Fig. 2: Receiver operating characteristic (ROC) curves of methods on classifying adenocarcinoma and squamous cell carcinoma. | Nature Communications

Fig. 2: Receiver operating characteristic (ROC) curves of methods on classifying adenocarcinoma and squamous cell carcinoma.

From: An annotation-free whole-slide training approach to pathological classification of lung cancer types using deep learning

Fig. 2

Area under the ROC curve (AUC) was measured for each data set and method. All the measurements were taken from distinct samples. a, b ROC curves on the standard testing data set (n = 1397). c, d ROC curves on the small lesion data set (n = 476). e, f ROC curves on TCGA-diagnostic data set (n = 1044). g, h ROC curves on TCGA-tissue data set (n = 2167). The dark blue, cyan, and red lines represent CNN-MaxFeat-based RF, MIL-RNN, and whole-slide training method, respectively.

Back to article page