Clinical implementation of surgical liver navigation: Advancing ultrasound-guided techniques

Surgical treatment is the gold standard with the best long-term survival for patients with colorectal liver metastases. During the procedure it can be challenging to find all lesions and achieve a radical resection or complete ablation. Ultrasound (US) is used during intraoperative localization and surgical decision making, however, this is not always sufficient. A possible solution to provide the surgeon with improved guidance is surgical navigation, where surgical instruments are virtually displayed with respect to preoperative imaging and patient-specific 3D models. However, nowadays these surgical navigation systems are not implemented in the surgical workflow as accurate image registration and guidance are challenged by the natural flexibility of the liver.
This thesis aims to improve several aspects of surgical liver navigation. Ultimately, this should advance integration of surgical liver navigation in the clinical workflow and improve treatment options for patients with colorectal liver metastases.
First, intraoperative 3D US is assessed as an alternative for 2D US for image registration purposes. In this thesis, two types of 3D ultrasound are explored for intraoperative use. Freehand US volumes were acquired by electromagnetically-tracked 2D US transducers. Alternatively, a dedicated abdominal transducer for 3D US acquisition was used. Second, the use of electromagnetic tracking is examined as a possible solution to locally compensate for intraoperative liver movement caused by natural movement and surgical manipulation. Third, the use of deep learning-based segmentation of hepatic vasculature in US is assessed. Subsequently, nonrigid image registration is examined to compensate for deformation between the preoperative scan and the intraoperative shape of the liver.
This thesis demonstrates satisfactory accuracy by using electromagnetic tracking of the targeted organ in combination with 3D US. Feasibility of automatic vasculature segmentation followed by nonrigid registration improves accuracy when compared to manual image registration. The investigated navigation setup is particularly useful in patients with small liver lesions, as it guides the surgeons in retrieving their locations.