Enhanced Vertebra-GNN: A Graph-Based Deep Learning Framework for Anatomically Informed Spinal Disease Classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Spine Diagnostics

Background:

• Accurate classification of spinal pathologies from radiographic images is crucial for diagnosis and treatment.
• Current methods may not fully capture complex anatomical relationships in spinal X-rays.

Purpose of the Study:

• To develop an end-to-end deep learning framework for automated spinal pathology classification.
• To integrate convolutional neural networks (CNNs) and graph neural networks (GNNs) for enhanced feature extraction and relational modeling.

Main Methods:

• Proposed an Enhanced Vertebra-GNN framework integrating CNNs for regional features and GNNs for inter-vertebral dependencies.
• Explored GCN, GAT, and Graph Transformer variants for relational learning.
• Utilized lateral spine X-rays for model training and evaluation.

Main Results:

• The Enhanced Vertebra-GNN model outperformed conventional CNN baselines in F1-score and accuracy.
• Demonstrated substantial gains in classification performance across all spinal pathology classes.
• Visual interpretability analyses confirmed the model's focus on clinically relevant regions.

Conclusions:

• Enhanced Vertebra-GNN provides a reliable, interpretable, and anatomically aware tool for automated spine assessment.
• The integration of CNNs and GNNs effectively models both appearance and anatomical context.
• This approach holds promise for improving the accuracy and efficiency of spinal pathology diagnosis.