Artificial intelligence-assisted staging in Barrett's carcinoma.

Area of Science:

• Gastroenterology
• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• The T stage of Barrett's carcinoma is critical for treatment decisions, but current preoperative staging methods like endoscopic ultrasound have limitations.
• Novel tools are needed to improve the accuracy of staging Barrett's carcinoma.
• Artificial intelligence (AI) shows promise in interpreting endoscopic images for neoplasia detection.

Purpose of the Study:

• To investigate the accuracy of an AI system in determining the T stage of Barrett's carcinoma based on endoscopic images.
• To evaluate the AI system's performance in differentiating between various stages of esophageal cancer.

Main Methods:

• A convolutional neural network was trained and validated using 1020 endoscopic images from 577 patients with Barrett's adenocarcinoma.
• The dataset comprised 821 training images and 199 validation images.
• The AI model's performance was assessed based on sensitivity, specificity, and accuracy for different cancer stages.

Main Results:

• The AI system achieved 85% accuracy in recognizing Barrett's mucosa without neoplasia.
• For mucosal cancer, the AI demonstrated a sensitivity of 72% and accuracy of 68%.
• The AI system achieved 67% accuracy for early neoplasia (

Conclusions:

• The AI system demonstrated high accuracy in identifying esophageal cancer stages from endoscopic images.
• The findings suggest that AI can serve as a valuable tool to assist endoscopists in clinical decision-making for Barrett's carcinoma.
• Further research and validation are warranted to integrate AI into routine clinical practice for improved patient outcomes.