1. To go to the next image, press the Next button or the right arrow key or space bar. To go to the previous image, press the Previous button or the left arrow key. You can also type the case number in the edit box and press Enter or the Go To button. With the check boxes you can specify which results and which evaluation measures you want to be displayed.
2. The overlap and mean contour distance are listed for, respectively, right lung, left lung, heart, right clavicle, left clavicle. Note that the right lung and clavicle are on the left of the image.
3. All images shown here are subsampled to 256 by 256 for display purposes and are jpeg compressed. Brightness and contrast have been adjusted using an automatic algorithm. Segmentation results are visualised by plotting object pixels on a boundary in white and background pixels touching the boundary in black.
4. Manual outlines produced by a human observer. See  for details.
5. Manual outlines independently produced by a second human observer. See  for details.
6. This method is described in . Data provided by Bram van Ginneken.
7. This method is described in . Data provided by Mikkel Stegmann.
8. Results from a novel segmentation algorithm that combines shape modeling, appearance modeling and dynamic programming . Data provided by Dieter Seghers, Medical Image Computing, KU Leuven, Belgium.
9. The overlap of two object (the reference and the segmentation) is defined as the area of intersection divided by the area of the union.
10. For two objects A and B, the mean contour distance is the average distance from a point on the contour of A to the nearest point on the contour of B, and vice versa, averaged. See  for details. The distances listed are in millimeter, the dimension of the image is 358 by 358 millimeter.
11. Pixel error is defined as the part of all pixels in the image for which the segmentation is not the same as the reference for any of the five objects considered here.
B. van Ginneken, M.B. Stegmann, M. Loog,
Segmentation of anatomical structures in chest radiographs using supervised methods: a comparative study on a public database,
Medical Image Analysis, nr. 1, vol. 10, pp. 19-40, 2006.
- D. Seghers, D. Loeckx, F. Maes, P. Suetens, "Image segmentation using local shape and gray-level appearance models", in Medical Imaging 2006: Image Processing, Proceedings of SPIE, vol 6144, 2006.