Population-Specific Radiomics From Biparametric Magnetic Resonance Imaging Improves Prostate Cancer Risk...

Area of Science:

• Radiology
• Oncology
• Medical Imaging Analysis

Background:

• Prostate cancer (PCa) presentation varies significantly across different racial populations.
• Accurate risk stratification is crucial for effective PCa management.

Purpose of the Study:

• To quantify population-specific differences in PCa presentation between African American (AA) and White (W) men using MRI-based radiomics.
• To evaluate the efficacy of population-specific radiomic models for PCa risk stratification.

Main Methods:

• 149 men with PCa underwent 3T MRI; radiomic features were extracted from PCa regions of interest.
• Machine learning models were trained separately for AA and W men to distinguish clinically significant (csPCa) from insignificant (ciPCa) PCa.
• Models were validated and compared against a population-agnostic model, with and without clinical parameters.

Main Results:

• Radiomic features differed between AA and W men, particularly in the peritumoral region.
• Population-specific radiomic models outperformed the population-agnostic model for both AA and W men (AUCs 0.84 vs 0.57 for AA; 0.71 vs 0.60 for W).
• Integration of clinical and radiomic data further improved risk stratification for both groups (AUCs 0.90 for AA; 0.75 for W).

Conclusions:

• Radiomic features on MRI show population-specific differences in PCa presentation.
• Population-specific radiomic models enhance PCa risk stratification compared to population-agnostic approaches.
• Tailoring radiomic analysis to specific populations may improve diagnostic accuracy for PCa.