CrossSpine: Multi-scale Cross-sequence Attention with Anatomical Priors for Automated Pfirrmann Grading

2026-07-22

AI Summary

This study addresses the limited performance of existing models in automated Pfirrmann grading of lumbar intervertebral disc degeneration by proposing CrossSpine, a novel framework that adaptively fuses multimodal MRI features through a multi-scale cross-sequence attention mechanism. CrossSpine further incorporates anatomical priors of intervertebral discs to guide hierarchical-aware classification, effectively modeling structural prior knowledge and significantly enhancing the objective quantification of degeneration associated with low back pain. Evaluated on a high-quality dataset with expert annotations, CrossSpine achieves substantial improvements over baseline models, yielding a relative increase of over 125% in Macro F1 score, 99% in Mean AUPRC, and 36% in Mean AUROC.

Abstract

Automated grading of Lumbar Disc Degeneration is essential for the objective quantification of structural changes associated with low back pain. Observing that baseline models underperformed on our data, we propose a framework designed to overcome these limitations. First, we present the Cross-sequence Attention Spine (CrossSpine) framework, a novel architecture that employs a cross-sequence attention mechanism to adaptively fuse features from different MRI sequences at multiple spa- tial scales. Second, we contribute a meticulously curated dataset aimed at automated Pfirrmann grading. Finally, we introduce an IVD-aware classification technique that integrates anatomical disc-level information, enabling the model to learn level-specific degeneration priors. Our experi- ments demonstrate the superiority of this approach: CrossSpine achieved a relative improvement exceeding 125% in the Macro F1 score, while boosting the Mean AUPRC by 99% and the Mean AUROC by 36% com- pared to the baseline.

Problem

• Lumbar Disc Degeneration
• Pfirrmann Grading
• Automated Grading
• MRI
• Anatomical Priors

Innovation

• Cross-sequence Attention
• Multi-scale Fusion
• Anatomical Priors
• IVD-aware Classification
• Pfirrmann Grading