle involvement (93%), index lesion location (93%), PI-RADS category of index lesion (82%), number of suspicious lesions (78%), significance of index lesion PI-RADS category (53%), and PI-RADS category of non-index lesions (52%). Preferred methods for lesion localization were slice/image number (68%), 3-part craniocaudal level (68%), zonal location (65%), anterior vs. posterior location (57%), and medial vs. lateral position (56%). Least preferred methods for localization were numeric sector from the PI-RADS sector map (8%), annotated screen capture (10%), and graphical schematic of PI-RADS sector map (11%).

CONCLUSION

Radiologists generally deemed a high level of detail warranted in prostate MRI reports. The PI-RADS v2 sector map was disliked for lesion localization.

Follow-up of negative MRI-targeted prostate biopsies: when are we missing cancer?

INTRODUCTION

Multiparametric magnetic resonance imaging (mpMRI) has improved clinicians' ability to detect clinically significant prostate cancer (csPCa). Combining or fusing these images with the real-time imaging of transrectal ultrasound (TRUS) allows urologists to better sample lesions with a targeted biopsy (Tbx) leading to the detection of greater rates of csPCa and decreased rates of low-risk PCa. In this review, we evaluate the technical aspects of the mpMRI-guided Tbx procedure to identify possible sources of error and provide clinical context to a negative Tbx.

METHODS

A literature search was conducted of possible reasons for false-negative TBx. This includes discussion on false-positive mpMRI findings, termed "PCa mimics," that may incorrectly suggest high likelihood of csPCa as well as errors during Tbx resulting in inexact image fusion or biopsy needle placement.

RESULTS

Despite the strong negative predictive value associated with Tbx, concerns of missed disease often remain, especially with MR-visible lesions. This raises questions about what to do next after a negative Tbx result. Potential sources of error can arise from each step in the targeted biopsy process ranging from "PCa mimics" or technical errors during mpMRI acquisition to failure to properly register MRI and TRUS images on a fusion biopsy platform to technical or anatomic limits on needle placement accuracy.

CONCLUSIONS

A better understanding of these potential pitfalls in the mpMRI-guided Tbx procedure will aid interpretation of a negative Tbx, identify areas for improving technical proficiency, and improve both physician understanding of negative Tbx and patient-management options.

Risk of Upgrading from Prostate Biopsy to Radical Prostatectomy Pathology-Does Saturation Biopsy of Index Lesion during Multiparametric Magnetic Resonance Imaging-Transrectal Ultrasound Fusion Biopsy Help?

PURPOSE

We sought to determine whether saturation of the index lesion during magnetic resonance imaging-transrectal ultrasound fusion guided biopsy would decrease the rate of pathological upgrading from biopsy to radical prostatectomy.

MATERIALS AND METHODS

We analyzed a prospectively maintained, single institution database for patients who underwent fusion and systematic biopsy followed by radical prostatectomy in 2010 to 2016. Index lesion was defined as the lesion with largest diameter on T2-weighted magnetic resonance imaging. In patients with a saturated index lesion transrectal fusion biopsy targets were obtained at 6 mm intervals along the long axis of the index lesion. In patients with a nonsaturated index lesion only 1 target was obtained from the lesion. Gleason 6, 7 and 8-10 were defined as low, intermediate and high risk, respectively.

RESULTS

Included in the study were 208 consecutive patients, including 86 with a saturated and 122 with a nonsaturated lesion. Median patient age was 62.0 years (IQR 10.0) and median prostate specific antigen was 7.1 ng/ml (IQR 8.0). The median number of biopsy cores per index lesion was higher in the saturated lesion group (4 vs 2, p <0.001). The risk category upgrade rate from systematic only, fusion only, and combined fusion and systematic biopsy results to prostatectomy was 40.9%, 23.6% and 13.8%, respectively. The risk category upgrade from combined fusion and systematic biopsy results was lower in the saturated than in the nonsaturated lesion group (7% vs 18%, p = 0.021). There was no difference in the upgrade rate based on systematic biopsy between the 2 groups. However, fusion biopsy results were significantly less upgraded in the saturated lesion group (Gleason upgrade 20.9% vs 36.9%, p = 0.014 and risk category upgrade 14% vs 30.3%, p = 0.006).

CONCLUSIONS

Our results demonstrate that saturation of the index lesion significantly decreases the risk of upgrading on radical prostatectomy by minimizing the impact of tumor heterogeneity.

Changes in prostate cancer detection rate of MRI-TRUS fusion vs systematic biopsy over time: evidence of a learning curve

BACKGROUND

To determine the effect of urologist and radiologist learning curves and changes in MRI-TRUS fusion platform during 9 years of NCI's experience with multiparametric magnetic resonance imaging (mpMRI)/TRUS fusion biopsy.

METHODS

A prospectively maintained database of patients undergoing mpMRI followed by fusion biopsy (Fbx) and systematic biopsy (Sbx) from 2007 to 2016 was reviewed. The patients were stratified based on the timing of first biopsy. Cohort 1 (7/2007-12/2010) accounted for learning curve. Cohort 2 (1/2011-5/2013) and cohort 3 (5/2013-4/2016) included patients biopsied prior to and after debut of a new software platform, respectively. Clinically significant (CS) disease was defined as Gleason 7 (3+4) or higher. McNemar's test compared cancer detection rates (CDRs) of Sbx and Fbx between time periods.

RESULTS

1528 patients were included in the study with 230, 537 and 761 patients included in three respective cohorts. Median age (interquartile range) was 61.0 (±9.0), 62.0 (±7.3), and 64.0 (±11.0) years in three cohorts, respectively (P<0.001). Fbx and Sbx had comparable CS CDR in cohort 1 (24.8 vs 22.2%, P=0.377). Fbx detected significantly more CS disease compared to Sbx in the following two periods (cohort 2: 31.5 vs 25.0%, P=0.001; cohort 3: 36.4 vs 30.3%, P<0.001) and detected significantly less low risk disease in the same period (cohort 2: 14.5 vs 19.6%, P<0.001; cohort 3: 12.6 vs 16.7%, P<0.001). Even after multivariate adjustment with age, PSA, race, clinical stage and MRI suspicion score, Fbx CS cancer detection increased in successive cohorts (cohort 2: OR 2.23, P=0.043; cohort 3: OR 2.92, P=0.007).

CONCLUSIONS

In the past 9 years, there has been significant improvement in the accuracy of Fbx. Our results show that after an early learning period, Fbx detected higher rates of CS cancer and lower rates of clinically insignificant cancer than Sbx. Software advances allowed for even greater detection of CS disease.