Early identification of acute rejection for renal allografts: a machine learning approach

Abstract

The goal of this chapter is to explore the developed computer-aided diagnostic (CAD) system that helps to determine the functionality of renal transplant using diffusion-weighted magnetic resonance imaging (DW-MRI). This study is based on the integration of DW-MRI-based biomarkers with clinical-based biomarkers. The data are acquired at different durations and strengths of the magnetic field (b-values). The DW-MRI data were collected from Egypt and the United States, using different scanners such as GE and Philips. Using a level-set approach, the kidney is first segmented, then estimates the apparent diffusion coefficients (ADCs). Then serum creatinine and creatinine clearance (the clinical biomarkers that we used) are combined with the ADCs, creating new image markers known as integrated ADCs (IADCs). Finally, the IADCs make cumulative distribution functions at the different b-values. Lastly, any classifier may be used to distinguish between nonrejection and acute rejection renal transplant status. More importantly, our proposed CAD system exhibits 93% accuracy, 93% sensitivity, and 92% specificity in distinguishing between the renal transplant status. This concluded that the developed CAD system is completely independent of classifiers, geographical areas, and scanner type. These results support that the developed CAD system might be noninvasively able to assess the status of renal allograft dysfunction.