3d Diffusion Mri-Based Cad System For Early Diagnosis Of Acute Renal Rejection

This paper introduces a novel computer aided diagnostic (CAD) system for early diagnosis of acute renal rejection (ARR) from 3D diffusion-weighted MRI (DW-MRI) data (2D + b-value). The proposed CAD system starts with aligning the DW-MRI data using B-splines based registration to handle the motion effects, which come from breathing and heart beats. This is followed by the segmentation of the kidney tissue using geometric (level-sets based) deformable model, which is guided by a new stochastic speed relationship that takes into account an adaptive kidney shape prior and the visual appearance of the kidney. The pixel-wise guidance of the level-sets is obtained by integrating these image features into a joint Markov-Gibbs random field (MGRF) model of the kidney and its background. The final step of the proposed CAD system is to calculate the apparent diffusion coefficients (ADCs) between different b-values of the segmented DW-MRI data to distinguish between rejection and nonrejection renal transplants. Experimental results on 36 subjects, using a KStar classifier and leave-one-subject-out, have classified 87% of the subjects correctly (26 out of 30 rejection kidneys and 5 out of 6 nonrejection kidneys). These initial diagnostic results hold promise of the proposed CAD system as a reliable non-invasive diagnostic tool.