Pneumonia detection

Authors

• Leonid
• [@snevyazh] (https://github.com/snevyazh/)
• Yosef Schoen (https://github.com/YosefSchoen/)

Features

• Take the recorded auscultation and returns pneumonia diagnosis
• Trained on hundreds of records labeled by professional doctors
• Works in real-time
• Works with any smartphone that can record an audio
• Tells of the recording is good (Future Version)

About the project

Article

• Lung auscultation - one of the most basic diagnostic tools used for diagnosing respiratory diseases
• Other basic diagnostics (body temperature, blood pressure, pulse) can be taken by the patient himself
• Lung auscultation requires the participation of a doctor Our instrument allows for self-diagnosis based on the lung auscultation:
• Improve the early diagnosis of pneumonia
• Save unnecessary visits to the doctor - reduce load to Healthcare System, saves patient's time

Files to use

• You need just inference.py, model.pkl and requirements.txt
• The rest of the files are working material on the project

API Reference

Submit audio for prediction

  GET /predict_pneumonia?path={audio_path}

Parameter	Type	Description
path	string	Required. path to audio file

returns

1 in case of positive for pneumonia and 0 in case of negative

References

[https://towardsdatascience.com/learning-from-audio-the-mel-scale-mel-spectrograms-and-mel-frequency-cepstral-coefficients-f5752b6324a8]

[https://surfboard.readthedocs.io/_/downloads/en/latest/pdf/]

[https://librosa.org/]

[https://en.wikipedia.org/wiki/Mel-frequency_cepstrum]