Non-invasive Detection of Bowel Sounds in Real-life Settings Using Spectrogram Zeros and Autoencoding.

Area of Science:

• Biomedical Engineering
• Gastroenterology
• Signal Processing

Background:

• Gastrointestinal (GI) diseases pose significant health risks due to diagnostic challenges.
• Traditional bowel sound (BS) analysis requires extensive resources and patient immobility.
• There is a need for non-invasive, accessible methods for long-term BS monitoring.

Purpose of the Study:

• To develop and validate a non-invasive system for detecting bowel sounds (BS) in real-life environments.
• To overcome the limitations of traditional auscultation methods for GI disease diagnosis.
• To enable early detection of GI conditions through advanced signal processing and AI.

Main Methods:

• An end-to-end solution using a smart-belt apparatus was designed for BS detection.
• Advanced signal processing techniques were employed to analyze the recorded sounds.
• Deep neural network algorithms were utilized for high-rate BS identification and noise separation.

Main Results:

• The proposed smart-belt system achieved a high rate of bowel sound (BS) identification.
• Effective separation of BS from ambient domestic and urban noise was demonstrated.
• The system proved feasible for real-life BS recordings and analysis.

Conclusions:

• The smart-belt system offers a non-invasive and practical approach for monitoring bowel sounds (BS).
• This technology has the potential to improve early diagnosis and management of GI diseases.
• The study highlights the efficacy of combining wearable technology with AI for remote health monitoring.