Current status of artificial intelligence analysis for endoscopic ultrasonography

Area of Science:

• Medical imaging analysis
• Artificial intelligence in medicine
• Gastroenterology

Background:

• Endoscopic ultrasonography (EUS) is crucial for diagnosing pancreatic diseases but lacks specificity.
• Artificial intelligence (AI) automates pattern recognition in data for improved diagnostic accuracy.
• Deep learning, a subset of AI, utilizes neural networks for complex medical image analysis.

Purpose of the Study:

• To review the principles and applications of AI and deep learning in EUS for pancreatic disease diagnosis.
• To explore the potential of AI in enhancing the specificity of EUS imaging.
• To summarize current research and future directions for AI in EUS-based pancreatic diagnostics.

Main Methods:

• Review of AI and deep learning algorithms and their application in medical image diagnosis.
• Analysis of AI's role in EUS for detecting anatomical features, pancreatic tumors, and cysts.
• Examination of conventional machine learning and deep learning architectures in EUS studies.

Main Results:

• AI and deep learning show high diagnostic abilities (85-95%) in exploratory EUS studies for pancreatic diseases.
• AI aids in detecting anatomical landmarks, differentiating tumors, and identifying cysts in EUS.
• Current evidence is limited, with deep learning applied in only a few reports.

Conclusions:

• AI is a rapidly advancing technique with the potential to become essential for medical diagnosis, particularly in EUS.
• Further research with substantial evidence is needed to fully integrate AI into routine EUS practice for pancreatic diseases.
• AI-driven EUS may significantly improve the diagnostic accuracy and specificity for pancreatic tumors and chronic pancreatitis.