Enhancing bowel sound recognition with self-attention and self-supervised pre-training
2024

Improving Bowel Sound Recognition with Deep Learning

Sample size: 19 publication 10 minutes Evidence: high

Author Information

Author(s): Yu Yansuo, Zhang Mingwu, Xie Zhennian, Liu Qiang

Primary Institution: Academy of Artificial Intelligence, Beijing Institute of Petrochemical Technology, Beijing, China

Hypothesis

Can a deep learning-based method enhance the recognition of bowel sounds?

Conclusion

The study demonstrates that the Branchformer model significantly improves bowel sound recognition accuracy compared to existing methods.

Supporting Evidence

  • The Branchformer model outperformed traditional models like CNN and LSTM in recognizing bowel sounds.
  • Self-supervised pre-training improved the model's performance even with limited labeled data.
  • The study validated the effectiveness of the proposed methods through experiments on public bowel sound datasets.

Takeaway

This study shows that a new computer program can help doctors listen to and understand bowel sounds better, making it easier to check if someone is healthy.

Methodology

The study used a deep learning model called Branchformer, which processes audio signals to recognize bowel sounds, trained on a dataset of bowel sound recordings.

Potential Biases

Potential biases may arise from the limited dataset and the specific conditions under which data was collected.

Limitations

The study's dataset may not fully represent the diversity of clinical environments and patient demographics.

Participant Demographics

The study included bowel sound recordings from 19 participants.

Statistical Information

P-Value

p<0.001

Statistical Significance

p<0.001

Digital Object Identifier (DOI)

10.1371/journal.pone.0311503

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication