Pneumonia lung opacity detection and segmentation in chest x-rays by using transfer learning of the Mask R-CNN

As people become advanced day by day and they move towards automation for their comfort. The machine is also being smarter and accurate due to the availability of a large amount of data and fast computing power. So as a result machine learning becomes an important pillar in life. People now become more reliable to machine than man decision. So here we apply Mask R-CNN to chest x-rays for detection and segmentation of the area of the lungs where pneumonia lungs opacity are present. For this, we take a dataset from the radiological society of North America (RSNA) that is also available on the Kaggle website for pneumonia detection competition. We apply transfer learning of the Mask R-CNN model i.e. pre-trained model on the COCO dataset we modify the model according to our needs and requirement for the RSNA x-rays dataset. We trained the Mask R-CNN model for this dataset for medical investigation and drew the different types of losses at the time of training. To check the performance of the trained model we pass the test x -rays image to the model as input that gives segmented and detected images region with confidence score affected from pneumonia lungs opacity as the output. Thus this is a step towards automation of pneumonia lung opacity detection and segmentation in x-rays. So this method makes fast and accurate pneumonia lungs opacity detection and segmentation automatically.