Year: 2017
Type: dissertation
Abstract: Purpose: To develop and test methods of correlating whole-mount pathology of head and neck (H) samples to pre-operative radiological imaging. The ultimate goal of this study is to correlate pathology samples of H cancers to pre-operative computed tomography (CT) imaging and navigated endoscopy. Methods: Tongue and mandible samples from non-survival pig studies were used. We assessed tissue shrinkages and proposed methods to enhance the registration accuracy. A slicing apparatus was designed that generates consistent tissue sections to aid histology to pre-operative registration. Finally, the whole registration pipeline was tested using intermediate steps. Results: The fiducial registration error (FRE) from histology to pre-operative imaging was 1.94 mm for tongue and 1.07 mm for mandible. Conclusion: We demonstrated the feasibility of registering pathology findings with volumetric imaging. The registration error can further improve by utilizing techniques of deformable registration and adding more landmarks in our specimens to calculate the registration error reliably.
Source: TSpace (University of Toronto)
Author: Niousha Aflatouni
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2781425798