Skip to main content
. 2022 Dec 21;12:22059. doi: 10.1038/s41598-022-26328-2

Figure 3.

Figure 3

Processing of single-modal (top row) and multi-modal (bottom row) MRI sequence data at the input level of the 3D U-Net in use (nnU-Net23). The implementation uses 2D kernels at the first convolutional layer #1 since it treats the MRI sequences as anisotropic data.