Improving lung region segmentation accuracy in chest X-ray images using a two-model deep learning ensemble approach

Md Fashiar Rahman, Yan Zhuang, Tzu-Liang Bill Tseng, Michael Pokojovy, Peter McCaffrey, Eric Walser, Scott Moen, Alex Vo

Abstract: Highlights•Intensive bacterial/viral infection reduces lung segmentation accuracy in CXR films.•Patching technique is effective to train deep learning architecture with a small dataset.•Ensemble of U-Net and CNN models complements each other to capture missing pixels.•Major advantages observed for heavily infected CXRs with improved accuracy.