Automatic Characterization of Prostate Suspect Lesions on T2-Weighted Image Acquisitions Using Texture Features and...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Oncology and Cancer Research

Background:

• Prostate cancer (PCa) is a common male neoplasm with varied prognoses.
• Multiparametric MRI (mpMRI) aids PCa assessment but lacks direct histopathological correlation.
• AI and radiomics show promise in bridging the gap between imaging and tumor grading.

Purpose of the Study:

• To develop a machine learning algorithm for predicting International Society of Urological Pathology (ISUP) grades of prostate nodules.
• To classify prostate nodules into clinically significant and indolent groups using T2-weighted MRI.
• To leverage radiomics and AI for improved PCa prognostication.

Main Methods:

• Manual segmentation of 76 prostate nodules from T2-weighted MRI in 55 patients.
• Extraction of radiomic features using PyRadiomics (version 3.0.1).
• Development and application of machine learning classifiers for ISUP grade prediction and classification.

Main Results:

• The developed AI algorithm achieved 87.2% accuracy in classifying indolent versus clinically significant PCa.
• The algorithm demonstrated 80.3% accuracy when differentiating specific ISUP grade groups.
• A high proportion (85.52%) of nodules were PI-RADS 4 or higher, indicating significant disease.

Conclusions:

• An AI-based decision-support system was successfully developed.
• The system accurately differentiates PCa prognostic groups using only T2 MRI and radiomics.
• This approach offers a robust tool for PCa assessment and treatment planning.