Background: The quality of donor’s kidney is one of the crucial elements for Kidney Transplantation. Donor factors including age, height, weight, history of hypertension, history of diabetes, causes of death, serum creatinine, these factors are used to calculate Kidney Donor Profile Index (KDPI), are known to be associated with kidney graft quality. In addition to KDPI, kidney biopsies at the procurement remain the important role to assess the kidney graft quality, associated with graft function. While histopathological assessment for the kidney is important, but traditional histopathological analysis is labor-intensive, subject to intra- or inter-observer variability, and limited by the availability of expert nephropathologists. The artificial intelligence (AI) technologies, particularly using deep learning technology, have demonstrated transformative potential in medical imaging field, however, AI applications in kidney pathology remain underexplored.
Methods: Retrospectively we collect the donor kidney biopsy images in our archives. We annotate the following characteristics in the kidney biopsy: glomeruli, sclerotic glomeruli, nodular mesangial glomerulosclerosis, interstitial fibrosis, vascular disease, arteriolar hyalinosis, cortical necrosis and fibrin thrombi. Using a deep learning technology, we create an AI tool to analyze these pathological characteristics in the kidney biopsy. These results will be available in the secure cloud platform.
Results: We have successfully developed the AI model to analyze the kidney biopsies. We have calculated the number of glomeruli and percentage of sclerotic glomeruli. These results have been shared in the secure cloud platform and help the decisions of OPO and transplant centers.
Conclusions: Our new kidney donor pathology tool would improve the workflow of OPO and transplant centers. Further validation would be necessary to analyze the kidney graft outcomes based on our results.