Fig. 6
From: Dimension reduction and outlier detection of 3-D shapes derived from multi-organ CT images
Maximum number of feature matrices (organs) per individual located inside the same 2-dimensional bagplot region, comparing the dimension reduction methods MCIA, PCA and t-SNE. For both datasets, CT-ORG (n = 41) and AbdomenCT-1k (n = 50), MCIA and t-SNE show a significantly (p < 0.05) higher robustness of multiple organs per individual being projected into the same bagplot region in contrast to PCA. The data points were jittered for visualization purposes: the real count value of a data point corresponds to the nearest integer