Clinical Report: Utilizing Self-Supervised 3D Deep Learning on Preoperative CT

Overview

This study demonstrates that a self-supervised 3D deep learning model can accurately predict high pathologic nodal burden in esophageal squamous cell carcinoma using preoperative contrast-enhanced CT. The model showed robust performance across multiple cohorts, outperforming traditional CT criteria.

Background

Accurate preoperative nodal staging is crucial in esophageal squamous cell carcinoma (ESCC) as it influences treatment decisions and prognosis. Current imaging techniques, particularly contrast-enhanced CT, have limitations in sensitivity for detecting high nodal burden. This study explores the potential of advanced deep learning techniques to enhance preoperative assessment and improve patient outcomes.

Data Highlights

Key Findings

• The self-supervised model achieved an AUC of 0.881 in the internal test cohort.
• In the external cohort, the model demonstrated a sensitivity of 0.581 and specificity of 0.845.
• Calibration improved with self-supervised pretraining, indicated by a Brier score of 0.148.
• Traditional CT criteria showed lower sensitivity (0.339) and specificity (0.784) compared to the self-supervised model.
• The model was validated across internal, temporal, and external cohorts, indicating its robustness.

Clinical Implications

The findings suggest that self-supervised deep learning models can significantly enhance the preoperative identification of high nodal burden in ESCC, potentially guiding treatment decisions. This approach may be particularly beneficial in settings with limited staging resources or ambiguous CT findings.

Conclusion

The study highlights the efficacy of a self-supervised 3D deep learning model in predicting high pathologic nodal burden in ESCC, offering a promising tool for improving preoperative staging and treatment planning.

Related Resources & Content

1. The ASCO Post, External Validation Confirms Ability of AI Model to Stratify Recurrence Risk in Early-Stage Lung Cancer, 2025 -- External Validation Confirms Ability of AI Model to Stratify Recurrence Risk in Early-Stage Lung Cancer
2. European Radiology, Validation of CT Grey-Level Co-Occurrence Matrix Features Across Multiple Centers for Predicting Overall Survival in Primary Oesophageal Adenocarcinoma, 2024 -- Validation of CT Grey-Level Co-Occurrence Matrix Features Across Multiple Centers for Predicting Overall Survival in Primary Oesophageal Adenocarcinoma
3. ESMO Clinical Practice Guideline interim update on the treatment of locally advanced oesophageal and oesophagogastric junction adenocarcinoma and metastatic squamous-cell carcinoma - PMC, 2025 -- ESMO Clinical Practice Guideline interim update
4. Neoadjuvant chemoradiotherapy plus surgery versus surgery alone for oesophageal or junctional cancer (CROSS): long-term results of a randomised controlled trial - PubMed, 2015 -- Neoadjuvant chemoradiotherapy plus surgery versus surgery alone for oesophageal or junctional cancer (CROSS)
5. The ASCO Post — External Validation Confirms Ability of AI Model to Stratify Recurrence Risk in Early-Stage Lung Cancer
6. The ASCO Post — External Validation Confirms Ability of AI Model to Stratify Recurrence Risk in Early-Stage Lung Cancer
7. ESMO Clinical Practice Guideline interim update on the treatment of locally advanced oesophageal and oesophagogastric junction adenocarcinoma and metastatic squamous-cell carcinoma - PMC
8. Neoadjuvant chemoradiotherapy plus surgery versus surgery alone for oesophageal or junctional cancer (CROSS): long-term results of a randomised controlled trial - PubMed