A Practical Deep Learning Model in Differentiating Pneumonia-Type Lung Carcinoma from Pneumonia on CT Images: ResNet Added with Attention Mechanism

Figure 3

5-fold validation ROC curves: the black dash line means the average of 5 curves, and the gray dash line means a model without any predictive ability. The average AUC value of FROC was 0.82 after five-fold cross-validation. The average accuracy of cross-validation was 74.2%. The model introduced false-positive in the process of distinguishing pneumonia and lung cancer, but the overall accuracy had a relatively strong reference value. It shows that the model has high credibility.