# MRI-Based Deep Learning Algorithm Detects csPCa Comparable to Radiologists

A new study published in Radiology suggests that deep learning assessment of prostate biparametric MRI scans can be as effective as assessment by experienced genitourinary radiologists in detecting clinically significant prostate cancer (csPCa). The research compared the performance of a deep learning algorithm and a radiologist in diagnosing csPCa in 658 men with a total of 1,029 MRI-visible lesions. The deep learning algorithm detected csPCa in 96% of cases compared to 98% by the radiologist. The algorithm also showed comparable sensitivity and positive predictive value to the radiologist. The researchers believe that AI can standardize lesion detection and reduce variability in diagnosis, potentially enhancing consistency across different radiologists and healthcare settings. The algorithm focuses on central lesion regions but also filters out false positives beyond tumor foci with a benign prostatic hyperplasia filter. The study highlights the algorithm’s efficiency in detecting high-grade lesions and potential for automation in lesion segmentation, biopsy, and radiation therapy planning. Limitations of the study include inconsistent MRI acquisition affecting the model’s performance. Overall, the findings suggest that deep learning algorithms could serve as valuable decision support tools for radiologists in diagnosing csPCa and improving patient management.

Source link

Source link: https://www.diagnosticimaging.com/view/mri-based-deep-learning-algorithm-comparable-detection-cspca-radiologists