2020 Volume 38 Issue 3 Pages 126-131
Children have a high risk of pneumonia infection due to low immunity and collective life living in a group. Therefore, accurate diagnosis and early treatment are required. The purpose of this study is to develop the decision support system for thoracic diseases using chest X-ray images. As a pilot study, we propose the extraction method of the lung region using Mask R-CNN. Mask R-CNN contains the processing of object detection and semantic segmentation. In this method, 1000 images (child images; 200, adult images; 800) from the open database published by NIH were used to train Mask R-CNN. As a result of evaluation, average of Jaccard index was 93.3% and Dice index was 96.5%. Therefore, high accuracy of lung field extraction has been obtained even if various chest X-ray images of children such as lung field size exist.
