A Deep-Learning Approach to Find Respiratory Syndromes in Infants using Thermal Imaging

Respiratory syndromes being one of the most recurrent issues in a neonate, our methodology involves detection of respiratory rates to identify different types of respiratory syndromes in an infant. Most of the monitoring techniques involve an invasive monitoring approach, which may bring uneasiness to the patient. In our approach, we use a non-invasive method to classify respiratory diseases using a deep learning model in thermal imaging. A deep learning neural network is created using Keras which classifies the respiratory rates into four classes namely Tachypnea, Bradypnea, Healthy and No information. This model gives a recall and precision of 0.92.