Usefulness of Functional MRI Textures in the Evaluation of Renal Function

Abstract

Non-invasive assessment of kidney function and structure remains of clinical importance in the diagnosis and prognosis of chronic kidney disease. This work aims to evaluate the role of textures extracted from functional magnetic resonance imaging in renal dysfunction detection by differentiating healthy and chronic kidney disease patients. Textural descriptors are extracted from apparent diffusion coefficient, blood oxygenation level dependent images and T2 maps. Synthetic resampling technique is performed to account for imbalanced classes and increase the variety of sample domain. Principal component analysis projection is applied to eliminate irrelevant features and compact the dataset. The performance of linear discriminant analysis, logistic regression and Naïve Bayes classifiers in terms of discriminating healthy and affected kidney is evaluated. The results of this preliminary study support the fact that chronic kidney disease affects texture parameters significantly. Textures-based predictive models have shown promise in accurate and safe renal function evaluation (accuracy, sensitivity and AUC up to 98%, 98% and 1 respectively).