Multimodal Integration of Medical Ultrasound for Treatment Planning and Interventions

Ultrasound is a popular, cost-effective and non-invasive medical imaging modality. If an ultrasound probe is equipped with a 3D position sensor, the acquired images can be obtained in their spatial context, a technique commonly denoted “3D freehand ultrasound”. Combining ultrasound and a pre-operative Computed Tomography (CT) scan of the same patient can be beneficial for a number of clinical applications. The core of this thesis is the development of novel methods for fully automatic alignment (i.e. registration) of 3D freehand ultrasound and CT data, based on the image content and the physical properties of both modalities. In particular, a new method is proposed which simultaneously optimizes the linear combination of different ultrasonic effects simulated from CT, and the parameters for spatial alignment. This original concept allows local as well as global optimization of the simulation parameters, resulting in optimal registration of any modalities, where usual registration solutions do not succeed and explicit simulation of complex effects are necessary. Furthermore, we introduce new techniques for 3D freehand ultrasound calibration and reconstruction, as well as visualization of fused CT and ultrasound data. Two clinical applications are investigated in detail. We use a designated version of an automatic registration algorithm to integrate diagnostic ultrasound into radiation treatment planning for head and neck cancer. Our simultaneous optimization of simulation and registration is applied, in the context of treating liver and kidney metastases, for fusing CT with both diagnostic and interventional ultrasound of the abdomen. While diagnostic fusion helps doctors to assess indeterminate lesions in those organs, interventional fusion using our techniques allows for advanced image-guided navigation, in particular, for needle biopsies and radio-frequency ablations.