To improve the quality of upper gastrointestinal endoscopic procedures through a modified photo-documentation guideline that integrates both British and Japanese standards, and a deep learning-based classification system.
Key Findings:
Proposed a modified guideline for upper GI endoscopic photo-documentation that enhances standardization.
Introduced a new annotated dataset crucial for AI system development and validation, addressing a significant gap in existing resources.
Developed a complete workflow for EGD image classification, including data collection and ROI extraction, which can be adapted for future AI applications.
Interpretation:
The study highlights the potential of AI in enhancing endoscopic procedures by providing a structured approach to image documentation and classification, which may lead to improved detection rates of lesions and better patient outcomes.
Limitations:
The dataset is limited to images from a single clinical institution, which may affect the generalizability of the findings.
Variability in image resolution and quality due to different clinical systems may impact classification accuracy.
Conclusion:
The proposed modifications and deep learning approach can significantly improve the detection and classification of lesions during upper GI endoscopy, potentially leading to better patient outcomes.
