An Efficient Transfer Learning With Prompt Learning for Brain Disorders Diagnosis.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Informatics

Background:

• Deep learning models for brain disease diagnosis require extensive training data, which is often limited.
• Cross-disease transfer learning is essential for extracting more information from scarce datasets.

Purpose of the Study:

• To develop a novel prompt learning framework, BPformer, for brain disease diagnosis.
• To leverage cross-disease knowledge transfer using specific prompts within the brain network analysis.

Main Methods:

• Proposed BPformer framework integrating mask, disorder, and adaptive instance-level prompts.
• Utilized prompt learning to model consistent and disorder-specific knowledge and account for inter-individual variations.
• Evaluated on Nanjing Medical University, Autism Brain Imaging Data Exchange, and Alzheimer's Disease Neuroimaging Initiative datasets.

Main Results:

• Demonstrated BPformer's effectiveness in classifying major depressive disorder, bipolar disorder, Alzheimer's disease, and autism spectrum disorder.
• Showcased the model's ability to extract valuable information from limited training data through cross-disease transfer learning.
• Achieved accurate diagnoses across multiple brain network analysis tasks.

Conclusions:

• BPformer offers a powerful approach for brain disease diagnosis with limited data.
• The framework enables disease interpretability and subtype analysis for personalized medicine.
• BPformer facilitates more accurate and fine-grained treatment planning for neurological disorders.