Guidelines for Parameter Value Selection

The multiscale enhancement filter includes several parameters. In defining the similarity measures to local structures, yst and a in Eqs. (10.3) and (10.4) need to be specified, while in the multiscale integration, the smallest scale a1, the scale factor s, and the number of scale levels n in Eq. (10.17) need to be specified.

With regard to the parameters in the line measure, if there are strong sheet structures to be removed from an image, y23 should be 1.0. If the cross sections of line structures of interest vary not only in size but also in shape, y23 should be 0.5. y12 also should be determined to optimize the trade-off between the preservation of branches and the reduction of noise and spurious branches. However, the performance has been found to be relatively insensitive to the values of y23 and y12 as long as they are between 0.5 and 1.0. Also, we have empirically found that a = 0.25 can be considered a good compromise.

With respect to the parameters in multiscale integration, we experimentally found that s should be 1.5, V2, or less for reasonable approximation of the multiscale integration of the continuous scale. The minimum value of a1 for discrete samples of volume data was 0.8 (voxels), which are experimentally shown in [7]. Considering the above observations, suitable values for a1 and n should be found out for each case taking account of the anatomical structure of interest based on the width response curves as shown in Fig. 10.3. We confirmed that the results were quite stable for different images obtained under similar conditions once suitable values have been determined.

The detailed analyses of the effects of parameter values on the filter responses, which are the bases of the above guidelines, are discussed for the line case in the next section, and more thorough analyses of them are found in [7].

10.2.4 Examples

0 0

Post a comment