Automatic classification of prostate cancer Gleason scores from multiparametric magnetic resonance images

摘要: Noninvasive, radiological image-based detection and stratification of Gleason patterns can impact clinical outcomes, treatment selection, the determination disease status at diagnosis without subjecting patients to surgical biopsies. We present machine learning-based automatic classification prostate cancer aggressiveness by combining apparent diffusion coefficient (ADC) T2-weighted (T2-w) MRI-based texture features. Our approach achieved reasonably accurate scores (GS) 6(3 + 3) vs. ≥7 7(3 4) 7(4 despite presence highly unbalanced samples using two different sample augmentation techniques followed feature selection-based classification. method distinguished between GS cancers with 93% accuracy for occurring in both peripheral (PZ) transition (TZ) zones 92% PZ alone. from TZ In comparison, a classifier only ADC mean top 58% distinguishing 63% The same an 59% 60% Separate analysis alone was not performed owing limited number samples. results suggest that features derived T2-w MRI together help obtain patterns.