Fig. 5: Comparison of 1D CNN and HD-CNN performance for dual-cancer classification. | npj Digital Medicine

Fig. 5: Comparison of 1D CNN and HD-CNN performance for dual-cancer classification.

From: Advanced breath analysis through hierarchical deep convolutional neural network for multi-cancer screening

Fig. 5: Comparison of 1D CNN and HD-CNN performance for dual-cancer classification.

a–c Classification results of the 1D CNN model on the 5-fold cross-validation dataset. a Scatter plots showing predicted class probabilities for HC, LC, and GC groups. b Confusion matrix for 3-class classification. c One-vs-rest ROC curves and AUC values with 95% confidence intervals for each class. d–f Corresponding results of the HD-CNN model on the 5-fold cross-validation dataset. d Scatter plots of predicted probabilities, e confusion matrix, and f one-vs-rest ROC curves and AUC values of the HD-CNN model. Radar plots comparing accuracy, precision, recall, and F1-score between two models for g training dataset, and h validation dataset.

Back to article page