Foundational Models for Pathology and Endoscopy Images: Application for Gastric Inflammation

Area of Science:

• Medical Diagnostics
• Artificial Intelligence
• Gastroenterology

Background:

• Upper gastrointestinal (GI) cancer, particularly gastric cancer (GC), is a leading cause of global cancer mortality.
• Chronic inflammation drives mucosal changes, progressing from atrophy and intestinal metaplasia (IM) to dysplasia and cancer.
• Early detection via endoscopic surveillance is critical for improving patient survival rates.

Purpose of the Study:

• To review recent advancements, applications, and challenges of foundation models (FMs) in endoscopy and pathology imaging for upper GI cancer.
• To elucidate the principles, architectures, and training methodologies of FMs in medical diagnostics.
• To provide a roadmap for integrating FMs into clinical practice for GC prevention and management.

Main Methods:

• Exploration of core principles and architectures of foundation models (FMs).
• Analysis of training methodologies and the role of large-scale data in FM development.
• Review of current applications in endoscopy and pathology image analysis.

Main Results:

• Foundation models show promise in enhancing the accuracy of endoscopic and pathology image analysis for upper GI cancer.
• FMs leverage large-scale, diverse data for robust predictive capabilities.
• The review highlights the potential of FMs to improve early detection and management strategies.

Conclusions:

• Foundation models represent a significant advancement in AI-driven medical diagnostics for upper GI cancers.
• Integrating FMs into clinical workflows can improve the accuracy and efficiency of cancer detection.
• Future research should focus on multimodal data integration and developing equitable, real-time diagnostic tools.