A new artificial intelligence system for both stomach and small-bowel capsule endoscopy

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Artificial intelligence (AI) offers benefits for small-bowel (SB) capsule endoscopy (CE) image analysis.
• Limited data exists on AI's application in gastric and SB CE.
• This study addresses the need for advanced AI in gastrointestinal (GI) CE.

Purpose of the Study:

• To evaluate the efficacy of an upgraded AI algorithm, SS Plus, in detecting GI lesions.
• To assess the impact of SS Plus on the diagnostic performance of endoscopists, particularly juniors.
• To determine if SS Plus reduces the workload in CE image interpretation.

Main Methods:

• A deep learning-based AI algorithm, SmartScan (SS), was enhanced with gastric imaging data to create SS Plus.
• The study involved training on 1069 GI CE examinations (2,672,542 images) and validation on 342 CE examinations.
• Performance was compared between SS Plus-assisted reading (SSP-AR) and conventional reading (CR) by senior and junior endoscopists.

Main Results:

• SS Plus is designed to detect 5 gastric and 17 SB lesion types.
• SSP-AR significantly reduced the number of CE images requiring review (median 1000 vs 42,393) and reading time (median 8.51 minutes).
• SS Plus identified more gastric (411 vs 400) and SB lesions (422 vs 368) compared to CR, with notable improvement for junior endoscopists.

Conclusions:

• The upgraded AI algorithm, SS Plus, effectively detects GI lesions in CE videos.
• SS Plus enhances diagnostic accuracy and efficiency in CE interpretation.
• The AI tool particularly improves the performance of junior endoscopists in CE image reading.