Project Lead(s): German Comina Bellido
Issue
Tuberculosis (TB) is an infectious disease that was responsible for 1.5 million deaths in 2014, with 9.6 million new cases reported that year. TB is transmitted though aerosol droplets, with cough being the most important cause of infection.
Cough in people with active pulmonary TB disease arises as a result of the inflammatory response to the mycobacterial pulmonary infection and a reduction in cough is assumed to represent an adequate response to treatment.
A laboratory-free test for assessing recovery from TB is critically needed in regions of the world where laboratory facilities are lacking.
Solution
The project team proposed to develop a smartphone-integrated cough frequency monitoring device that could be used during first-line tuberculosis (TB) treatment and could serve as a remote, laboratory-free non-invasive indicator of treatment efficacy.
The application would run on a smartphone and monitor audio input. When the intensity of the signal is greater than a set limit, it acquires data from the input, records it and sends it to a web server to be analyzed.
Modification of a previously developed algorithm, which recognized cough in an audio recording and signal from an accelerometer, was used. To evaluate the algorithmic performance, a set of data files were randomly selected. A pair of experienced nurses listened to all the recordings and manually annotated the start of each cough event.
In total, 32 files comprising 16 hours of patient recordings were used for evaluation.
Outcome
The team demonstrated that adding accelerometers to a low-cost cough-monitoring system can significantly reduce false-positive rates for automated analysis of noisy real-world recordings. The resulting system is much more robust, due to the wide variety of environmental noises encountered in real-world TB patient monitoring settings.
Improving data quality is a key issue that needs to be addressed through engineering and through providing users with real-time feedback on data quality.
Unfortunately, the research team was not able to test the smartphone app with actual TB patients. However, knowledge gained from the studies has been widely disseminated in journal publications.