Leveraging sMRI, Self-Attention Mechanisms, and Evolving Spiking Neural Networks for Enhanced Suicide Ideation Detection in Depressed Young Adults
Accurate assessment of suicide ideation (SI) risk in depressed young adults remains a critical challenge, with existing methods exhibiting limited effectiveness. This study proposes a novel machine learning approach leveraging Evolving Spiking Neural Networks (ESNN) to enhance SI risk detection util...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Thesis |
| Language: | English English English |
| Published: |
2024
|
| Subjects: | |
| Online Access: | http://ir.unimas.my/id/eprint/45396/3/DSVA_Corrine%20Francis.pdf http://ir.unimas.my/id/eprint/45396/4/Thesis%20Master_%20Corrine%20Francis%20-%2024%20pages.pdf http://ir.unimas.my/id/eprint/45396/5/Thesis%20Master_%20Corrine%20Francis.ftext.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate assessment of suicide ideation (SI) risk in depressed young adults remains a critical challenge, with existing methods exhibiting limited effectiveness. This study proposes a novel machine learning approach leveraging Evolving Spiking Neural Networks (ESNN) to enhance SI risk detection utilizing structural magnetic resonance imaging (sMRI) data. ESNNs, inspired by the brain's information processing mechanisms, excel at capturing temporal and spatial patterns in data, making them well-suited for modeling the complex dynamics of SI risk factors. Unlike traditional neural networks, ESNNs employ spiking neurons and adaptive learning mechanisms that continuously update internal representations, enhancing their robustness to changing risk factors and individual SI trajectories. However, their application in SI detection has been largely underexplored, creating a gap in leveraging their unique capabilities for this critical task.
To address this gap, the self-AM-ESNN model is introduced, which integrates self-attention mechanisms (self-AM) with ESNN to enable effective feature extraction and learning from sMRI data. By integrating self-AM with ESNN's dynamic learning capabilities, the model can capture complex neuroanatomical patterns associated with SI risk while adapting to individual variations. Evaluated on a dataset of 20 depressed individuals and 60 healthy controls, the self-AM-ESNN model demonstrated exceptional performance in classifying depression, achieving 94% test accuracy, 100% sensitivity, 92% specificity, and an area under the curve of 0.96.
These promising results highlight the potential of ESNN-based approaches to augment clinical decision-making and mental health interventions for SI risk assessment. Furthermore, the study incorporates a user-centric evaluation framework that enables mental health professionals and service users to assess the model's detections and rationale, facilitating informed decision-making processes. By providing interpretable insights into the underlying factors contributing to SI risk, this approach empowers stakeholders to make more informed choices and tailor interventions accordingly. |
|---|