Application of artificial intelligence in gastroenterology

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Deep learning (DL), a subset of artificial intelligence (AI), excels in analyzing large datasets, particularly medical images.
• Gastroenterology generates vast amounts of clinical data and images (endoscopy, ultrasound), necessitating efficient analytical tools.
• Convolutional neural networks (CNNs) are a prominent DL model for image analysis.

Purpose of the Study:

• To review the application of AI and DL in gastroenterology for diagnosis, prognosis, and image analysis.
• To highlight the challenges and requirements for robust validation of AI models in this field.
• To emphasize the need for further research into AI interpretability in clinical settings.

Main Methods:

• Review of AI and DL applications in gastroenterology literature.
• Discussion of potential biases in retrospective studies, including selection bias, overfitting, and spectrum bias.
• Emphasis on the necessity of external validation and prospective studies for reliable AI model verification.

Main Results:

• AI and DL, particularly CNNs, have demonstrated significant potential in gastroenterological applications.
• Retrospective studies are prone to biases (selection, overfitting, spectrum bias) that can inflate accuracy.
• External validation and prospective studies are crucial for confirming AI model efficacy and generalizability.

Conclusions:

• AI and DL offer powerful tools for gastroenterology, but rigorous validation is essential.
• Minimizing spectrum bias and ensuring representative datasets are critical for developing reliable AI models.
• Addressing the interpretability of DL models is vital for safety, bias detection, and clinical acceptance.