Skip to main content
. 2022 Jul 29;12:13087. doi: 10.1038/s41598-022-16976-9

Figure 1.

Figure 1

t-SNE plot for all the extracted features of the training set from the first-phase training. The features are extracted from the fine-tuned 3D ResNet-18 model and are 512-d vectors, which are then embedded in the 2D space by t-SNE for visualization. The two groups of positive (orange) and negative (blue) samples are separated well in the 2D space.