3D shape comparison -> % of difference

Hi,

I have created two 3D matrix from two different dicom files and I would like to know whether I could compare their shape and deduce the percentage of difference between them.

I cannot really make the difference (A-B) between the 2 matrix since their orientation and the "position" within the matrix will be different.

is there a way to do so ?

thank you for your time.

Answer 1

[Jan]

This is not uniquely defined: imagine two spheres with different radius and equal center. What is the "percentage of difference"? Or two spheres with same radius but different centers?

You can define a certain number of "equally distributed" point on the inner shape and measure the mean/maximum/geometric average of the distances to the nearest point (or the point on the line perpendicular to the surface) on the other shape. But another distribution of the points will result in different distances. Therefore you need a "sufficiently" high number of points, such that small deviations of the locations do not change the result significantly anymore.

But even then a "percentage" difference need a normalization of any kind. You have to dived the result by a specific number to get a "percentage", but what could this be? E.g. for the statement "less than 50% of the Matlab Answers users are female" you have to divide the number of female users by the number of all users.

Finally I think, that there are too many choices you have to define, such that the result will be more or less arbitrary and not obviously or intuitive anymore.

Answer 2

[Sean de Wolski]

You could use imregister() (if you have R2012b) to align the two image volumes. From here you could do the typical math.

[ocelote]

Thx, it seems the R2012a just work with 2D :(

Answer 3

[ocelote]

Thank you for your reply.

You are right, my way to dealing with my issue is not correct.