Ensemble Method of Lung Segmentation in Chest Radiographs

Lung segmentation plays a crucial role in computer-aided diagnosis using chest radiographs (CRs). In this research, we present an ensemble approach of segmenting lungs using U-Net and DeepLabV3+ in CRs across multiple datasets. We utilize publicly available datasets such as Japanese Radiological Scientific Technology (JRST) and Shenzhen datasets for testing. Our overall performance in terms of global accuracy is 98.6%, an IoU (intersection over union) of 0.97 for a set of 100 test cases thereby setting a new benchmark for future research efforts.