Related Experiment Video
Updated: May 28, 2026
12:50
Lesion Explorer: A Video-guided, Standardized Protocol for Accurate and Reliable MRI-derived Volumetrics in Alzheimer's Disease and Normal Elderly
Published on: April 14, 2014
Development of a serverless, interactive application for Alzheimer's disease detection and visualization using MRI
Tri Huynh1,2, My Nguyen3, Huong T T Ha1,2
1School of Biomedical Engineering, International University, Ho Chi Minh City, Vietnam.
The Neuroradiology Journal
|May 27, 2026
Summary
This study developed an AI system for early Alzheimer's disease detection using MRI scans. The cost-effective, cloud-based tool offers high accuracy, aiding diagnosis in resource-limited settings.
Area of Science:
- Neuroimaging and Artificial Intelligence
- Computational Neuroscience
- Medical Diagnostics
Background:
- Early Alzheimer's disease (AD) detection is crucial for effective intervention with disease-modifying treatments.
- Clinical implementation faces barriers like limited computational resources and the research-to-practice gap.
- Resource-constrained healthcare facilities require accessible diagnostic tools.
Purpose of the Study:
- To develop and evaluate a computer-aided diagnosis (CADx) system for classifying cognitive states (Alzheimer's disease, Mild Cognitive Impairment, Cognitively Normal) from structural MRI.
- To integrate automated preprocessing and cost-effective cloud deployment for resource-limited settings.
- To assess the system's accuracy, specificity, and operational costs.
Main Methods:
- A multi-view MRI analysis model was optimized using neural architecture search (NAS) with Universal Inverted Bottleneck blocks and Kolmogorov-Arnold Networks.
- Automated MRI preprocessing via FSL was deployed on serverless cloud functions for scalable processing.
- The system was evaluated on the ADNI dataset (1687 individuals) and a web application was developed for patient management and prediction.
Main Results:
- The model achieved 86.7% accuracy and 0.900 AUC for three-class classification (AD, MCI, CN).
- High specificity was observed across all classes (CN: 91.0%, MCI: 91.8%, AD: 97.3%), with 100% specificity for distinguishing AD from CN.
- Operational costs were approximately $0.028 USD per diagnosis, demonstrating cost-effectiveness.
Conclusions:
- The developed system offers a cost-effective solution for early Alzheimer's diagnosis, accessible to resource-constrained environments.
- The integration of NAS and serverless deployment advances automated AD detection for clinical use.
- Future research should focus on prospective clinical validation and incorporating interpretability features.
