Manual Alignment of CT and Ultrasound Images for Computer- Assisted Shoulder Arthroscopy

Many computer-assisted orthopaedic surgery applications employ a three-dimensional surface model of the anatomy of the patient for pre-operative planning. Registering the model to the coordinates of the patient in the operating room usually requires additional intra-operative images of the patient. To achieve accurate registration results using registration algorithms such as iterative closest point, a good initial alignment of the pre-operative and intra-operative data is essential. This paper presents an approach for obtaining an optimal initial alignment to register pre-operative computed tomography images to intra-operative ultrasound images of the shoulder. First, sequential computed tomography images of the shoulder bone are acquired from which a three-dimensional model is built. Then, ultrasound images of three landmarks of the scapula bone are captured. An initial alignment is obtained by matching three manually selected points on the model to their corresponding points identified on the ultrasound images. Results of laboratory experiments on a shoulder phantom indicate acceptable initial alignment to be used with iterative closest point registration algorithm. In addition, the technique requires only three ultrasound images, making the initial registration process fast and applicable for an ultrasound-guided navigation system for shoulder arthroscopic surgery.