ResNet50V2: A Transfer Learning Model to Predict Pneumonia with chest X-ray images
Sashikanta Prusty, Srikanta Patnaik, Sujit Kumar Dash
Abstract
In the Covid-19 pandemic days, it is critical to diagnose the disease due to less availability of medical beds and of having a large number of infected people. Thus, initial screening of the chest is much needed to avoid the diseases like Covid-19, lung cancer, heart disease, and many other conditions before these are caused to severe. However, the biggest advancement of the X-ray machine is that it uses a very modest dosage of ionizing radiation to obtain images of the inside of the chest. In the meantime, Deep learning (DL) in healthcare provides a prominent solution to identify the disease from the huge amount of medical images. The DL-based Transfer Learning (TL) model acts as a powerful technique for extracting feature when there have been less amount of data, and also has the potential to provide a promising solution. Thus, a TL model called ResNet50V2 has been proposed for detecting pneumonia from 5216 sample images. This model has been evaluated successfully with 10 epochs and predicts pneumonia with an accuracy of 99.69%. This might also help the doctors to predict other diseases like lung cancer, Covid-19, and Heart failure before these become to death.