Functional Near-Infrared Spectroscopy-Based Computer-Aided Diagnosis of Major Depressive Disorder Using Convolutional...

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Functional near-infrared spectroscopy (fNIRS) offers a portable, cost-effective method for major depressive disorder (MDD) screening.
• fNIRS's low susceptibility to motion artifacts makes it suitable for clinical applications.
• Deep learning methods for fNIRS-based computer-aided diagnosis (CAD) of MDD remain underexplored.

Purpose of the Study:

• To develop and evaluate a novel deep learning framework for fNIRS-based CAD of MDD.
• To achieve high accuracy in differentiating MDD patients from healthy controls using fNIRS data.
• To investigate the effectiveness of a convolutional neural network (CNN) architecture for this diagnostic task.

Main Methods:

• Utilized fNIRS data from 48 MDD patients and 68 healthy controls (HCs) during a Stroop task.
• Employed an ensemble CNN architecture with 1D depth-wise convolutional layers to capture interhemispheric asymmetry.
• Applied a leave-one-subject-out cross-validation strategy for performance evaluation.

Main Results:

• The proposed CNN model achieved an accuracy of 84.48%, sensitivity of 83.33%, and specificity of 85.29%.
• Outperformed conventional machine learning algorithms, including shrinkage linear discriminator analysis (73.28%) and regularized support vector machine (74.14%).
• Demonstrated superior performance compared to other CNN models like EEGNet (62.93%) and ShallowConvNet (62.07%).

Conclusions:

• The novel deep learning model significantly enhances the accuracy of fNIRS-based CAD for MDD.
• The findings support the potential of this approach for developing advanced diagnostic systems for MDD.
• The developed CNN framework effectively utilizes fNIRS data to distinguish between MDD patients and HCs.