Figure 4 | Scientific Reports

Figure 4

From: 2D and 3D convolutional neural networks for outcome modelling of locally advanced head and neck squamous cell carcinoma

Figure 4

Ensemble Kaplan–Meier curves: Kaplan–Meier curves for patient groups at low risk (blue) and high risk (orange) of loco-regional recurrence for training and internal test folds as well as for the independent validation cohort. The stratification was created using the median of the training ensemble predictions as cutoff. The top row shows the curves obtained from an ensemble of 3D-CNN models trained from scratch based on the architecture of Hosny et al.23 with \(\tanh\) as final activation. The centre row shows the curves obtained from an ensemble of 2D-CNN models trained from scratch without batch normalisation and \(\tanh\) as final activation. The bottom row shows the curves obtained from an ensemble of transfer learning models based on DenseNet201 with the last convolutional layer as foundation.

Back to article page