Artificial Intelligence- and Physician-Interpreted Stool Image Characteristics Correlate With C-Reactive Protein...

Area of Science:

• Gastroenterology
• Medical Artificial Intelligence
• Digital Health

Background:

• Stool characteristics are commonly used to assess ulcerative colitis (UC) disease activity.
• Current methods lack validation against objective inflammation markers.
• Objective validation of stool characteristics for UC monitoring is needed.

Purpose of the Study:

• To determine if stool characteristics measured by AI and physicians correlate with inflammation in UC patients.
• To explore the utility of smartphone-based AI for UC disease activity assessment.

Main Methods:

• Hospitalized acute severe UC (ASUC) patients captured stool images via a smartphone app (Dieta®).
• Validated AI measured stool characteristics (e.g., Bristol stool scale, fragmentation, consistency).
• Physicians scored images for blood, mucus, and stool location; AI and physician data correlated with C-reactive protein (CRP) levels.

Main Results:

• Bristol stool scale and fragmentation positively correlated with CRP.
• Stool consistency negatively correlated with CRP.
• Median CRP was higher in stool images containing mucus.

Conclusions:

• AI-derived stool characteristics (Bristol scale, consistency, fragmentation) significantly correlate with CRP in ASUC patients.
• The presence of mucus in stool is associated with higher inflammation.
• Smartphone-based AI shows potential as a novel, noninvasive tool for UC monitoring.