SynaptiQ: Detection and Prevention of Hospital Induced Delirium
ISEF · 2025 Embedded Systems
Overview
Hospital-induced Delirium (HID) affects 30% of patients causing fluctuating cognitive disturbances, tripling mortality risk and costing North American healthcare systems $164 billion annually. Despite the effectiveness of the Confusion Assessment Method (CAM), 72% of cases go undetected due to sporadic testing and staff limitations. To address this, SynaptiQ was developed, the first AI-powered monitoring system for HID detection and prevention. SynaptiQ integrates predictive modeling on electronic medical records (EMR), non-contact physiological monitoring, facial micro-expression recognition for cognitive assessment, and an AI chatbot to enhance delirium screening. A Non-linear Support Vector Machine trained on MIMIC-III EMR data achieved 93.0% accuracy in predicting delirium, identifying heart and respiratory rates as key predictive factors. Using real-time computer vision with wall or bed-mounted cameras, an algorithm for non-contact measurement of heart and respiratory rate was developed, while prioritizing patient comfort, achieving root mean square errors of 2.26 beats-per-minute and 0.58 breaths-per-minute. For micro-expression recognition to detect emotional cues tied to mental state, a hybrid ResNet18 and dual Long Short-Term Memory network was built, achieving 98.3% accuracy on the SAMM dataset, the highest reported results. The model achieved 93.0% accuracy on fully independent CASME-II testing, showing model robustness. An AI chatbot was built to automate the CAM, interact with patients and provide cognitive reorientation – interventions shown to reduce risk by 50%. By shifting reactive care to proactive, SynaptiQ enables early detection, reduces risk through automated preventive strategies, and transforms delirium management through AI-driven innovation.
Competition history
- ISEF 2025
Resources
Related projects
ISEF · 2023
MemSpark: Artificially Intelligent Virtual Reality System for Non-Intrusive Therapy and Evaluation of Dementia
ISEF · 2024
PoDE - A Platform Leveraging Characteristic Time Series Patterns of Oculomotor Control Attributable to Cholinergic Neuron Destruction in the Parietal Lobe, Indicating Early Alzheimer's Disease Utilizing the PoDE Neural Network
ISEF · 2022
NeuraHealth: An Automated Screening Pipeline To Detect Undiagnosed Cognitive Impairment in Electronic Health Records Using Deep Learning and Natural Language Processing
CWSF · 2026
AD-istics: A Machine Learning Framework for Optimized Detection of Dementia and Alzheimer's Disease
Closest projects by meaning, across every fair and year in the corpus.
Source: Regeneron International Science and Engineering Fair