Conventional small-bowel capsule endoscopy reading vs proprietary artificial intelligence auxiliary systems:...

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Small-bowel capsule endoscopy (SBCE) is the standard for diagnosing small bowel pathologies.
• Current SBCE reading is time-consuming and prone to human error.
• Artificial intelligence (AI) systems offer potential improvements in lesion detection and reading efficiency.

Purpose of the Study:

• To evaluate the performance of proprietary AI auxiliary platforms in SBCE.
• To compare AI-assisted SBCE readings with conventional human-only readings.
• To assess AI's impact on diagnostic accuracy and reading times.

Main Methods:

• Systematic literature search to identify comparative studies.
• Extracted performance measures including accuracy, sensitivity, and specificity.
• Assessed methodological transparency using the MINORS tool.

Main Results:

• Six studies were included in the meta-analysis.
• AI-assisted reading demonstrated significantly higher sensitivity and comparable specificity.
• AI integration reduced reading times by a mean of 12-fold.

Conclusions:

• AI-assisted SBCE reading surpasses conventional human review in accuracy and sensitivity.
• AI significantly reduces SBCE reading times, offering efficiency gains.
• AI holds potential to decrease endoscopy workload and aid in training new readers.