Fig. 2
From: Dimension reduction and outlier detection of 3-D shapes derived from multi-organ CT images
Multiple Co-Inertia Analysis (top: CT-ORG dataset, bottom: AbdomenCT-1k dataset). A, C Sample space derived from MCIA. Individuals are projected by the geometry of their selected organs into the same 2-D space. Different point shapes illustrate the organ features a sample point is based on. The liver, left lung, right lung, left kidney and right kidney (A) as well as liver, pancreas, spleen, left kidney and right kidney (C) originating from the same individual are connected by lines that meet at a common center point. The shorter the lines, the higher the correlation of samples. Each individual is labeled by a number. B, DÂ Variable space projecting each feature from all feature matrices into the same 2-dimensional space. The further away a feature is projected from the point of origin in the same direction as a sample, the higher the value of that feature in the respective sample