X-SiT: Inherently Interpretable Surface Vision Transformers for Dementia Diagnosis

2025-06-25

AI Summary

To address the poor interpretability and clinical intractability of 3D voxel-based brain imaging data, this paper introduces the first intrinsically interpretable cortical surface vision transformer. Methodologically, we propose a case-reasoning–driven prototypical surface patch decoder that enables visual attribution of classification decisions via spatially aligned cortical prototypes; further, we integrate surface patch embedding, geometry-aware positional encoding, and prototype learning to jointly preserve cortical geometry and enhance discriminative capacity. Evaluated on multi-center Alzheimer’s disease and frontotemporal dementia diagnosis tasks, our model achieves state-of-the-art performance. The learned disease-specific cortical prototypes exhibit strong concordance with established neuropathological priors, and the model reliably localizes regions prone to misclassification. These capabilities provide clinicians with intuitive, neurobiologically grounded, and trustworthy interpretability support for diagnostic decision-making.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Interpretable models are crucial for supporting clinical decision-making, driving advances in their development and application for medical images. However, the nature of 3D volumetric data makes it inherently challenging to visualize and interpret intricate and complex structures like the cerebral cortex. Cortical surface renderings, on the other hand, provide a more accessible and understandable 3D representation of brain anatomy, facilitating visualization and interactive exploration. Motivated by this advantage and the widespread use of surface data for studying neurological disorders, we present the eXplainable Surface Vision Transformer (X-SiT). This is the first inherently interpretable neural network that offers human-understandable predictions based on interpretable cortical features. As part of X-SiT, we introduce a prototypical surface patch decoder for classifying surface patch embeddings, incorporating case-based reasoning with spatially corresponding cortical prototypes. The results demonstrate state-of-the-art performance in detecting Alzheimer's disease and frontotemporal dementia while additionally providing informative prototypes that align with known disease patterns and reveal classification errors.

Problem

• Develop interpretable model for dementia diagnosis
• Improve visualization of 3D brain structures
• Classify surface patches for disease detection

Innovation

• Inherently interpretable Surface Vision Transformer
• Prototypical surface patch decoder
• Human-understandable cortical feature predictions