Skip to main content


Figure 2 | BMC Bioinformatics

Figure 2

From: Method: automatic segmentation of mitochondria utilizing patch classification, contour pair classification, and automatically seeded level sets

Figure 2

Automatic results for mitochondria segmentation in SBFSEM images. XY slices shown here are 2.1 μm × 3.1 μm. (a) Original data for Test 1. (b) Step 1: Voxel classification result and contour detection. (c) Step 2: Detection of isocontours. (d) Salient contours after contour classification. Notice that many contours that do not belong to mitochondria have been eliminated. (e) Step 3: Result of level set operation seeded with inner region of salient contours. Note that the small isolated sets of pixels in (e) are a result of the 3D level set operation spreading into a few pixels when seeded from an adjacent plane. Also, observe that contours occasionally are eliminated in the contour pair classification set but re-appear after the level set operation. This is the result of contour classification eliminating a contour in one plane but accepting contours in nearby planes above or below. Then salient contours above and below act as seed point regions, and the 3D level set operation tends to fill in the region where the contour was eliminated.

Back to article page