Accurate Diagnosis of the Invasion Depth in ESCC by a Deep Neural Network Analysis of NBI Endoscopy Data

Not yet recruiting

• Conditions: Esophageal Squamous Cell Carcinoma

• Registration Number: NCT06252974
• Lead Sponsor: RenJi Hospital

Brief Summary

The goal of this observational study is to accurate diagnose the stage of esophageal squamous cell carcinoma in order to help physicians to decide the appropriate clinical treatment. The main question it aims to answer is:

• To get early accurate diagnosis of the invasion depth of esophageal squamous cell carcinoma by narrow-band imaging endoscopy data.

Participants' clinical informations from routine examinations and treatments will be collected, there will be no harm to participants.

Detailed Description

Not available

Study Timeline

• Study First Submitted: 2024-01-22
• Study First Submitted QC: 2024-02-01
• Study First Posted: 2024-02-12
• Status Verified: 2024-04
• Study Start: 2024-04-25
• Last Update Submitted: 2024-04-25
• Last Update Posted: 2024-04-29
• Primary Completion: 2024-05-31
• Study Completion: 2024-12-31

Recruitment & Eligibility

• Status: NOT_YET_RECRUITING
• Sex: All
• Target Recruitment: 500

Inclusion Criteria

1. Age ≥ 18 years old, regardless of gender;
2. Performing esophageal ESD surgery due to esophageal mucosal lesions;
3. Pathological evaluation of ESD specimens.

Exclusion Criteria

Pathological examination after ESD surgery ruled out esophageal squamous cell carcinoma.

Study & Design

• Study Type: OBSERVATIONAL
• Study Design: Not specified

Primary Outcome Measures

Name	Time	Method
Accurate diagnose the invasion depth of early esophageal squamous cell carcinoma by endoscopy NBI images through deep neural network analysis	2024/12/31	We will compare the predictive performance of InvaDepNet before and after incorporating data generated using GAN and demonstrated that including the generated data in the training dataset effectively improves the accuracy of the predictive model. Additionally, we will train six commonly used CNN models on two datasets with different shooting angles (including NBI without magnifying and NBI with magnifying), and we will propose a ResNet model to analysis the clinical informations combine with NBI images.