A comprehensive non‐invasive framework for automated evaluation of acute renal transplant rejection using DCE‐MRI

Abstract

The objective was to develop a novel and automated comprehensive framework for the non-invasive identification and classification of kidney non-rejection and acute rejection transplants using 2D dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI). The proposed approach consists of four steps. First, kidney objects are segmented from the surrounding structures with a geometric deformable model. Second, a non-rigid registration approach is employed to account for any local kidney deformation. In the third step, the cortex of the kidney is extracted in order to determine dynamic agent delivery, since it is the cortex that is primarily affected by the perfusion deficits that underlie the pathophysiology of acute rejection. Finally, we use an analytical function-based model to fit the dynamic contrast agent kinetic curves in order to determine possible rejection candidates. Five features that map the data from the original data space to the feature space are chosen with a k-nearest-neighbor (KNN) classifier to distinguish between acute rejection and non-rejection transplants. Our study includes 50 transplant patients divided into two groups: 27 patients with stable kidney function and the remainder with impaired kidney function. All of the patients underwent DCE-MRI, while the patients in the impaired group also underwent ultrasound-guided fine needle biopsy. We extracted the kidney objects and the renal cortex from DCE-MRI for accurate medical evaluation with an accuracy of 0.97 ± 0.02 and 0.90 ± 0.03, respectively, using the Dice similarity metric. In a cohort of 50 participants, our framework classified all cases correctly (100%) as rejection or non-rejection transplant candidates, which is comparable to the gold standard of biopsy but without the associated deleterious side-effects. Both the 95% confidence interval (CI) statistic and the receiver operating characteristic (ROC) analysis document the ability to separate rejection and non-rejection groups. The average plateau (AP) signal magnitude and the gamma-variate model functional parameter α have the best individual discriminating characteristics.

Developing advanced mathematical models for detecting abnormalities in 2D/3D medical structures.

“…• A future work of this dissertation is to investigate the integration of the proposed work with the BioImaging lab work for the early detection of acute renal rejection [128,129,[288][289][290][291][292][293][294][295][296][297][298][299][300][301][302].…”

A novel NMF-based DWI CAD framework for prostate cancer.

“…For example, Ghose et al [181] proposed a probabilistic graph-cut-based framework for 3D T2-MRI prostate segmentation based on a probabilistic atlas. Firjany et al [147] proposed a Markov random field (MRF) image model [182][183][184][185][186][187][188][189][190][191][192][193][194][195][196] for 2D DCE-MRI prostate segmentation that combined a graph-cut approach with a prior shape model of the prostate and the visual appearance of the prostate image, modeled using a linear combination of discrete Gaussian (LCDG) [197][198][199][200][201][202][203][204][205][206][207][208] Their method was later extended in [209,210] to allow for 3D prostate segmentation from DCE-MRI volumes. The main limitation of graph-cut techniques is that they are prone to minimizing the size of the segmented region [211].…”

Fast and robust hybrid framework for infant brain classification from structural MRI : a case study for early diagnosis of autism.

“…• Another future work will be to apply the developed models in other clinical applications such as: acute renal rejection [144][145][146][147][148][149][150][151][152][153], lung cancer detection , and cancerous cells detection in the prostate [188][189][190][191][192][193][194].…”