| Popis: |
Asthma is an inflammatory disease of the airways which causes cough, chest tightness, wheezing and other distinct sounds during breathing. Spirometry is a golden standard lung function test, is used to monitor and diagnose asthma. Spirometry is very time-consuming and requires a lot of patient's efforts. Therefore, an alternate method which can overcome spirometry limitations is required. Sound based method can be one such alternative as it is less tedious, less time consuming and suitable for patients of all ages. It has been shown in the past that breath, among other vocal sounds, performs the best for an asthma vs healthy subject classification task. Breath consists of two phases, namely, inhale and exhale. Experiments in this work show, exhale performs better for classification task compared to the entire breath cycle as well as the inhale. However, this requires manual marking of the breath boundaries, which is a very time-consuming task. We, in this work, investigate how critical are the breath cycle and breath phase boundaries for the classification task. Experiments with chunks of random duration shows that they perform on par or better than the exhale. However, a segment comprising the second and third quarters of a breath cycle results in the highest classification accuracy of 80.64%. This suggests that, while breath phase boundaries may not be important, breath cycle boundaries could benefit in the classification task. |
