Fig. 4: Neural network architecture—Diagram showing the ILD-model architecture used in this study. | npj Digital Medicine

Fig. 4: Neural network architecture—Diagram showing the ILD-model architecture used in this study.

From: Handling missing MRI sequences in deep learning segmentation of brain metastases: a multicenter study

Fig. 4

Five contiguous axial slices of each of the four pulse sequences (BRAVO, pre-Gd and post-Gd CUBE, and T2-weighted FLAIR) are concatenated in the color-channel dimension to create an input tensor with channel dimension 20. This is fed into the neural network to predict the segmentation on the center slice. The pipeline is built largely on the DeepLab V3 architecture21.

Back to article page