AI-enabled Stroke Prediction in Patients with Chronic Kidney Disease

PROJECT SUMMARY Chronic Kidney Disease (CKD) affects 14% of the U.S. population and is associated with a high risk of both ischemic and hemorrhagic strokes and a mortality rate of up to three times that of the general population. Effective stroke risk prediction is flawed in CKD patients, because 1) comorbidities often remain undiagnosed, 2) stroke risk stratification schemes do not consider stages of CKD, and 3) the detection of stroke risk variations due to dialysis requires real-time risk monitoring, which is unavailable to date. The real-time monitoring of the stroke risk in CKD patients, and in particular for those undergoing dialysis, is likely to influence adopted therapeutic strategies and promote a more personalized therapeutic approach. Better care and prevention through monitoring of these high-risk patients will reduce mortality rates and their high per capita cost, which is five times higher than the average healthcare spending. Biotricity is developing Bioflux-AI, an innovative system for real-time monitoring and prediction of stroke episodes in CKD patients. Bioflux-AI combines an FDA-approved, high-precision, small mobile cardiac telemetry (MCT) device with AI-driven algorithms specifically trained for the prediction of stroke in stage 4 and 5 CKD patients. Biotricity has previously generated and validated algorithms for the automated detection of ECG abnormalities, including Atrial Fibrillation (AF). Given the strong association of AF with increased risk of blood clot formation and hence, ischemic stroke in CKD patients, Biotricity proposes to combine the detection of this arrhythmia with other stroke risk factors of CKD patients (age, weight, height, BMI, CKD status, diabetes, heart disease) and ECG parameters to predict stroke risk in real-time. To this aim, in this SBIR Phase I project a convolutional neural network algorithm, which will incorporate all these risk factors, will be developed, trained and validated. The accomplishment of this feasibility study will pave the road for further development and optimization of the AI-based algorithm for stroke prediction in CKD patients, while widening the application to a larger patient demographic, validating the predictive algorithm for patients with other chronic diseases.