Skip to main content

Table 1 Performance metrics of the Resnet50 and Xception architecture on our dataset

From: Assessment of deep learning algorithms to predict histopathological diagnosis of breast cancer: first Moroccan prospective study on a private dataset

Average accuracy (%) Confusion matrices Performance evaluation (%)
Actual Predicted Metric Class
Group 0 Group 1 Group 2 All Carcinoma vs. non-carcinoma Group 0 vs. Group 1 Group 0 vs. Group 2 Group 1 vs. Group 2
Resnet 50 model
 F1 91           
 F2 84 Group 0 142 4 6 152 Sensitivity 93 92 93 88
 F3 85 Group 1 9 50 11 70 Specificity 87 94 92 84
 F4 80 Group 2 12 9 85 106 Precision 88 84 87 90
 F5 91 All 163 63 102 328 Accuracy 90 93 92 87
 F6 76  
Xception model
 F1 90           
 F2 85 Group 0 144 4 4 152 Sensitivity 95 93 95 93
 F3 81 Group 1 10 54 6 70 Specificity 88 93 94 90
 F4 87 Group 2 9 6 91 106 Precision 89 84 91 93
 F5 95 All 163 64 101 328 Accuracy 91 93 94 92
 F6 82  
  1. The bold data in the confusions matrices have a significance, It means the number of cases that were correctly predicted in each group
  2. Accuracy: average accuracy for three-classification task, using Resnet50 and Xception models, evaluated over sixfolds via cross-validation
  3. Confusion matrices without normalization using Resnet50 and Xception models: vertical axis—ground truth, horizontal—predictions
  4. Performance evaluation: performance metrics of ResNet50 and Xception models for the binary and 3-class classification