Artificial Intelligence for Upper Gastrointestinal Endoscopy: A Roadmap from Technology Development to Clinical...

Area of Science:

• Medical Technology
• Artificial Intelligence
• Oncology

Background:

• Stomach cancer is a leading cause of death globally, with projected increases in incidence and mortality.
• Upper GI endoscopy (UGIE) is crucial for early stomach cancer detection, but misdiagnosis can occur due to human and technical factors.
• Artificial intelligence (AI) offers potential solutions to enhance UGIE accuracy and effectiveness.

Purpose of the Study:

• To review current AI algorithms applied to gastroscopy for stomach cancer diagnosis.
• To focus on AI's role in ensuring exam completeness and detecting/characterizing precancerous and neoplastic lesions.
• To discuss future challenges for integrating AI into clinical UGIE practice.

Main Methods:

• Review of state-of-the-art AI algorithms, particularly deep learning architectures for computer vision.
• Analysis of AI applications in recognizing endoscopic patterns from UGIE video data.
• Focus on AI for detecting blind spots, identifying gastric precancerous conditions, and characterizing neoplastic changes.

Main Results:

• AI, using deep learning, has demonstrated early promise in analyzing endoscopic video data for gastroscopy.
• Algorithms show potential in assuring exam completeness and assisting in the detection and characterization of gastric lesions.
• Current results are promising but highlight remaining algorithmic challenges for widespread AI adoption in UGIE.

Conclusions:

• AI holds significant potential to improve the accuracy and completeness of upper GI endoscopy for stomach cancer diagnosis.
• Further development of robust deep learning models and availability of large, annotated datasets are crucial for clinical integration.
• AI-assisted UGIE could significantly enhance early detection and improve patient survival rates for stomach cancer.