Microphone based Smartphone enabled Spirometry Data Augmentation using Information Maximizing Generative Adversarial Network

Spirometry is gaining popularity in the primary care and point-of-care scenarios in the medical community for constant monitoring of lung condition and early diagnosis of chronic lung ailments. Accurate classification of spirometry data is necessary to diagnose lung ailments properly. Deep learning-based classification models, free from manual feature extraction, are quite accurate. One of the major challenges of deep learning-based classification approach in spirometry is its small size of the datasets. In this paper, a user-friendly smartphone application that performs spirometry using built-in mobile microphone has been developed and an information maximizing generative adversarial network (InfoGAN) model has been proposed that can generate spirometry signal to augment the dataset and enhance the performance of the deep learning-based classification models.