-
1
The proposed framework integrates anatomy-guided parsing, multi-sequence representation learning, and spine-context encoding for lumbar degenerative disease assessment.
-
2
An anatomical parsing module segments vertebral bodies, intervertebral discs, and the spinal canal, providing stable localization for analysis.
-
3
The framework achieves a Macro F1-score of 0.783, a Cohen's Kappa of 0.765, and a patient-level AUC of 0.891, outperforming existing methods.
-
4
Key limitations in existing methods include anatomical ambiguity, underutilization of anatomical priors, and lack of cross-level contextual modeling.
-
5
The design of the framework enhances diagnostic performance and interpretability, aligning with recommendations for trustworthy radiology AI.
