“TimBre” Phase-II Pilot Study Conducted Using Multi-Country Longitudinal Training Data for Screening of Pulmonary Tuberculosis Using Cough (Acoustic Sounds), Clinical & Demographic Inputs

Authors

  • Rahul Pathri Docturnal Inc, India Author
  • Sasidhar R Sannadi Docturnal Inc, India Author
  • Samarth Tandon Docturnal Inc, India Author
  • Aparna Sykam Docturnal Inc, India Author

DOI:

https://doi.org/10.47363/JPRR/2025(7)188

Keywords:

Cough Sounds, Pulmonary Tuberculosis, Sensitivity, Specificity, XGB, RUS, Ensemble, Spectral Analysis, FFT. STFT, XAI

Abstract

TimBre from docturnal offers screening for multiple lung diseases – Pulmonary Tuberculosis, Pneumonia & Lung Cancer. Detailed studies of TimBre in the past used a third-party Microphone Array with an XY arrangement that recorded high fidelity cough sounds with an average length of >5 seconds and basic demographic data such as Gender, Age, Height, Weight, BMI. In the current study, cough sounds were collected from 7 different countries (India, Vietnam, Philippines, Uganda, Tanzania, Madagascar, and South Africa) using Mobile Phones from different manufacturers & recorded solicited & longitudinal cough sounds for a duration of 0.5 seconds as a part of the phase-I study. We used longitudinally obtained cough sounds numbering 724,694 .WAV files with a resolution of 44.1 kHz & 16 bits. The duration was of 0.5 seconds, with the subject clinical and demographic variables added to the model, which anticipated an improved accuracy over the phase-1 study that used 10,000 plus cough files that obtained a sensitivity & specificity of 75.41% and 68.30% respectively with an AUC of 0.78. The current study (Phase-2) resulted in an overall sensitivity of 68.8%, a specificity of 70.4%, and an AUC of 70% when compared with the microbiological reference standard MRS (sputum culture or Xpert, 25% TB prevalence), and sensitivity of 73.8%, specificity of 69.5%, and an AUC of 71% compared to a sputum Xpert Ultra reference standard (22% prevalence). The phase-1 study used true labels of GeneXpert results as some of the culture results were labelled indeterminate. A comparison between the phases is unwarranted given the fact that the MRS definition is GeneXpert OR Culture while, in Phase-1 it is was strictly GeneXpert. The study and its earlier phase demonstrate the usage of mobile phone-based screening for low resource settings or home-based TB screening or countries that are addressing TB elimination or eradication goals.

Author Biographies

  • Rahul Pathri, Docturnal Inc, India

    Docturnal Inc, India

  • Sasidhar R Sannadi, Docturnal Inc, India

    Docturnal Inc, India

  • Samarth Tandon, Docturnal Inc, India

    Docturnal Inc, India

  • Aparna Sykam , Docturnal Inc, India

    Docturnal Inc, India

Downloads

Published

2025-04-30