Radiomics Driven Diffusion Weighted Imaging Sensing Strategies for Zone-Level Prostate Cancer Sensing

Area of Science:

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

Background:

• Prostate cancer is a common diagnosis in North American men, necessitating efficient and reliable sensing methods.
• Diffusion weighted imaging (DWI) is a non-invasive technique gaining prominence for cancer detection.
• Apparent diffusion coefficient (ADC) imaging and computed high-b value DWI (CHB-DWI) are key DWI modalities.

Purpose of the Study:

• To evaluate the efficacy of ADC and CHB-DWI in zone-level prostate cancer sensing.
• To introduce novel radiomics-driven sensing strategies, including hand-engineered (Zone-X) and deep learning-based (Zone-DR) approaches.
• To compare these new strategies against traditional clinical heuristics.

Main Methods:

• Development of Zone-X sensing strategies using hand-engineered radiomic sequences from DWI data.
• Proposal of Zone-DR, a deep radiomic sequencer for discovering novel radiomic sequences.
• Analysis of 12,466 pathology-verified prostate zones from 101 patients using ADC and CHB-DWI.

Main Results:

• Zone-X and Zone-DR significantly outperformed traditional clinical heuristics in terms of Area Under the Curve (AUC).
• Zone-DR and Zone-SVM achieved the highest sensitivity and specificity for ADC, respectively.
• Zone-DR and Zone-LR demonstrated the highest sensitivities for CHB-DWI, with Zone-DR, Zone-LR, and Zone-SVM showing the highest specificities.

Conclusions:

• Radiomics-driven strategies, particularly Zone-DR, offer significant improvements for zone-level prostate cancer sensing using DWI.
• The choice of strategy can be optimized to balance sensitivity and specificity based on clinical needs (e.g., screening vs. surgical planning).
• Explainability methods applied to Zone-DR provide insights into critical biomarkers for prostate cancer detection.