Using AI on CT scans to predict complete response in esophageal squamous cell carcinoma after neoadjuvant immunochemotherapy

Deep Learning Model Predicts Pathological Complete Response of Esophageal Squamous Cell Carcinoma Following Neoadjuvant Immunochemotherapy

This project will test whether a deep-learning model using chest CT images and clinical data can predict which esophageal squamous cell carcinoma patients will have a complete pathological response after neoadjuvant immunochemotherapy.

Quick facts

• Study type: Observational
• Enrollment: 300 (estimated)
• Ages: 18 Years and up
• Sex: All
• Sponsor: Tongji Hospital (Academic / other)
• Drugs / interventions: chemotherapy, immunotherapy
• Locations: 1 site (Wuhan, Hubei)
• Trial ID: NCT07088354 on ClinicalTrials.gov

What this trial studies

This multicenter retrospective analysis will collect contrast-enhanced chest CT scans and clinical data from ESCC patients who had surgery after neoadjuvant chemotherapy combined with immunotherapy between January 2019 and July 2025. High-throughput deep learning features will be extracted from pre- and post-treatment CT images to build a model that predicts pathological complete response (pCR). Model performance will be quantified by AUC, accuracy, sensitivity, specificity, PPV, and NPV, and SHAP analysis will be used to explain which imaging features drive predictions. The goal is to create a reproducible imaging-based tool to help identify responders to neoadjuvant immunochemotherapy and support personalized care decisions.

Who should consider this trial

Good fit: Ideal candidates are adults with pathologically confirmed esophageal squamous cell carcinoma who received at least one cycle of neoadjuvant chemo-immunotherapy, completed contrast-enhanced chest CT scans before and after treatment, and proceeded to surgery.

Not a fit: Patients with other concurrent malignancies, who received other anti-tumor therapies before or during neoadjuvant immunochemotherapy, with incomplete clinical data, or with poor-quality CT images are unlikely to benefit from this model.

Why it matters

Potential benefit: If successful, the model could help identify patients likely to achieve a complete pathological response and inform more personalized treatment or surgical planning.

How similar studies have performed: Similar radiomics and deep-learning approaches for predicting pathological response in ESCC and other cancers have shown promising but variable results in retrospective cohorts, while application specifically to neoadjuvant immunochemotherapy remains relatively novel.

Eligibility criteria

Inclusion Criteria:

1. Pathologically confirmed esophageal squamous cell carcinoma (ESCC).
2. Received at least one cycle of neoadjuvant chemotherapy combined with immunotherapy.
3. Underwent contrast-enhanced chest CT before initiation of neoadjuvant treatment.
4. Underwent contrast-enhanced chest CT after completion of neoadjuvant treatment and prior to surgery.

Exclusion Criteria:

1. Diagnosis of other malignancies.
2. Received other anti-tumor therapies before or during neoadjuvant chemo-immunotherapy.
3. Incomplete clinical data.
4. Poor-quality CT imaging.

Where this trial is running

Wuhan, Hubei

• Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology — Wuhan, Hubei, China (Recruiting)

Study contacts

• Principal investigator: Yangkai Li, MD, PhD — Tongji Hospital
• Study coordinator: Yangkai Li, MD, PhD

Conditions Esophageal Squamous Cell Carcinoma; Neoadjuvant Immunochemotherapy; Pathological Complete Response; Deep Learning