Non-iterative partial view 3D ultrasound to CT registration in ultrasound-guided computer-assisted orthopedic surgery

In image-guided orthopedic surgery, rigid registration of intra-operative ultrasound (US) to a pre-operative plan, developed using computed tomography (CT) scans, is an important step for providing real time surgical guidance. The ability to perform this registration accurately, automatically, and in real time is critical for enabling more effective image guidance and anatomic restoration in a number of orthopedic procedures. Several surface-based and intensity-based registration methods have been proposed before to align the US and CT data. Although relatively successful results were reported, both methods require accurate segmentation or localization of the bone surface in US data, which is a challenging task. Furthermore, typically, only partial views of the three-dimensional (3D) bone anatomy are visible in US data, and registration would only converge if a good estimation of the initial alignment between the US and CT datasets is known.

We propose a 3D rigid CT to US registration method based on the alignment of local phase bone image projections. The registration is achieved by transforming the local phase bone features, calculated using 3D Log-Gabor filters, to a projection space obtained using 3D Radon transform. Validation experiments show the capability of the method in registering partial view US volumes to full view CT volume.

Feasibility experiments, carried out on a phantom and ten volunteer subjects, show an average surface registration, in the region where the US scans were acquired, of 0.42 and 0.78  mm, respectively.

The proposed US to CT registration method is fully automatic, non-iterative and requires no initial alignment between the two registering datasets.