Abstract:Introduction and objectives: The Prostate Imaging Reporting and Data System (PI-RADS) score reports the likelihood of a clinically significant prostate cancer (CsPCa) based on various multiparametric prostate magnetic resonance imaging (mpMRI) characteristics. The PI-RADS category 3 is an intermediate status, with an equivocal risk of malignancy. The PSA density (PSAD) has been proposed as a tool to facilitate biopsy decisions on PI-RADS category 3 lesions. The objective of this study is to determine the frequency of CsPCa, assess the diagnostic value of targeted biopsy and identify clinical predictors to improve the CsPCa detection rate in PI-RADS category 3 lesions. Methods: Between 1st January 2017 and 31st December 2022, a total of 1661 men underwent a prostate biopsy at our institution. Clinical and mpMRI data of men with PI-RADS 3 lesions was reviewed. The study population was divided into two groups: target group, including those submitted to systematic plus targeted biopsy versus non-target group when only systematic or saturation biopsy were performed. Patients with PI-RADS 3 lesions were divided into three categories based on pathological biopsy results: benign, clinically insignificant disease (score Gleason = 6 or International Society of Urologic Pathologic (ISUP) 1) and clinically significant cancer (score Gleason ≥ 7 (3+4) or ISUP ≥ 2) according to target and non-target group. Univariate and multivariate analyses were performed to identify clinical predictors to improve the CsPCa detection rate in PI-RADS category 3 lesions. Results: A total of 130 men with PIRADS 3 index lesions were identified. Pathologic results were benign in 77 lesions (59.2%), 19 (14.6%) were clinically insignificant (Gleason score 6) and 34 (26.2%) were clinically significant (Gleason score 7 or higher). Eighty-seven of the patients were included in the target group (66.9%) and 43 in the non-target group (33.1%). The CsPCa detection was higher in the non-target group (32.6%, n = 14 vs 23.0%, n = 20 respectively). When systematic and target biopsies were jointly performed, if the results of systematic biopsies are not considered and only the results of target biopsies are taken into account, a CsPCa diagnosis would be missed on 9 patients. The differences of insignificant cancer and CsPCa rates among the target or non-target group were not statistically significant (p = 0.50 and p = 0.24, respectively). on multivariate analysis, the abnormal DRE and lesions localized in Peripheral zone (PZ) were significantly associated with a presence of CsPCa in PI-RADS 3 lesions (oR = 3.61, 95% CI [1.22,10.72], p = 0.02 and oR = 3.31, 95% CI [1.35, 8.11], p = 0.01, respectively). A higher median PSAD significantly predisposed for CsPCa on univariate analyses (p = 0.05), however, was not significant in the multivariate analysis (p = 0.76). In our population, using 0.10 ng/ml/ml as a cut-off to perform biopsy, 41 patients would have avoided biopsy (31.5%), but 5 cases of CsPCa would not have been detected (3.4%). We could not identify any statistical significance between other clinical and imagiological variables and CsPCa detection. Conclusions: PI-RADS 3 lesions were associated with a low likelihood of CsPCa detection. A systematic biopsy associated or not with target biopsy is essential in PI-RADS 3 lesions, and targeted biopsy did not demonstrate to be superior in the detection of CsPCa. The presence of abnormal DRE and lesions localized in PZ potentially predict the presence of CsPCa in biopsied PI-RADS 3 lesions.

Machine Learning and Clinical-Radiological Characteristics for the Classification of Prostate Cancer in PI-RADS 3 Lesions

Clinico-radiological characteristic-based machine learning in reducing unnecessary prostate biopsies of PI-RADS 3 lesions with dual validation

Prostate Imaging-Reporting and Data System Version 2 in Combination with Clinical Parameters for Prostate Cancer Detection: a Single Center Experience

Machine Learning for the Prediction of Prostate Cancer Biopsy Based on 3D Dynamic Contrast-Enhanced Ultrasound Quantification

Navigating the gray zone: Machine learning can differentiate malignancy in PI-RADS 3 lesions

Development and Validation of Interpretable Machine Learning Models for Clinically Significant Prostate Cancer Diagnosis in Patients With Lesions of PI‐RADS v2.1 Score ≥3

Machine and Deep Learning Prediction Of Prostate Cancer Aggressiveness Using Multiparametric MRI

Machine learning-based analysis of a semi-automated PI-RADS v2.1 scoring for prostate cancer

Utility of Clinical-Radiomic Model to Identify Clinically Significant Prostate Cancer in Biparametric MRI PI-RADS V2.1 Category 3 Lesions

Machine learning models for discriminating clinically significant from clinically insignificant prostate cancer using bi-parametric magnetic resonance imaging

Enhancing Prostate Cancer Detection in PI-RADS 3 Cases: An In-depth Analysis of Radiological Indicators from Multiparametric MRI

Texture analysis based on PI-RADS 4/5-scored magnetic resonance images combined with machine learning to distinguish benign lesions from prostate cancer

MP49-02 DEVELOPMENT AND EXTERNAL VALIDATION OF INTERPRETABLE MACHINE LEARNING MODELS FOR CLINICALLY SIGNIFICANT PROSTATE CANCER DIAGNOSIS IN PATIENTS WITH LESIONS OF PI-RADS V2.1 SCORE ≥3

Magnetic resonance imaging radiomics-based prediction of clinically significant prostate cancer in equivocal PI-RADS 3 lesions in the transitional zone

Refining clinically relevant cut-offs of prostate specific antigen density for risk stratification in patients with PI-RADS 3 lesions

Challenges in the Use of Artificial Intelligence for Prostate Cancer Diagnosis from Multiparametric Imaging Data

Biparametric MRI-based radiomics classifiers for the detection of prostate cancer in patients with PSA serum levels of 4∼10 ng/mL

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

A Novel Machine Learning-Based Predictive Model of Clinically Significant Prostate Cancer and Online Risk Calculator

PI-RADS Score of Prostate Multi-Parametric Magnetic Resonance Imaging Predicts Pathological Features of Prostate Cancer:An Imaging to Whole-Mount Pathology Matching Study

Predictors of prostate cancer detection in MRI PI-RADS 3 lesions - Reality of a tertiary center