Detection of Renal Rejection after Kidney Transplantation
|
|
Acute rejection is the most important reason of graft failure after kidney transplantation, and early detection is crucial to survive the kidney function. This project aims to combine the information from Ultrasound and Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) modalities to early detect the acute rejection. Our current studies are focused on the Dynamic Contrast Enhanced Magnetic Resonance Imaging results. DCE-MRI is an imaging modality where a contrast agent (such as Gd-DTPA) is introduced into the kidney and rapid and repeated images are taken, which allows both the anatomic and functional information of the kidney to be monitored. Using these images, we are aiming to follow the intensity changes in the cortex of the kidney to classify the acute rejected and accepted kidneys using their patterns of contrast agent intake. To allow this classification, the abdomen images should be registered to each other to cancel the motion of the kidneys during breathing, then the kidneys need to be segmented from the surrounding structures, and finally, the perfusion curves obtained from the cortex that show the transportation of the contrast agent into the tissue should be used to classify the normal and acute rejection transplants. |
Research Team
MethodsOur image analysis approach consists of three major steps:
- Motion compensation using rigid registration based on Mutual Information;
- Segmentation of kidney from the DCE-MRI scans;
- Generation of the perfusion curve from the cortex of the kidneys.
The most difficult part in these three steps is the segmentation of the kidney from the DCE-MRI scans because of the non-uniformly changing contrast and the low-resolution in the images. To overcome this problem, we have proposed an accurate method by introducing a shape model of the kidney into the external energy component of the deformable models. The proposed external energy component is a multiplication of two densities: the gray level and the signed distance map density of the shape model. These two densities are approximated using our novel modified EM algorithm; therefore, the deformable model moves with both the gray level and the shape model information depending on a Bayesian classification.
Results
In this study, gradient echo T-1 imaging is employed by a Signa Horizon GE 1.5T scanner and the contrast agent Gadolinium DTPA is introduced via a wide bore veno-catheter placed at antecubital vein at a rate of 3-4 ml/sec with a dose of 0.2 ml/kg.BW. Images are taken at 5mm thickness with no interslice gap, repetition time (TR) 34 msec, field of view (FOV) 42x42 cm and the matrix is 256x160. For each patient, 12 temporal sequences of coronal scans are taken, each sequence consisting of 6 images. The first image is taken at pre contrast, second image is at 0 second of injection and then the imaging is repeated every 30 sec for 10 times with the last image acquired 15 minutes after the start of injection.
An example of the segmentation results is given in Fig.1. From these results, for two patients, two perfusion curves are obtained as shown in Fig. 2 by the manual selection of a window in the cortex. Although the perfusion curves of accepted and rejected kidneys of two patients in Fig. 2 are easily distinguishable, we haven’t been able to come to a general solution in detecting acute rejection especially considering the fact that there are other diseases such as acute tubular necrosis or polyoma virus infection which look like rejection. Moreover, because of the non-uniform behavior of the kidney in the contrast agent uptake, plots of the perfusion curves depend highly on the manual window selection and therefore discriminating the accepted and rejected kidneys has been left as an open question, which will constitute our future work.
Fig. 1. Segmentation Results
Fig. 2. Perfusion curve of a patient with biopsy proven acute rejection compared to an accepted kidney
Publications
Acknowledgement
/ Sponsors