A Prediction Model of Defecation Based on BP Neural Network and Bowel Sound Signal Features.

Area of Science:

• Biomedical Engineering
• Physiological Monitoring
• Artificial Intelligence in Healthcare

Background:

• Incontinence management for the disabled is challenging due to invasive, costly, and bulky monitoring methods.
• Bowel sound monitoring offers a non-invasive alternative for intestinal disease monitoring and potential clinical support.
• Existing methods lack the precision and accessibility needed for effective incontinence care.

Purpose of the Study:

• To develop a non-invasive system for predicting bowel movements using bowel sound analysis.
• To enhance the quality of life for individuals with fecal incontinence through early prediction.
• To provide a cost-effective and user-friendly monitoring solution.

Main Methods:

• A portable, non-invasive bowel sound collection system was developed.
• Iterative kurtosis and Kalman filtering were employed for noise reduction and signal processing.
• Feature extraction was performed in time, frequency, and time-frequency domains.
• A Backpropagation (BP) neural network was utilized for classification and prediction.

Main Results:

• The proposed method demonstrated convergence to a stable state during experiments.
• A prediction accuracy of 88.71% was achieved on a dataset of 232 records.
• The performance surpassed other classification methods evaluated.

Conclusions:

• The developed bowel sound monitoring system offers high-precision defecation prediction.
• This non-invasive approach can significantly aid patients with fecal incontinence.
• Advance preparation for defecation is enabled, improving patient care and management.