Promising Results About the Possibility to Identify Prostate Cancer Patients Employing a Random Forest Classifier: A...

Area of Science:

• Radiology and Oncology
• Artificial Intelligence in Medicine
• Medical Imaging Analysis

Background:

• Accurate prediction of lymph node involvement is crucial for prostate cancer (PCa) staging and treatment planning.
• Current methods may not fully capture the metastatic potential of PCa, necessitating advanced diagnostic tools.

Purpose of the Study:

• To evaluate the accuracy of a Random Forest (RF) machine learning model in predicting lymph node metastasis in PCa using MRI-derived radiomic features.
• To identify key radiomic features and MRI sequences that contribute to accurate lymph node status prediction.

Main Methods:

• Analysis of mp-MRI data from 95 PCa patients who underwent prostatectomy and lymphadenectomy.
• Extraction of radiomic features from T2-weighted, Diffusion-Weighted Imaging (DWI), and Apparent Diffusion Coefficient (ADC) sequences.
• Application of a Random Forest model, incorporating clinical data (PSA, Gleason score) and radiomic features.

Main Results:

• The RF model achieved 84% accuracy in the peripheral zone and 87% in the transitional zone for predicting lymph node involvement.
• Key predictive features included ADC shape, T2 noduloglcm, DWI glcm, and first-order features for whole-gland analysis.
• DWI and ADC sequences were identified as particularly important for lymph node assessment.

Conclusions:

• AI-driven radiomic analysis using DWI and ADC MRI sequences effectively predicts lymph node involvement in PCa.
• This approach shows promise for preoperative patient selection, particularly for identifying those with negative lymph node status, enabling lymph node-sparing strategies.
• Further validation in larger patient cohorts is recommended due to the study's retrospective nature and limited sample size.