FIGURE 10.

Qualitative results from the performance comparison between our CAD system versus state-of-the-art CAD system on Clinical Dataset having a confidence threshold of 95%. Green represents the Ground truth box. Red as predicted by Retina U-Net [37] and blue represents our proposed model predictions. Three random cases out of 120 cases from Shanghai Sixth People’s Hospital are selected. Left-most images (a) show the ground truth of lung CT scan, images (b) in all cases show the box-predictions of Retina U-Net (2 foreground classes) as well as the prediction of our proposed 3DDCNN model, images (c) represent the remaining predictions after prediction of nodule along with some False Positive predictions, images (d) After the False Positives reduction step, leaving the correct prediction with confidence scores as high as 87%.