Suriya Dev Saravanakumar
Built an ML-powered eye-tracking framework for early Alzheimer's detection as a 9th grader with zero research experience. Accepted at IEEE EMBC 2026 in Toronto - the world's top biomedical engineering conference.
Where Suriya Started
His Background
- • 9th grader at Doha College, Qatar
- • Interested in AI and machine learning
- • Comfortable with coding but new to research
- • No prior research, mentorship, or science fair experience
His Goals
- • Win an ISEF-level science fair
- • Publish research in a top journal
- • Use AI to detect Alzheimer's from eye movements
- • Eventually start a company based on his research
The Problem He Wanted to Solve
Alzheimer's disease gradually damages brain cells, but detecting it early - before clear symptoms emerge - remains one of medicine's biggest challenges. Eye movements reflect attention, memory, and cognitive function, and Alzheimer's disrupts all of these. Suriya wanted to build an AI system that could detect Alzheimer's through eye-tracking patterns - a non-invasive, scalable approach to early diagnosis.
The Research
Working with YRI mentors, Suriya developed a machine learning framework that analyzes simulated eye-tracking data - fixation duration, saccade amplitude, and blink rate - to distinguish Alzheimer's patients from healthy controls. He tested multiple ML models including Gradient Boosting, Random Forest, and Neural Networks, using rigorous ten-fold cross-validation to evaluate reliability under signal degradation conditions.
Calibration Reliability of Eye-Tracking-Based Alzheimer's Disease Models Under Signal Degradation
Early Alzheimer's detection is difficult before clear symptoms emerge, and large standardized clinical datasets are scarce
Literature-informed synthetic eye-tracking dataset with ten-fold cross-validation across multiple ML classifiers
Random Forest, XGBoost (Gradient Boosting), Feed-Forward Neural Network, and Ensemble methods
ROC-AUC of 0.75, with horizontal eye-position measures identified as strongest diagnostic indicators
From Eye Movements to Early Diagnosis
Suriya's key insight: individual eye-tracking features like fixation duration, saccade amplitude, and blink rate showed weak correlations with each other - meaning each captures a different aspect of cognitive decline. Combined through ensemble ML, they provide complementary diagnostic signals that no single measure could achieve alone. He also built a Streamlit prototype demonstrating real-time eye-tracking data visualization, showing a path from research to practical clinical tools.
ML Models
ROC-AUC
Cross-Validation
Features Analyzed
The Outcome
Accepted at the 48th Annual International Conference of IEEE Engineering in Medicine and Biology Society
IEEE EMBC 2026, Toronto, Canada
July 26-30, 2026
1706
Oral or poster presentation
9th grader with no research experience, no mentor, and no science fair background
Accepted at IEEE EMBC - the world's top biomedical engineering conference - as a 9th grader from Qatar
The Bigger Picture
People worldwide living with dementia - early detection could transform treatment outcomes for millions
The world's largest and most prestigious biomedical engineering conference - accepted as a 9th grader
From zero research experience to IEEE-published in under a year through YRI mentorship
A Father's Perspective
Hear from Suriya's father about the impact of the YRI Fellowship
Ready to Start Your Research Journey?
Join the YRI Fellowship and work with expert mentors to conduct original research, get published, and build a profile that stands out.
Apply Now