Development, Validation, and Integration of an Ultrasound-guided Interventional System for Liver Cancer
Abstract
Tumour ablations are a curative, minimally invasive treatment option for early-stage liver cancer, particularly for patients ineligible for surgery. Ultrasound (US) is a commonly used imaging modality for guiding therapy applicator placement due to its real-time imaging, wide availability and radiation-free nature. Conventionally, physicians have to mentally relate two-dimensional (2D) US images to the three-dimensional (3D) location of the internal target, while simultaneously manipulating the US transducer and the applicator for in-plane insertion. This process heavily relies on the physician’s experience and long-term training to achieve consistent procedure performance. Additionally, the difficulty of accurately identifying tumours further constrains the clinical applicability of US-guided ablation procedures, particularly in scenarios involving similar acoustic characteristics between tumours and surrounding tissues, irregular tumour locations, and tumour mimics (e.g., cirrhotic regenerative nodules). To address these limitations, this dissertation aims to develop clinically feasible, intra-procedural US-guided approaches to improve tumour identification and lesion targeting. Specifically, it focuses on three key aspects: intra-procedural tumour coverage assessment, tumour identification, and therapy applicator localization.
To ensure complete tumour coverage, we introduced 3D US image-based volumetric metrics to identify any uncovered tumour region. Additionally, a novel margin uniformity-based approach was developed to provide prioritized applicator adjustment instructions when necessary. To improve tumour identification, we proposed a 2D US-computed tomography (CT)/magnetic resonance imaging (MRI) registration workflow. This included an automatic, liver vasculature-based approach for rigidly aligning 3D US with diagnostic CT/MRI images, followed by the development of a deep regression 2D-3D US registration model for real-time correction of rigid liver motion. Furthermore, a thin plate spline-based deformable refinement tool was integrated with a novel visualization approach, enabling intuitive confirmation of alignment prior to guidance and facilitating interactive volume deformation when required. To improve applicator identification, we investigated the feasibility of integrating magnetic tracking into the operating room, particularly for C-arm-equipped. As an extended study, we demonstrated the clinical efficacy of the integration of magnetic tracking in benefiting fluoroscopy-guided interventions.
Overall, we believe these advancements can improve and facilitate US-guided liver tumour ablation, thereby expanding the capabilities of 3D US imaging in US-guided interventions for liver cancer.