Automated Ultrasound Technology to Diagnose Traumatic Retinal Detachment
Abstract
Background/Rationale: Our proposal aims to improve diagnosis of a vision-threatening eye disease called retinal detachment (RD) in Service Members who sustain eye trauma. RD describes an emergency situation in which a thin layer of tissue (the retina) at the back of the eye pulls away from its normal position. Recent reports suggest that a significant proportion of Soldiers who sustained ocular trauma in the battlefield or during military exercises were reported to have traumatic RD. The longer RD goes untreated, the greater risk of permanent vision loss in the affected eye. Early diagnosis of RD, before the central area of the retina, called macula, is detached, is crucial for successful reattachment and prevention of vision loss. However, a majority of healthcare providers lack expertise to examine the entire retina thoroughly (as ophthalmologists do), and current diagnostic tools at their disposal are inadequate. While physician-performed eye ultrasound has been shown to be an effective tool to rapidly diagnose RD, the experience and expertise to perform and interpret such ultrasound findings are not universally available, especially in resource-limited settings. The skill level of the healthcare provider is a principal limitation in both the use of eye ultrasound to detect ultrasound findings related to RD and arranging for essential treatment in a timely manner. How, therefore, can we deliver the expertise of a specialist (eye ultrasound image interpretation) to any place in the world? One of the potential tools to overcome these limitations is artificial intelligence (AI), which is a cutting-edge type of computer program. AI-enabled computer programs can fill the void between inadequate training and lack of experience and provide inexperienced operators reliable automated tools capable of detecting RD. AI can improve diagnosis by delivering expertise in ultrasound image interpretation. Objective: Our objective is to develop and validate a prototype mobile AI-enabled diagnostic tool that can accurately detect ultrasound signs of RD by the end of the project period. Our rationale is that the proposed deep-learning and reinforcement learning technology tools will provide a relatively simple and time-efficient strategy that can be implemented in most healthcare settings, including austere environments. Our goal is to make the AI tools available for use in the battlefield and clinical environment by non-experts to rapidly detect RD and status of macula to direct appropriate therapy. Aims: To achieve our goals, we need to develop AI software using large eye ultrasound image datasets and test the accuracy of the software in detecting RD and status of macula. Study Design: Our team of AI scientists, physicians, and biostatistician will employ advanced AI technology and do research using thousands of ultrasound images that have been collected in the clinical environment. We will work with expert physician sonographers and test accuracy of the software in detecting RD and assessing the status of macula. Short-Term Outcomes: We expect to develop a mobile AI-enabled ultrasound software platform that can be used by healthcare providers with minimal experience in ultrasound in diverse healthcare settings (e.g., emergency department, office, inpatient, austere environments, battlefield care, rural healthcare). The results of this study will serve as a platform to plan and implement future AI research effectively in resource-limited environments. The technology will be commercially available for use in the clinical environment to streamline workflow and direct appropriate therapy in patients with RD. Our study findings will lay the foundation for research aimed at using AI technology to detect other ocular pathologies related to eye trauma. The proposed research will contribute to development of new clinical protocols incorporating use of ocular ultrasound AI technology to assess patients (active
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Dec 28, 2022
- Source ID
- W81XWH2210803
Entities
People
- Srikar Adhikari
Organizations
- United States Army
- University of Arizona