CMPT 340 Final Project

Topic: Diagnosing lung diseases using respiratory sounds

Project Summary

Using machine learning (ML) to automatically classify respiratory sounds (RS)is not a new application. However, the bottleneck to developments in this field results from the lack of publicly-available databases containing RS, both healthyand unhealthy, and expert classification of these sounds. To address this issue,the International Conference on Biomedical and Health Informatics (ICBHI) has publicly published a large dataset in 2017 containing 920 respiratory soundrecorded from 126 subjects as part of a Kaggle challenge.

We will be using this dataset by applying ML classifiers to features extracted from this dataset. The classification of RS using ML can lead to a quick and improved diagnosis of respiratory conditions, sometimes providing a better and more consistentclassification than a manual diagnosis by experts. It can also lead to discoveriesof new and improved methods to diagnose conditions.

Enivironment used

• Python - version 3.0 or higher
• GoogleColab (Optional)

Required installations

pip install -r requirements.txt

Data

• All files used and produced are in All_Features folder

Getting Audio Files

• In order to get the audio files you must download and unzip them Link: https://www.kaggle.com/vbookshelf/respiratory-sound-database/download

To run instantly

If you want to run the code instantly just

cd All_Features_And_Models

Then run all_features.py on the audio_and_txt_files_run_instantly

Unzipping

Make sure you are in the correct path when running the command.

cd All_Features_And_Models

Add the downloaded archive.zip into All_Features_And_Models. Using "unzip_files.py" function on the download will put all the folders in the correct place

And all you must do on line 39 of all_features.py located in All_Features is uncomment that line. It should however just be named "audio_and_txt_files .Then make sure you run your code inside /All_Features_And_Models.

Plots

• This file contains all the plots and visualizations

Order of Execution

• Optional: Run audio_files/unzip_files.py

  • extracts audio files from zip to audio_files

• Run All_Features/all_features.py

  • Extracts the Requred features from the kaggle dataset

• Run All_Features/Models.py

  • This file performs various supervised and unsupervised machine learning algorithms on the features.

Aside

Since all of us worked on GoogleColab for this Project, A comment with the name of the developer is assigned to their respective contribution in 'all_features.py' and 'Models.py'. Since we already set up a discord channel (which was our main source of communication), we didn't use Slack. Please refer to the 'Contributions' part of the report for individual contributions.

Developers

• Pranav Sood (301335687)
• Anuj Rattam (301339825)
• Jyotiraditya Mayor (301401591)
• Nathaniel Chan (301314801)
• Vaibhav Saini (301386847)

G-Drive Link - This is where we completed the project.

https://drive.google.com/drive/folders/1e1OTACsf9h9Mi5uoqIsDyb3h904F6VDk?usp=sharing

Discord Discussion

All of the discussions are on discord and will be provided in the supplementary materials.