How to automatically detect a cough in a portable spirometer?
Female patient during spirometry
The innovative solution has been described by Mateusz Soliński and Michał Łepek – PhD students from the Faculty of Physics at the Warsaw University of Technology – and Łukasz Kołtowski, MD, PhD from the Medical University of Warsaw. You can read about the results of their work in the “Informatics in Medicine Unlocked” journal.
Spirometry is the most important and the most common of tests allowing to diagnose and monitor lung function. It consists in breathing through a mouthpiece of a device. This is how the volume of air exhaled from and inhaled into the lungs is assessed.
It is important that the test result is not disturbed by coughing (especially in the first seconds). And this is a very common ailment of patients struggling with lung diseases, which is why it is so difficult to eliminate.
Another problem is the availability of equipment. To make testing easier, portable spirometers are used increasingly often. The patient does not need to appear in person at the doctor’s office. However, both the patient and the doctor need to know if the test result has been disturbed or distorted by coughing. That is why automatic algorithms are needed to accurately detect coughing and warn in case of incorrect measurement in real time.
Our PhD students and a doctor from the Medical University of Warsaw decided to face the challenge.
In their research, they used an algorithm based on neural networks. They “taught” it to detect coughing thanks to data collected by the AioCare mobile spirometry system (created by a Polish HealthUp company) and spirometry curves from the NHANES database of the American National Center for Health Statistics. In this way, the solution will be useful for different patients.
What is more, in their work they focused on the signals of air flow passing through the spirometer rather than on the much more popular sound analysis. Thus, they significantly minimized the impact of environmental noise which is troublesome for test results.
The authors of the work point out that, according to their knowledge, this is the first publication on a fully repeatable description of the algorithm of automatic cough detection based entirely on air flow signals.
The described functionality is, on the one hand, a chance to improve the quality of measurements conducted at home using portable spirometers and, on the other, it may be helpful in quickly assessing the quality of the test.
