NOT-OD-23-070: Empowering Cloud Computing for Non-image-based Diabetic Retinopathy Screening by Designing an EHR-oriented Incremental Learning Framework

点击翻译按钮获取中文摘要

Project Summary/Abstract Despite the high prevalence of diabetic retinopathy (DR), the recommended annual ophthalmic exam for diabetic patients has a very low compliance rate, only around 43%. Many patients do not seek proper medical attention because DR is asymptomatic in the early stage, and thus miss the most effective period to halt DR progression and prevent vision loss. Moreover, ophthalmic equipment for DR exams is predominantly limited to urban areas, restricting access by patients in rural communities with limited incomes. All of these create an urgent need to develop innovative approaches that enable early detection of DR. Our long-term goal is to develop a cost-effective, non-image based, artificial intelligence (AI) tool for primary care physicians to assess patients’ risk for DR using comorbidity data and routine lab results, which are widely available. It will help physicians recommend ophthalmic exams and individual screening frequency for at-risk patients confidently. The aim of our parent NEI project is to improve data quality and prediction accuracy for DR screening by harnessing tensor information. The parent project demonstrated the feasibility of detecting the existence of DR with about 92% accuracy. Our approach is promising to increase the compliance rate of the recommended ophthalmic exams among asymptotic patients, break the barrier to ubiquitous diabetic eye care in rural communities, and save thousands of people from blindness. The deep-learning models developed in the parent project were trained with data from Cerner Real-World Data (CRWD). CRWD is a nation-wide, comprehensive, relational database of real-world, de-identified, HIPAA- compliant patient data with over 100 million patients and 1.5 billion encounters from diverse care settings. Starting from 2023, Cerner moved CRWD to Amazon Web Services (AWS) completely. Cerner updates CRWD every quarter. The updated data brings opportunities to continuously improve the predictive models, but also the high computational burden to re-train the models with huge amount of data on the cloud. The technical objective of this supplemental project is to design an innovative sample recycling-assisted incremental learning (SR-IL) framework to update deep neural network models with newly added EHR data, without requiring completely re-training the models. We will implement the proposed SR-IL framework on AWS to continuously improve our predictive models for non-image-based DR screening. More and more Electronic Health Record (EHR) data are being moved to the cloud today. However, there is no EHR-oriented incremental learning framework to reduce the computational burden caused by updating healthcare analytics models on the cloud. The methodology developed from this study will contribute to increasing the efficiency of healthcare analytics and help continuously improve the performance of the predictive models of other diseases without adding high computational burden to completely re-train the deep- learning models with huge amount of data on the cloud.