Fig. 7: A visual explanation of the DNN’s decision for representative BOS patients.
From: Harnessing deep learning to detect bronchiolitis obliterans syndrome from chest CT
a Left: a CT slice; Right: the same slice with the attention map obtained using Grad-CAM algorithm. b DNN interpretation on CTs reconstructed with thin and thicker slices obtained at the same radiology visit. c DNN interpretation prior BOS diagnosis. Scale bar, 1 cm. DNN deep neural networks, BOS bronchiolitis obliterans syndrome.
