Multi-Scale Temporal-Frequency Attention Network Based on Ocular Imaging for Depression Detection.

Area of Science:

• Neuroscience
• Computer Science
• Psychiatry

Background:

• Depression is a prevalent mental disorder with diverse symptoms impacting mood, cognition, and physiology.
• Ocular imaging reveals distinct eye movement patterns in individuals with depression compared to healthy controls.
• Existing deep learning models, like convolutional neural networks, have limitations in capturing complex spatio-temporal features from eye movement data.

Purpose of the Study:

• To propose a novel deep learning model, MTFNet, for enhanced depression recognition using ocular imaging.
• To address the limitations of existing models in capturing global and local features from sequential eye movement data.
• To improve the accuracy and effectiveness of AI-driven depression detection.

Main Methods:

• Developed the Multi-Scale Temporal-Frequency Attention Network (MTFNet), integrating Multi-Scale time-frequency domain attention with the Video Swin Transformer.
• Introduced the Multi-Scale Temporal-Frequency Attention Module (MTFAM) to focus on salient regions within eye movement images.
• Utilized sequential eye movement image data for training and evaluating the MTFNet model.

Main Results:

• The proposed MTFNet model achieved a high accuracy of 76.8% on a self-collected eye movement image dataset.
• MTFNet demonstrated superior performance compared to most existing depression recognition models.
• The model effectively captured crucial features from sequential eye movement data, enhancing understanding of underlying structures.

Conclusions:

• MTFNet offers a promising and novel approach for depression recognition based on eye movement imaging.
• The integration of multi-scale temporal-frequency attention significantly improves feature extraction from ocular data.
• This research contributes to the development of objective, AI-powered tools for mental health assessment.