Artificial Intelligence Diagnosis of Different Histopathological Growth Patterns of Colorectal Cancer Liver Metastasis

Sun Yat-sen University

Status: Active, not recruiting

Conditions: Liver Metastases of Colorectal Cancer

Study type: Observational

Funder types: Other

Identifiers

NCT07088393

2023ZSLYEC-256

Details and patient eligibility

About

This study selected cases of colorectal cancer liver metastasis patients who underwent liver metastasis tumor resection, retrieved the pathological HE sections of the metastatic lesions, and constructed a predictive model. AI software was applied to delineate different types of regions, achieving full automation of HGP prediction and constructing a predictive model. Statistical analysis was conducted on the classification of histopathological growth patterns (HGP) of liver metastasis and the survival prognosis of patients, and the differences in prognosis among different HGP classification methods were compared. This provides a new method for judging prognosis and treatment for clinical treatment of colorectal cancer liver metastasis patients.

Enrollment: 437 patients

Sex: All

Volunteers: No Healthy Volunteers

Inclusion criteria

• Patients with colorectal cancer liver metastases who underwent resection of liver metastases;
• Confirmed by a pathologist as having liver metastases from colorectal cancer;

Exclusion criteria

• Cases of colorectal cancer liver metastasis that cannot be classified by histopathology.

Trial design

437 participants in 1 patient group

Colorectal cancer liver metastasis cohort

Description:

A total of 437 cases of colorectal cancer liver metastasis patients who underwent liver metastasis tumor resection were selected, with a total of 1205 tumor lesions. Pathological HE sections were retrieved and a predictive model was constructed. Among them, 301 cases were in the training set and 106 cases were in the validation set. After constructing the model, it was used to prospectively interpret 30 lesions. The interpretation result of a senior pathologist with a high professional title was taken as the standard to evaluate the accuracy and interpretation time of the model.