Project

CODA-TB DREAM

We participated in a DREAM challenge to use audio recordings of cough to diagnose the presence of Tuberculosos in a patient.

CODA-TB DREAM image

Authors

Gautam Ahuja [1], Aakash Rao [1], Diya Khurdiya [1], Shalini Balodi [1], Ashwin Salampuriya [1], Rintu Kutum [1,2,*]

Affiliations

  1. Department of Computer Science, Ashoka University, Haryana, India
  2. Trivedi School of Biosciences, Ashoka University, Haryana, India

*. Corresponding Author

Description

MFCC based neural network classifier to predict TB status from cough sound. We achieved best class-wise accuracy of 67.20% for positive with mel-frequency cepstral coefficients (MFCC) as features and artificial neural network (ANN) model.

Methods

MFCC: Features were extracted with the librosa mfcc function, then linearised to form model-ready inputs for ANN and CNN1D models.

Scaled log2 mel-frequency spectrogram: Spectrogram points were extracted, transformed with log2, and scaled for learning.

Training and testing: Data was split 80:20 (train:test), and 5-fold validation was used during training.

ANN model: Three dense layers (100, 200 with dropout, 100) using ReLU, with softmax output and categorical cross-entropy.

CNN 1D model: Three Conv1D layers (32, 64, 128 filters), flatten and dense layers, with ReLU and softmax output.

Results

CODA TB challenge results

The results of our submission to the CODA TB Dream Challenge 2023.

Code repository

https://github.com/rintukutum/dream-coda-tb-sc1-chsxashoka

My role

I adviced on the project in terms of the design of the Artificial Neural Network as well as the benchmarking of other algorithms on the dataset.

Acknowledgements

We would like to acknowledge Mphasis F1 foundation, Centre for Computation Biology & Bioinformatics (BIC) computing facility, and High Performance Computing (HPC) at Ashoka University for support.

License

Please refer to the CODA-TB licensing guidelines.