Classifying Arteriovenous Fistula Sounds with Deep Learning
Author Information
Author(s): Zhang Jun, Zhang Rongxi, Shu Xinming, Zhang Hongtao
Primary Institution: Zhengzhou University, Zhengzhou, China
Hypothesis
Can deep learning techniques improve the classification of arteriovenous fistula sounds for better monitoring of AVF stenosis?
Conclusion
The CBAM-LSTM model effectively classifies arteriovenous fistula sounds, achieving high accuracy and precision in detecting stenosis.
Supporting Evidence
- The CBAM-LSTM model achieved an Area Under the Receiver Operating Characteristic curve of 99%.
- Precision of the model was 99%, with a recall of 97% and an F1 Score of 98%.
- Comparative analysis showed that the CBAM-LSTM model outperformed other models like VGG and ResNet50.
Takeaway
This study shows that a special computer program can listen to sounds from blood vessels and tell if they are healthy or not, helping doctors take better care of patients.
Methodology
The study used a CBAM-LSTM neural network to analyze sounds from 800 patients, combining Mel-frequency cepstral coefficients and Mel-spectrogram features for classification.
Potential Biases
Potential biases may arise from the subjective classification of sound samples and the limited diversity of the patient sample.
Limitations
The study may have limitations related to the generalizability of the findings due to the specific patient population and the reliance on audio recordings.
Participant Demographics
{"age_range":"12–86","gender_distribution":{"male":426,"female":374},"history_of_avf":591}
Statistical Information
P-Value
0.01
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website