Abstract
[Summary] Measurement of 3D deformation is indispensable to orthopedic image diagnoses. But so far it is difficult to quantify the deformation between two images obtained at different opportunities accurately, because correspondence between them is obscure. In this study we developed a new method to match such images by elastic matching and quantify the deformity. [Method] We implemented an elastic matching method, which measures the matching energy measure E as sum of voxel similarity measure Em and elastic strain energy R. We defined Em as root mean square of the differences of voxel value between two images. After matching, displacement vectors for all voxels in pre-injury image are obtained and the difference in two of them is the relative change in location of the two points. [Tests] We used MRI of ACL deficient knee case (suspected). Firstly, as an accuracy test, arbitrarily deformed MRI scans was restored by elastic matching and restoration errors in lengths and angles were measured. The results showed that the restoration errors in length were less than the size of a voxel. The accuracy of this method was sufficient. Secondly, as a clinical test, we measured the changes in coordinates of ACL bony insertions and length of ACL from MRI scans of the same case obtained before and after re-injury. The measured coordinates of tibial insertion were less accurate by usual manual method than by our method. The measured lengths were almost equal. [Conclusion] We developed a new method to quantify 3D orthopedic deformity by elastic matching. Non-linear biological deformities can be quantified three-dimensionally by this method.
