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