skip to content

3D Multimodal Medical Imaging

Supervisors: Prof. Laurence S. Dooley, Dr. Patrick Wong and Dr.-Ing. Anko Börner (German Aerospace Center (DLR), Germany).

Student: Mr. Parminder Singh Reel

Medical image registration is the systematic process of aligning medical images of same or different modalities (such as computed tomography (CT) or magnetic resonance imaging (MRI)). A source image is geometrically transformed to achieve the best physical alignment with a reference target image. An optimization method is applied to maximize some predefined similarity measure with known transformations between these source and reference images. This process becomes challenging in the presence of intensity non-uniformity and large homogeneous non-vascular regions for various regions of human anatomy such as brain and retina.
This research explores the role of similarity measure in the medical image registration framework. It proposes a new similarity measure which efficiently incorporates spatial information together with mutual information without incurring high computational overheads. Test datasets of brain and retinal (mono and multimodal) images have been used for investigating the new similarity measure. Both quantitative and qualitative experimental results clearly demonstrate both improved robustness and lower computational complexity of the proposed paradigm compared with existing similarity measures, for various test datasets.