Utility of an Artificial Intelligence Enabled Electrocardiogram for Risk Assessment in Liver Transplant Candidates

• Published in: Digestive diseases and sciences Vol. 68; no. 6; pp. 2379 - 2388
• Main Authors: Zaver, Himesh B., Mzaik, Obaie, Thomas, Jonathan, Roopkumar, Joanna, Adedinsewo, Demilade, Keaveny, Andrew P., Patel, Tushar
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-06-2023; Springer; Springer Nature B.V
• Subjects: Adult; Algorithms; Artificial Intelligence; Atrial fibrillation; Atrial Fibrillation - complications; Biochemistry; Cardiac arrhythmia; Electrocardiogram; Electrocardiography; Gastroenterology; Health aspects; Hepatology; Humans; Liver; Liver diseases; Liver Transplantation - adverse effects; Liver transplants; Medical research; Medicine; Medicine & Public Health; Medicine, Experimental; Mortality; Oncology; Original; Original Article; Performance evaluation; Prognosis; Retrospective Studies; Risk Assessment; Stroke Volume; Transplant Surgery; Transplantation; Ventricular Dysfunction, Left - complications; Ventricular Function, Left; End stage liver disease; Artificial Intelligence enabled electrocardiogram; Artificial intelligence; Convolutional neural network; Receiver operator characteristic; Liver transplantation
• ISSN: 0163-2116; 1573-2568
• DOI: 10.1007/s10620-023-07928-y

Background
Post-operative cardiac complications occur infrequently but contribute to mortality after liver transplantation (LT). Artificial intelligence-based algorithms based on electrocardiogram (AI-ECG) are attractive for use during pre-operative evaluation to screen for risk of post-operative cardiac complications, but their use for this purpose is unknown.
Aims
The aim of this study was to evaluate the performance of an AI-ECG algorithm in predicting cardiac factors such as asymptomatic left ventricular systolic dysfunction or potential for developing post-operative atrial fibrillation (AF) in cohorts of patients with end-stage liver disease either undergoing evaluation for transplant or receiving a liver transplant.
Methods
A retrospective study was performed in two consecutive adult cohorts of patients who were either evaluated for LT or underwent LT at a single center between 2017 and 2019. ECG were analyzed using an AI-ECG trained to recognize patterns from a standard 12-lead ECG which could identify the presence of left ventricular systolic dysfunction (LVEF < 50%) or subsequent atrial fibrillation.
Results
The performance of AI-ECG in patients undergoing LT evaluation is similar to that in a general population but was lower in the presence of prolonged QTc. AI-ECG analysis on ECG in sinus rhythm had an AUROC of 0.69 for prediction of de novo post-transplant AF. Although post-transplant cardiac dysfunction occurred in only 2.3% of patients in the study cohorts, AI-ECG had an AUROC of 0.69 for prediction of subsequent low left ventricular ejection fraction.
Conclusions
A positive screen for low EF or AF on AI-ECG can alert to risk of post-operative cardiac dysfunction or predict new onset atrial fibrillation after LT. The use of an AI-ECG can be a useful adjunct in persons undergoing transplant evaluation that can be readily implemented in clinical practice.
Graphical Abstract