Automated Disease Detection in Gastroscopy Videos Using Convolutional Neural Networks

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Gastric diseases, including cancer, affect a significant global population.
• Early detection and treatment are crucial for managing conditions like H. pylori infections.
• Current upper gastroscopy relies on manual video analysis, increasing the risk of human error.

Purpose of the Study:

• To develop an automated deep learning system for detecting gastric diseases from gastroscopy videos.
• To improve the sensitivity and specificity of gastric disease screening.
• To reduce diagnostic errors associated with manual video inspection.

Main Methods:

• Collected and utilized anonymous patient case reports and gastroscopy videos for training.
• Developed and evaluated a convolutional neural network (CNN) model.
• Implemented a sliding window technique to enhance video analysis stability.

Main Results:

• The CNN model achieved 84.92% sensitivity, 88.26% specificity, and 85.2% F1-score on the test set.
• Achieved a 97% true positive rate and a 16.2% false positive rate on a separate video test set.
• Demonstrated the potential of AI in improving automated gastric disease detection.

Conclusions:

• Deep learning models can significantly enhance the accuracy of gastric disease detection during upper gastroscopy.
• The developed automated system shows promise in reducing diagnostic errors and improving patient outcomes.
• Further research and implementation of AI can revolutionize gastrointestinal disease screening.