Research Article

Semiautonomous Medical Image Segmentation Using Seeded Cellular Automaton Plus Edge Detector

Figure 3

Comparison of segmentation accuracy resulting from SCAPE using edge images from four different methods and three sets of seeds. The long and short parameter searches both determined edge images that resulted in high quality SCAPE segmentations. Manually selecting parameters resulted in relatively low quality segmentations. Using the same “fixed” parameters for each edge image did reasonably well with more seeds. Quality measured by comparing with a gold standard using Jaccard index, for which the scale is shown ranging from 0.7 to 0.95 to show more detail. Error bars show standard deviations.
914232.fig.003