AI model to predict risk of liver spread after colorectal cancer surgery

A Multicenter, Prospective, Observational Study for the Validation of a Multimodal Deep Learning Model to Predict Metachronous Liver Metastasis in Patients With Colorectal Cancer After Curative Resection

This project will test whether an AI model using preoperative CT scans, postoperative pathology images, and routine clinical data can predict the risk of liver metastasis within two years for people having curative surgery for stage I–III colorectal cancer.

Quick facts

• Study type: Observational
• Enrollment: 160 (estimated)
• Ages: 18 Years to 75 Years
• Sex: All
• Sponsor: Tongji Hospital (Academic / other)
• Locations: 1 site (Wuhan, Hubei)
• Trial ID: NCT07392567 on ClinicalTrials.gov

What this trial studies

This prospective, observational study at Tongji Hospital enrolls patients with stage I–III colorectal adenocarcinoma undergoing curative resection to validate a pre-developed multimodal deep learning model. The model combines preoperative contrast-enhanced abdominal/pelvic CT scans, digitized postoperative pathology slides, and routine clinical data to predict the risk of liver metastasis within two years after surgery. Participants receive standard-of-care treatment and no experimental interventions; routinely generated images and clinical records are used to generate model predictions which are later compared to actual outcomes. The primary goal is independent prospective validation of predictive accuracy and generalizability to inform whether the model could be integrated into clinical follow-up planning.

Who should consider this trial

Good fit: Adults aged 18–75 with primary colon or rectal adenocarcinoma (stage I–III) scheduled for curative radical resection, with a preoperative contrast-enhanced abdominal/pelvic CT within one month of surgery, ECOG 0–1, and willing to provide informed consent are ideal candidates.

Not a fit: Patients with stage IV disease, synchronous distant metastasis, non-adenocarcinoma pathology, prior liver surgery or transplant, palliative/non-R0 resections, poor-quality imaging, perioperative death, or refusal of consent are unlikely to benefit from this predictive model.

Why it matters

Potential benefit: If successful, the model could help tailor postoperative surveillance and enable earlier detection or treatment of liver metastasis for higher-risk patients.

How similar studies have performed: Retrospective radiomics and AI studies have shown promising predictive performance for metastasis risk, but prospective, independent validation of multimodal deep learning models remains limited.

Eligibility criteria

Inclusion Criteria:

* Age 18-75 years, any gender.
* Clinical diagnosis of primary colon or rectal adenocarcinoma (Stage I-III). Scheduled to undergo curative radical resection for colorectal cancer.
* Preoperative contrast-enhanced abdominal/pelvic CT scan performed within 1 month before surgery, with acceptable image quality.
* No evidence of distant metastasis (including synchronous liver metastasis) on preoperative examination.
* ECOG Performance Status of 0 or 1.
* Patient or their legal representative voluntarily participates and provides written informed consent.

Exclusion Criteria:

* Postoperative pathological confirmation of non-primary colorectal adenocarcinoma or presence of distant metastasis.
* Intraoperative determination of non-R0 resection, or performance of palliative surgery/ostomy only.
* History of other malignant tumors.
* Previous history of liver surgery or liver transplantation.
* Death within the perioperative period (within 30 days after surgery).
* Refusal to participate in follow-up, withdrawal of informed consent, or loss to follow-up.

Where this trial is running

Wuhan, Hubei

• Tongji Hospital — Wuhan, Hubei, China (Recruiting)

Study contacts

• Study coordinator: Yang WU, M.D.

Conditions Colorectal Cancer Liver Metastasis; colorectal cancer liver metastasis; deep learning; multimodal; predictive model