Objective:

To develop and validate a self-supervised 3D deep learning model for predicting high pathologic nodal burden (N2+) in esophageal squamous cell carcinoma using preoperative contrast-enhanced CT, highlighting its clinical significance.

Approach:

Key Findings:

• The self-supervised model achieved AUC values of 0.881 (95% CI: 0.793–0.955) in the internal test cohort and 0.860 (95% CI: 0.810–0.903) in both temporal and external cohorts.
• Sensitivity and specificity in the external cohort were 0.581 and 0.845 for the self-supervised model, compared to 0.339 and 0.784 for the guideline-inspired comparator.
• Calibration improved with self-supervised pretraining, indicated by a Brier score of 0.148.

Interpretation:

The self-supervised model demonstrated robust predictive performance for high pathologic nodal burden in esophageal squamous cell carcinoma, surpassing traditional CT criteria, with significant clinical implications.

Limitations:

• The study is retrospective and may be subject to selection bias, potentially affecting the generalizability of the findings.
• The model's performance may vary with different imaging protocols and scanner types, which should be considered in clinical applications.

Conclusion:

The self-supervised 3D model provides a promising tool for predicting high pathologic nodal burden, potentially aiding in staging and treatment planning.

Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.