Artificial intelligence and capsule endoscopy: automatic detection of enteric protruding lesions using a...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Capsule endoscopy (CE) is vital for small intestine examination but reviewing images is laborious and error-prone.
• Protruding lesions in the small bowel present diverse morphologies in CE images, posing diagnostic challenges.

Purpose of the Study:

• To develop a convolutional neural network (CNN) algorithm for automated detection of protruding lesions in capsule endoscopy images.
• To address the time-consuming and error-prone nature of manual CE image review.

Main Methods:

• A CNN model was trained and validated on a dataset of 18,625 CE images from 1,229 patients.
• The dataset included images with protruding lesions and normal mucosa, with an 80%/20% train/validation split.
• Model performance was assessed using sensitivity, specificity, and accuracy.

Main Results:

• The CNN achieved 92.5% accuracy in detecting small bowel protruding lesions.
• The model demonstrated high sensitivity (96.8%) and specificity (96.5%).
• The CNN processed the validation dataset rapidly, at approximately 70 frames per second.

Conclusions:

• An accurate CNN algorithm was developed for detecting enteric protruding lesions with varied morphologies.
• This AI-driven tool has the potential to significantly enhance diagnostic efficiency in capsule endoscopy.