Radiomics for peripheral zone and intra-prostatic urethra segmentation in MR imaging

Praful Hambarde, Sanjay N. Talbar, Nilesh Sable, Abhishek Mahajan, Satishkumar S. Chavan, Meenakshi Thakur

Abstract: Highlights•The paper presents a novel approach of automatic segmentation of peripheral zone (PZ) and intra-prostatic urethra in T2w MR images.•The technique aids the radiologists for accurate extraction of PZ and precise localization of intra-prostatic urethra.•Nonnegative matrix factorization (NMF) is preferred to extract radiomic features of PZ and urethra from prostate region which are used for segmentation using unsupervised learning through self organizing maps (SOMs).•The segmentation results are evaluated using dice similarity coefficient (DSC) and validated by radiologists. It is found that the proposed approach of the segmentation provides better DSC and subjective score compared to K-means clustering and fuzzy C-means clustering techniques.