Helios: A Platform for Early Childhood Amblyopia Detection using Fixation Eye Movements
Poster Number: P168
Presentation Time: 05:00 PM - 06:30 PM
Abstract Keywords: Large Language Models (LLMs), Workflow, Deep Learning, Imaging Informatics, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Amblyopia is a neurodevelopmental disorder that is the primary cause of preventable monocular visual impairment among children, with prevalence rates of 3% to 5% globally and 2% to 3% in the United States, that is, 3% to 4% in those aged 1 to 5 years with 677,000 children affected in the US and 99.2 million children globally. Our study offers a novel, machine learning-based framework that analyzes eye movement data with a level of objectivity and precision previously unattainable with standard screening techniques. Utilizing advanced signal processing, we converted raw eye-tracking data from a pediatric cohort into a series of detailed images that reflect the complex dynamics of ocular movement. These images, capturing subtle variations indicative of amblyopia, were interpreted by the fine-tuned Gemini model. A domain expert in ophthalmology was instrumental in crafting the prompts used to train the model, ensuring clinical relevance in the interpretations. The classification results of the Gemini model were promising, showing a notable ability to distinguish between normal and amblyopic subjects. The model was particularly adept at identifying more pronounced cases of amblyopia, highlighting its potential as a scalable and non-invasive diagnostic tool. Our interdisciplinary approach bridges computational innovation with practical medical expertise, demonstrating the utility of AI in enhancing diagnostic methodologies. This research advances medical informatics and underscores the transformative impact of AI in healthcare. It paves the way for further studies to refine the model and explore its integration into clinical practice, with the ultimate goal of improving amblyopia detection and outcomes in pediatric ophthalmology.
Speaker(s):
Dipak Upadhyaya, PhD Student
Case Western Reserve University
Poster Number: P168
Presentation Time: 05:00 PM - 06:30 PM
Abstract Keywords: Large Language Models (LLMs), Workflow, Deep Learning, Imaging Informatics, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Amblyopia is a neurodevelopmental disorder that is the primary cause of preventable monocular visual impairment among children, with prevalence rates of 3% to 5% globally and 2% to 3% in the United States, that is, 3% to 4% in those aged 1 to 5 years with 677,000 children affected in the US and 99.2 million children globally. Our study offers a novel, machine learning-based framework that analyzes eye movement data with a level of objectivity and precision previously unattainable with standard screening techniques. Utilizing advanced signal processing, we converted raw eye-tracking data from a pediatric cohort into a series of detailed images that reflect the complex dynamics of ocular movement. These images, capturing subtle variations indicative of amblyopia, were interpreted by the fine-tuned Gemini model. A domain expert in ophthalmology was instrumental in crafting the prompts used to train the model, ensuring clinical relevance in the interpretations. The classification results of the Gemini model were promising, showing a notable ability to distinguish between normal and amblyopic subjects. The model was particularly adept at identifying more pronounced cases of amblyopia, highlighting its potential as a scalable and non-invasive diagnostic tool. Our interdisciplinary approach bridges computational innovation with practical medical expertise, demonstrating the utility of AI in enhancing diagnostic methodologies. This research advances medical informatics and underscores the transformative impact of AI in healthcare. It paves the way for further studies to refine the model and explore its integration into clinical practice, with the ultimate goal of improving amblyopia detection and outcomes in pediatric ophthalmology.
Speaker(s):
Dipak Upadhyaya, PhD Student
Case Western Reserve University
Helios: A Platform for Early Childhood Amblyopia Detection using Fixation Eye Movements
Category
Poster Invite
Description
Date: Tuesday (11/12)
Time: 05:00 PM to 06:30 PM
Room: Grand Ballroom (Posters)
Time: 05:00 PM to 06:30 PM
Room: Grand Ballroom (Posters)