Radiomics based Machine Learning Models for Classification of Prostate Cancer Grade Groups from Multi Parametric MRI...

Area of Science:

• Radiology and Oncology
• Artificial Intelligence in Medicine

Background:

• Prostate cancer grading is crucial for treatment decisions.
• Accurate noninvasive grading methods are needed to complement histopathology.

Purpose of the Study:

• To evaluate multiparametric MRI (mpMRI) radiomic features for machine learning (ML)-based classification of prostate cancer Gleason grade groups (GG).

Main Methods:

• Retrospective analysis of 203 prostate cancer patients' mpMRI data.
• Extraction of radiomic features from T2-weighted and diffusion-weighted images.
• Development and evaluation of ML models combining feature selection and classifiers.

Main Results:

• A model using recursive feature elimination (RFE) and random forest on high b-value diffusion-weighted MRI features achieved 97.0% accuracy.
• This model also demonstrated 98.0% sensitivity, 98.0% precision, 97.0% F1-measure, and 98% AUC for classifying five Gleason grade groups.

Conclusions:

• Preoperative mpMRI radiomic analysis using ML is a promising noninvasive tool for prostate cancer grading.
• The developed radiomic model offers high accuracy for multiclass grading of prostate cancer.