Multi-center and Multi-modal Deep Learning Study of Gastric Cancer

Active, not recruiting

• Conditions: Stomach Neoplasms
• Interventions: Radiation: The whole abdomen contrast-enhanced CT scan; Other: H&E stained sections and slides

• Registration Number: NCT05001321
• Lead Sponsor: First Hospital of China Medical University

Brief Summary

To assist postoperative pathological diagnosis and classification of gastric cancer by machine learning; To improve the accuracy of pathological diagnosis of gastric cancer by machine learning; To predict the effectiveness of treatment for gastric cancer by deep learning; To construct a model to predict the survival of gastric cancer patients by multimodal deep learning.

Detailed Description

Not available

Study Timeline

• Study Start: 2021-07-01
• Status Verified: 2021-08
• Study First Submitted: 2021-08-01
• Study First Submitted QC: 2021-08-04
• Last Update Submitted: 2021-08-04
• Study First Posted: 2021-08-11
• Last Update Posted: 2021-08-11
• Primary Completion: 2022-01-31
• Study Completion: 2024-12-31

Recruitment & Eligibility

• Status: ACTIVE_NOT_RECRUITING
• Sex: All
• Target Recruitment: 3300

Inclusion Criteria

• The diagnosis of gastric cancer was confirmed by pathology;
• Preoperative enhanced abdominal CT;
• Available detailed clinical and pathological data;
• Integrated follow-up data.

Exclusion Criteria

• The patients had severe underlying disease;
• Overall survival was less than 3 months;
• No detailed information could be collected.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Arm && Interventions

Group	Intervention	Description
Training Group	The whole abdomen contrast-enhanced CT scan	Based on the inclusion criteria, 2000 gastric cancer patients will be recruited in the analysis. And a model will be constructed based on deep learning.
Internal Validation Group	H&E stained sections and slides	Based on the inclusion criteria, 1000 gastric cancer patients will be recruited in this group to verify the sensitivity and specificity of the constructed model.
External Validation Group	H&E stained sections and slides	Based on the inclusion criteria, 300 gastric cancer patients from 5 other medical centers will be recruited in this group to verify the sensitivity and specificity of the constructed model.
Training Group	H&E stained sections and slides	Based on the inclusion criteria, 2000 gastric cancer patients will be recruited in the analysis. And a model will be constructed based on deep learning.
Internal Validation Group	The whole abdomen contrast-enhanced CT scan	Based on the inclusion criteria, 1000 gastric cancer patients will be recruited in this group to verify the sensitivity and specificity of the constructed model.
External Validation Group	The whole abdomen contrast-enhanced CT scan	Based on the inclusion criteria, 300 gastric cancer patients from 5 other medical centers will be recruited in this group to verify the sensitivity and specificity of the constructed model.

Primary Outcome Measures

Name	Time	Method
Growth pattern	1 day	To assess the growth pattern on preoperative enhanced abdominal CT of patients with gastric cancer, including endophytic, exophytic and mixed.
Nucleus shape	1 day	To obtain the nucleus shape of postoperative H\&E stained sections and slides of gastric cancer by deep learning.
Enhancement pattern	1 day	To assess the enhancement pattern on preoperative enhanced abdominal CT of patients with gastric cancer, including homogeneous and heterogeneous.
Maximum diameter of tumor	1 day	To measure the maximum diameter of tumor on preoperative enhanced abdominal CT of patients with gastric cancer.
Enhancement degree	1 day	To assess the enhancement degree on preoperative enhanced abdominal CT of patients with gastric cancer, including hypoenhancement, isoenhancement and hyperenhancement.
Nucleus size	1 day	To obtain the nucleus size of postoperative H\&E stained sections and slides of gastric cancer by deep learning.
Distribution of pixel intensity	1 day	To obtain the distribution of pixel intensity of postoperative H\&E stained sections and slides of gastric cancer by deep learning.
Texture of nuclei	1 day	To obtain the texture of nuclei of postoperative H\&E stained sections and slides of gastric cancer by deep learning.

Secondary Outcome Measures

Name	Time	Method
Survival status	1 day	To analyze the survival status of patients with gastric cancer, involving dead and alive.
Recurrence/metastasis	1 day	To calculate the days to recurrence/metastasis of patients with gastric cancer.
Overall survival	1 day	To calculate the overall survival of patients with gastric cancer based on days to death and days to last follow-up.

Trial Locations

Locations (6)

The fourth People's Hospital of Changzhou; 🇨🇳 Changzhou, Jiangsu, China

Chaoyang Central Hospital; 🇨🇳 Chaoyang, Liaoning, China

The General Hospital of Fushun Mining Bureau; 🇨🇳 Fushun, Liaoning, China

First Hospital of Jinzhou Medical University; 🇨🇳 Jinzhou, Liaoning, China

The First Affiliated Hospital of China Medical University; 🇨🇳 Shenyang, Liaoning, China

The Second Hospital of Shandong University; 🇨🇳 Ji'nan, Shandong, China