A portable spirometer using machine learning approach

Wheezing is a prime non-communicable disease among all age groups. It affects the air flow towards lungs by inflaming and narrowing the passage. Asthma relevant mortality occur in low and lower-middle revenue countries. In global arena, according to a report by world health organization about 417,918 deaths occur due to asthma. These sort of chronic obstructive pulmonary disease affects more than 10 million people every year. Spirometry is the universally used test for pulmonary functions. The obstructive and restrictive disease diagnosis and prognosis could be evaluated with this test. In this work, a lab model of a portable spirometer is aimed to be developed. This spirometer utilizes a flow sensor to acquire the respiratory signal and transmits the signal to Raspberry Pi for further processing. The identification or prediction of certain lung disease could be identified with the help of demographic parameters and particular symptoms of the subjects. Prime attributes are extricated from the acquired signals. Automatic decision support is executed via algorithms such as support vector machine, Naive Bayes, voting classifier are implemented to demarcate wheeze from normal subjects. Results demonstrate that the methodology adopted here is a promising method for lung disease identification and classification.