Hybrid Apparent Diffusion Coefficient (HADC) Map

Samira Masoudi, Stephanie A. Harmon, Stephanie Walker, Sherif Mehralivand, Ulas Bagci, Peter L. Choyke, Baris Turkbey

Abstract: Multiparametric MRI (mpMRI) is an established framework for prostate cancer assessment which includes T2-weighted magnetic resonance (T2w-MR) and diffusion weighted (DW) sequences. Low quality of Apparent Diffusion Coefficient (ADC) maps from the diffusion sequence can hinder such clinical assessment. Herein, we propose to generate Hybrid ADC (HADC) maps from high-quality T2w-MRI using “lesion-aware cycle-consistent generative adversarial network (LA-CGAN)”. Our produced HADC maps contain anatomical information from T2w-MR, and high intra-prostatic contrast of cancerous vs normal tissue, similar to the acquired ADC. Initial results have satisfied the expert radiologist in producing HADC. This work can considerably improve the quality of mpMRI combined assessment for prostate cancer detection.

Paper Type: methodological development

TL;DR: We proposed to generate Hybrid Apparent Diffusion Coefficient (HADC) maps from T2-weighted magnetic resonance (T2w-MR) using lesion-aware cycle-consistent generative adversarial network (LAC-GAN).

Track: short paper

Keywords: Prostate cancer detection, Hybrid ADC map, lesion-aware Cycle-GAN