Supervised Machine Learning Segmentation and Quantification of Gastric Pacemaker Cells.

Area of Science:

• Gastroenterology
• Computational Biology
• Microscopy

Background:

• Interstitial Cells of Cajal (ICC) are vital pacemaker cells regulating gastrointestinal motility through slow waves.
• Degradation of ICC networks is linked to gastrointestinal motility disorders.
• Current imaging and segmentation methods limit accurate ICC network analysis.

Purpose of the Study:

• To apply supervised machine learning for extracting ICC networks from 3D confocal microscopy images.
• To accurately quantify ICC network structural properties and identify regional variations.

Main Methods:

• Supervised machine learning (Fast Random Forest, Decision Table, Naïve Bayes) applied to 3D confocal images.
• Trainable WEKA Segmentation used for image processing.
• Quantification of ICC network thickness, density, and process width in mouse gastric antrum.

Main Results:

• Fast Random Forest demonstrated superior performance in ICC network segmentation.
• Significant regional variations in ICC network density and thickness were found in the mouse antrum.
• An inverse relationship between proximal and distal antrum ICC density and thickness was observed.

Conclusions:

• Machine learning provides an accurate method for ICC network structural analysis.
• Regional variations in ICC structure correlate with gut motility.
• Findings offer insights for diagnosing and treating functional gastrointestinal motility disorders.