Digital archive of theses discussed at the University of Pisa


Thesis etd-01122023-084532

Thesis type
Tesi di laurea magistrale
Thesis title
Multimodal Image Registration Workflow for Postoperative Assessment of Transcatheter Mitral Valve Annular Reduction
Course of study
relatore Prof. Positano, Vincenzo
correlatore Prof.ssa Celi, Simona
supervisore Ing. Garzia, Simone
  • registration
  • multimodal
  • annulus
  • TMVI
  • software
Graduation session start date
The thesis proposes a multimodal image registration workflow for postoperative assessment of transcatheter mitral valve annular reduction. In the studied clinical case, various imaging modalities were used, such as Computed Tomography (CT) for planning the surgery, Fluoro-Angiography (XA) to guide the surgeon in inserting catheters and the device to be implanted, and Echocardiography (US) to visualize in detail the anatomy of the target of interest in the intraoperative phase. The different images acquired are commonly displayed separately, causing difficulties for physicians and surgeons when they need to merge them into a single visualization, such as for postoperative evaluations. Therefore, the combination of these imaging modalities with temporal and spatial consistency can be an important and useful tool in clinicians' hands. The presented workflow, tested on five imaging datasets of patients who have undergone surgery at "Fondazione G.Monasterio- Ospedale del Cuore", is composed of three fundamental steps which are mainly performed using an open-source software. The first step is a semi-automatic rigid landmarks-based registration between a preoperative CT volume and an intraoperative US volume, executed by manually placing fiducial points on the anatomic target. The second step consists of a rigid registration between the previous CT volume and an intraoperative XA image, performed through a projective reconstruction approach to find the CT image projection matching the XA image. By exploiting the two registration outcomes, the third step consists of obtaining an automatic rigid registration between the US volume and the XA image. Finally, a postoperative assessment, after the registration of post-interventional US and XA data, is executed by comparing intra- and postoperative anatomic areas of manually segmented ROIs. Results show a good registration accuracy consistency among the five patients, making the workflow reproducible on such data. The results also show that registration accuracy is highly dependent on several factors like image quality and user experience on correctly placing fiducial points.