[Bowel Sounds Detection Method and Experiment Based on Multi-feature Combination]

Area of Science:

• Medical Signal Processing
• Gastroenterology

Background:

• Traditional manual auscultation of bowel sounds requires significant expertise and is prone to subjective errors, time constraints, and labor intensity.
• Accurate monitoring of intestinal motor function is crucial for diagnosing gastrointestinal diseases and assessing postoperative recovery.

Purpose of the Study:

• To develop an automated, objective, and efficient algorithm for detecting and analyzing bowel sounds.
• To overcome the limitations of manual auscultation in terms of accuracy, time, and subjectivity.

Main Methods:

• Wavelet transform was employed for preprocessing bowel sound signals, focusing on noise reduction and signal enhancement.
• Three key features of intestinal sounds were extracted and used in a three-stage decision algorithm for joint threshold detection.
• The algorithm was designed to detect bowel sound signals and precisely locate their start and end points for effective extraction.

Main Results:

• The developed algorithm successfully detected bowel sound signals and identified their start and end points, enabling effective extraction of complete signals.
• Validation using a large dataset of clinical bowel sound data and labels demonstrated an accuracy rate of 83.51%.
• A novel evaluation method was proposed and utilized to rigorously verify the performance of the algorithm.

Conclusions:

• The proposed automated algorithm offers a systematic and theoretically sound approach for the diagnosis of intestinal diseases.
• This method provides reliable monitoring of postoperative intestinal function recovery in patients.
• The findings support the integration of advanced signal processing techniques into clinical practice for improved patient care.