UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Joint multimodal registration of medical images to a statistical model of the lumbar spine for spine anesthesia Behnami, Delaram


Facet joint injections and epidural needle insertions are widely used for spine anesthesia. Needle guidance is usually performed by fluoroscopy or palpation, resulting in radiation exposure and multiple needle re-insertions. Several ultrasound (US)-based guidance approaches have been proposed to eliminate such issues.However, but they have not widely accepted in clinics due to difficulties in interpretation of the complex spinal anatomy in US, which leads to clinicians' lack of confidence in relying only on information derived from US for needle guidance. In this thesis, a model-based multi-modal joint registration framework is introduced, where a statistical model of the lumbar spine is concurrently registered to intraprocedure US and easy-to-interpret preprocedure images. The goal is to take advantage of the complementary features visible in US and preprocedure images, namely Computed Topography (CT) and Magnetic Resonance (MR) scans. Two versions of a lumbar spine statistical model are employed: a shape+pose model and a shape+pose+scale model. The underlying assumption is that the shape and size of the spine of a given subject are common amongst all imaging modalities . However, the pose of the spine changes from one modality to another, as the patient's position is different at different image acquisitions. The proposed method has been successfully validated on two datasets: (i) 10 pairs of US and CT scans and (ii) nine US and MR images of the lumbar spine. Using the shape+pose+scale model on the US+CT dataset, mean surface distance error of 2.42 mm for CT and mean Target Registration Error (TRE) of 3.14 mm for US were achieved. As for the US+MR dataset, TRE of 2.62 mm and 4.20 mm for the MR and US images, respectively. Both models models were equally accurate on the US+CT dataset. For US+MR, the shape+pose+scale model outperformed the shape+pose model. The joint registration allows augmentation of important anatomical landmarks in both intraprocedure US and preprocedure domains. Furthermore, observing the patient-specific model in preprocedure domains allows the clinicians to assess the local registration accuracy qualitatively. This can increase their confidence in using the US model for deriving needle guidance decisions.

Item Citations and Data


Attribution-NonCommercial-NoDerivatives 4.0 International