In this study the registration technique consist of fitting and tracking. The fitting approach mounts almost 800 t0 1000 landmarks on the surface of demanded objects (knee bones) using balloon-like strategy. Stating from an initial bone model the algorithm gradually reaches to the bone boundaries of undergoing images by iterative inflation constrained by voxel gray level gradient and etc.

Resulted patient depended bone model will be employed in next step in which the amount of object transform will be estimated. The registration technique is inherently a conventional tracking scheme to minimized voxel to voxel gray distortion at the landmark positions. Total transform parameters will be estimated from landmarks acquired motion vectors.

In order to deal with the different voxel aspect ratio between reference (knee extension)   and target (knee flexion position) images, a straight map (multiplication) transforms the coordination of landmarks before commencing the registration process. This factor is also applied to the transform parameters estimation by compensating the difference of VAR while computing the motion vectors.

 

Esfandiarpour F*, ShakooriRad A, Farahmand F, Eslami A

* Medical School,Tehran University of Medical Science, Tehran, Iran