Depth-Sequence Transformer (DST) for Segment-Specific ICA Calcification Mapping on Non-Contrast CT

2025-07-10

AI Summary

Current methods lack segmentation-specific quantification tools for intracranial carotid artery calcification (ICAC) in non-contrast CT, hindering region-specific stroke risk assessment. Method: This work proposes a novel paradigm that reformulates 3D vascular localization as a 1D axial sequence prediction task. It introduces, for the first time, a parallel probabilistic anatomical landmark localization task to circumvent localization ambiguity caused by downsampling in conventional 3D models. A Transformer-based deep sequence model (DST) processes raw-resolution 2D slice sequences end-to-end to output probability distributions for six key anatomical landmarks. Results: Evaluated on a clinical cohort of 100 cases, the method achieves a mean absolute error of only 0.1 slice, with 96% of predictions within ±1 slice; it also attains state-of-the-art classification performance on the Clean-CC-CCII dataset. This approach provides a highly accurate, interpretable tool for region-specific ICAC quantification and stroke risk stratification.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics
• Security and Privacy: Data transparency and provenance
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

While total intracranial carotid artery calcification (ICAC) volume is an established stroke biomarker, growing evidence shows this aggregate metric ignores the critical influence of plaque location, since calcification in different segments carries distinct prognostic and procedural risks. However, a finer-grained, segment-specific quantification has remained technically infeasible. Conventional 3D models are forced to process downsampled volumes or isolated patches, sacrificing the global context required to resolve anatomical ambiguity and render reliable landmark localization. To overcome this, we reformulate the 3D challenge as a extbf{Parallel Probabilistic Landmark Localization} task along the 1D axial dimension. We propose the extbf{Depth-Sequence Transformer (DST)}, a framework that processes full-resolution CT volumes as sequences of 2D slices, learning to predict $N=6$ independent probability distributions that pinpoint key anatomical landmarks. Our DST framework demonstrates exceptional accuracy and robustness. Evaluated on a 100-patient clinical cohort with rigorous 5-fold cross-validation, it achieves a Mean Absolute Error (MAE) of extbf{0.1 slices}, with extbf{96%} of predictions falling within a $pm1$ slice tolerance. Furthermore, to validate its architectural power, the DST backbone establishes the best result on the public Clean-CC-CCII classification benchmark under an end-to-end evaluation protocol. Our work delivers the first practical tool for automated segment-specific ICAC analysis. The proposed framework provides a foundation for further studies on the role of location-specific biomarkers in diagnosis, prognosis, and procedural planning. Our code will be made publicly available.

Problem

• Segment-specific ICAC quantification is technically infeasible
• Conventional 3D models sacrifice global context for localization
• Lack of automated tools for location-specific ICAC analysis

Innovation

• Parallel Probabilistic Landmark Localization for 3D challenge
• Depth-Sequence Transformer processes full-resolution CT volumes
• Predicts independent probability distributions for key landmarks