Artificial intelligence for prediction of donor liver allograft steatosis and early post-transplantation graft failure.

BACKGROUND: Donor livers undergo subjective pathologist review of steatosis before transplantation to mitigate the risk for early allograft dysfunction (EAD). We developed an objective, computer vision artificial intelligence (CVAI) platform to score donor liver steatosis and compared its capability for predicting EAD against pathologist steatosis scores. METHODS: Two pathologists scored digitized donor liver biopsy slides from 2014 to 2019. We trained four CVAI platforms with 1:99 training:prediction split. Mean intersection-over-union (IU) characterized CVAI model accuracy. We defined EAD using liver function tests within 1 week of transplantation. We calculated separate EAD logistic regression models with CVAI and pathologist steatosis and compared the models' discrimination and internal calibration. RESULTS: From 90 liver biopsies, 25,494 images trained CVAI models yielding peak mean IU = 0.80. CVAI steatosis scores were lower than pathologist scores (median 3% vs 20%, P < 0.001). Among 41 transplanted grafts, 46% developed EAD. The median CVAI steatosis score was higher for those with EAD (2.9% vs 1.9%, P = 0.02). CVAI steatosis was independently associated with EAD after adjusting for donor age, donor diabetes, and MELD score (aOR = 1.34, 95%CI = 1.03-1.75, P = 0.03). CONCLUSION: The CVAI steatosis EAD model demonstrated slightly better calibration than pathologist steatosis, meriting further investigation into which modality most accurately and reliably predicts post-transplantation outcomes.

Development of a portable device to quantify hepatic steatosis in potential donor livers.

An accurate estimation of liver fat content is necessary to predict how a donated liver will function after transplantation. Currently, a pathologist needs to be available at all hours of the day, even at remote hospitals, when an organ donor is procured. Even among expert pathologists, the estimation of liver fat content is operator-dependent. Here we describe the development of a low-cost, end-to-end artificial intelligence platform to evaluate liver fat content on a donor liver biopsy slide in real-time. The hardware includes a high-resolution camera, display, and GPU to acquire and process donor liver biopsy slides. A deep learning model was trained to label and quantify fat globules in liver tissue. The algorithm was deployed on the device to enable real-time quantification and characterization of fat content for transplant decision-making. This information is displayed on the device and can also be sent to a cloud platform for further analysis.