A fuzzy-based framework for diagnosing esophageal mobility disorder using high-resolution manometry.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• High-resolution manometry (HRM) is crucial for diagnosing esophageal and colonic motility disorders.
• Interpreting HRM data faces challenges due to device dependency and variable reference values.
• Existing diagnostic methods require refinement for complex motility disorder identification.

Purpose of the Study:

• To develop a decision support framework for diagnosing esophageal motility disorders using HRM data.
• To enhance the accuracy and interpretability of HRM-based diagnoses.
• To address the complexities in HRM data interpretation for medical professionals.

Main Methods:

• Spearman correlation was used to model spatio-temporal dependencies in HRM pressure data.
• Convolutional graph neural networks (CGNNs) embedded relation graphs into feature vectors.
• A novel Expert per Class Fuzzy Classifier (EPC-FC) with ensemble structure and negative correlation learning was developed.

Main Results:

• The framework achieved an average accuracy of 78.03% for single swallows and 92.54% at the subject level.
• The EPC-FC demonstrated superior performance compared to SVM and AdaBoost on benchmark datasets.
• The proposed method showed outstanding performance without limitations on HRM data or disorder types.

Conclusions:

• The developed decision support framework effectively aids in diagnosing esophageal motility disorders from HRM data.
• The EPC-FC offers a flexible, interpretable, and highly generalizable approach to classification tasks.
• This AI-driven framework represents a significant advancement in the clinical application of high-resolution manometry.