Emerging artificial intelligence applications in liver magnetic resonance imaging

Figure 1 A deep convolutional network. An example deep learning segmentation network, the U-Net. A series of convolutions, combined with downsampling and upsampling to learn feature maps at different scales, are used to output a segmentation map. An example of a convolution, and feature map are shown on the right.

Figure 2 Classification algorithms and their performance metrics. Artificial intelligence classification algorithms use the combination of data provided to them and output a class probability. They are often evaluated according to the metrics on the right. MR: Magnetic resonance; AUROC: Area under the receiver operating characteristic curve.