Facilitating Breast Conserving Surgery Using Preoperative MRI

prone to supine breast mri registration;image registration;image segmentation;breast mri;image processing

Rincon, Joseadvisor:Michailovich, Oleg ; affiliation1:Faculty of Engineering ; Michailovich, Oleg ;

University of Waterloo

关键词: image processing;    prone to supine breast mri registration;    Master Thesis;    image registration;    image segmentation;    breast mri;   

Others  :  https://uwspace.uwaterloo.ca/bitstream/10012/13886/3/Rincon_Jose.pdf

瑞士|英语

来源: UWSPACE Waterloo Institutional Repository

【 摘 要 】

Breast cancer is currently considered the most widespread malignancy in women, which costs the lives of approximately 400,000 people annually worldwide. While extremely useful for early detection and diagnosis of breast disease, the application of MRI to pre-operative planning of breast conservative surgeries is complicated due to the differences in the patient's posture at the time of imaging and surgery, respectively. Specifically, while MRI is standardly performed with patients positioned with their face down and their breast unrestricted and pendulous, breast surgeries normally require the patients to lie on their back, in which case the breast undergoes substantial deformations due to the effect of gravity. As a result of these deformations, pre-surgical MRI images frequently do not correspond with the actual anatomy of the breast at the time of surgery, which limits their applicability to pre-surgical planning. Accordingly, to overcome the above problem and make the MRI images align with the actual intra-surgical anatomy of the breast, the images need to be properly warped - a procedure that is known as prone-to-supine image registration. In many cases, this registration is carried out in two steps, prediction and correction. While the former involves bio-mechanical modeling used to describe the principal effect of tissue deformation, the latter refines the preceding results based on the image content. What is more important, however, is the fact that the accuracy of the correction step (and, hence, of the registration process as a whole) is strongly dependent on the accuracy of bio-mechanical modeling, which needstherefore be maximized as much as possible. Consequently, the fundamental objective of this research project has been the development of algorithmic solutions for reliable and accurate prediction. In particular, we propose an automatic detection of thelocation and geometry of the breast, and a breast image segmentation method to differentiate between adipose and dense tissue that is tractable, stable, and independent of initialization.