This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Structure Matters: Brain Graph Augmentation via Learnable Edge Masking for Data-efficient Psychiatric Diagnosis
Created by
Haebom
Author
Mujie Liu, Chenze Wang, Liping Chen, Nguyen Linh Dan Le, Niharika Tewari, Ting Dang, Jiangang Ma, Feng Xia
Outline
The limited amount of labeled brain network data presents challenges in making accurate and interpretable psychiatric diagnoses. Self-supervised learning (SSL) offers a promising solution, but existing methods often rely on augmentation strategies that can obscure important structural semantics of the brain graph. To address this, we propose SAM-BG, a two-stage framework for learning brain graph representations that preserve structural semantics. In the pretraining stage, edge maskers are trained on a small labeled subset to capture key structural semantics. In the SSL stage, the extracted structural prior information guides a structure-aware augmentation process, enabling the model to learn more meaningful and robust representations. Experiments on two real-world psychiatric datasets demonstrate that SAM-BG outperforms state-of-the-art methods, particularly in settings with limited labeled data, and discovers clinically relevant connectivity patterns that enhance interpretability. Code is available at https://github.com/mjliu99/SAM-BG .
We present SAM-BG, a brain graph representation learning framework that demonstrates superior performance even in environments with limited labeled data.
◦
Preserving structural meaning offers the possibility of more accurate and interpretable psychiatric diagnosis.
◦
Suggesting the possibility of improving diagnostic and treatment strategies through the discovery of clinically relevant connection patterns.
◦
Ensure reproducibility and scalability through open code
•
Limitations:
◦
Further validation of the scale and diversity of the experimental dataset is needed.
◦
Applicability to other types of brain imaging data (e.g., fMRI) needs to be examined.
◦
A more in-depth analysis of the structural semantic preservation mechanism of SAM-BG is needed.
