Related Experiment Video
Updated: Jul 26, 2026
06:18
Qualitative and Comparative Cortical Activity Data Analyses from a Functional Near-Infrared Spectroscopy Experiment Applying Block Design
Published on: December 3, 2020
3.9K
Deep Learning-Based Multilevel Classification of Alzheimer's Disease Using Non-invasive Functional Near-Infrared
Thi Kieu Khanh Ho1, Minhee Kim2, Younghun Jeon3
1Department of Software, Korea National University of Transportation, Chungju, South Korea.
Frontiers in Aging Neuroscience
|May 13, 2022
Summary
Functional near-infrared spectroscopy (fNIRS) combined with deep learning (DL) shows promise for early Alzheimer's disease (AD) diagnosis. DL models accurately classified AD stages using fNIRS data, highlighting potential for improved diagnostic systems.
Area of Science:
- Neuroscience
- Medical Imaging
- Artificial Intelligence
Background:
- Timely diagnosis of Alzheimer's disease (AD) is crucial for intervention and treatment.
- Functional near-infrared spectroscopy (fNIRS) is a promising technique for early AD detection.
- Existing diagnostic methods require validation with advanced analytical tools.
Purpose of the Study:
- To validate the capability of fNIRS coupled with Deep Learning (DL) models for multi-class AD classification.
- To examine hemodynamic responses in the prefrontal cortex across different subject groups and genders.
- To assess the performance of DL architectures on an imbalanced fNIRS dataset.
Main Methods:
- A comprehensive experimental design involving resting, cognitive, memory, and verbal tasks.
- Measurement of hemodynamic responses using fNIRS in the prefrontal cortex.
- Application of various DL architectures, including CNN-LSTM, to analyze fNIRS data.
Main Results:
- Statistical differences in hemodynamic responses were observed during memory and verbal tasks, correlating with AD severity.
- A gender effect on hemoglobin changes was identified.
- DL models significantly improved multi-class classification performance, with CNN-LSTM achieving the highest accuracy (0.909 ± 0.012).
Conclusions:
- DL frameworks effectively handle imbalanced class distributions in fNIRS data for AD diagnosis.
- fNIRS-based approaches show great potential for developing advanced AD diagnostic systems.
- The study validates the synergy of fNIRS and DL for accurate, early-stage Alzheimer's disease detection.
