Clinical Report: Investigation of Spatial Relationships Between Esophageal Lesions

Overview

This study developed a Grad-CAM–guided deep learning–radiomics framework to predict esophageal fistula risk in patients with esophageal squamous cell carcinoma (ESCC).

Background

Esophageal fistula is a severe complication of esophageal squamous cell carcinoma treatment, often leading to poor patient outcomes. Current imaging techniques have limited predictive capabilities.

Data Highlights

Key Findings

• The 3D-CNN achieved a mean accuracy of 0.809 in predicting esophageal fistula.
• Grad-CAM highlighted significant activation differences in the lungs and thoracic spine.
• Radiomics analysis showed significant textural differences in the lungs, thoracic spine, and esophageal tumor (P < 0.001).
• Higher Radscores were observed in patients who developed esophageal fistula.
• Conventional CT parameters lost statistical significance in predicting fistula risk.

Clinical Implications

The Grad-CAM–guided framework may provide clinicians with a tool for assessing the risk of esophageal fistula in ESCC patients.

Conclusion

The study presents a novel radiomics framework that could improve risk stratification for esophageal fistula in patients with esophageal squamous cell carcinoma.

Related Resources & Content

1. Frontiers in Oncology, 2026 -- CT-based radiomics combined with deep learning for predicting radiation pneumonitis in patients with esophageal cancer: A two-center study
2. Integrating Transrectal Ultrasound with a Radiomics Approach to Assess Neoadjuvant Chemoradiotherapy Outcomes in Locally Advanced Rectal Cancer
3. Radiomic Analysis of MRI for Preoperative T-Stage Assessment in Rectal Cancer: A Retrospective Study
4. ESMO Clinical Practice Guideline interim update on the treatment of locally advanced oesophageal and oesophagogastric junction adenocarcinoma and metastatic squamous-cell carcinoma - PMC
5. Risk Factors for Esophageal Fistula in Esophageal Cancer Patients Treated with Radiotherapy: A Systematic Review and Meta-Analysis | Oncology Research and Treatment | Karger Publishers
6. European Radiology — Radiomic Analysis of Unenhanced CT Scans for Identifying Endoleaks Following Endovascular Aneurysm Repair of Abdominal Aortic Aneurysms
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. Risk Factors for Esophageal Fistula in Esophageal Cancer Patients Treated with Radiotherapy: A Systematic Review and Meta-Analysis | Oncology Research and Treatment | Karger Publishers
9. BMC Medical Imaging