Ultrasound and computed tomography cardiac image registration
As the trend of the medical intervention moves towards becoming minimally invasive, the role of medical imaging has grown increasingly important. Medical images acquired from a variety of imaging modalities require image preprocessing, information extraction and data analysis algorithms in order for...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/78545/1/ChiengThionMingMFBME2017.pdf http://eprints.utm.my/id/eprint/78545/ http://dms.library.utm.my:8080/vital/access/manager/Repository/vital:110594 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Universiti Teknologi Malaysia |
| Language: | English |
| Summary: | As the trend of the medical intervention moves towards becoming minimally invasive, the role of medical imaging has grown increasingly important. Medical images acquired from a variety of imaging modalities require image preprocessing, information extraction and data analysis algorithms in order for the potentially useful information to be delivered to clinicians so as to facilitate better diagnosis, treatment planning and surgical intervention. This thesis investigates the employment of an affine registration method to register the pre-operative Computed Tomography (CT) and intra-operative Ultrasound cardiac images. The main benefit of registering Ultrasound and CT cardiac images is to compensate the weaknesses and combine the advantages from both modalities. However, the multimodal registration is a complex and challenging task since there is no specific relationship between the intensity values of the corresponding pixels. Image preprocessing methods such as image denoising, edge detection and contour delineation are implemented to obtain the salient and significant features before the registration process. The features-based Scale Invariant Feature Transform (SIFT) method and homography transformation are then applied to find the transformation that aligns the floating image to the reference image. The registration results of three different patient datasets are assessed by the objective performance measures to ensure that the clinically meaningful result are obtained. Furthermore, the relationship between the preoperative CT image and the transformed intra-operative Ultrasound image are evaluated using joint histogram, MI and NMI. Although the proposed framework falls slightly short of achieving the perfect compensation of cardiac movements and deformation, it can be legitimately implemented as an initialisation step for further studies in dynamic and deformable cardiac registration. |
|---|