Show simple item record

dc.identifier.urihttp://hdl.handle.net/1951/59783
dc.identifier.urihttp://hdl.handle.net/11401/71340
dc.titleImproving Cancer Detection Through Visualization
dcterms.abstractThe proliferation of medical scanning technology, especially computed tomography (CT) and magnetic resonance imaging (MRI), has led to various research works on methods of intuitively visualizing the large amounts of resulting slice data. The visualization of this data consists not only of display methods, but ideas covering topics such as navigation, exploration, surface parameterization, and analysis. These imaging modalities can play a large role in diagnosing a multitude of diseases, including carcinomas of the colon, lungs, and prostate. The work presented in this dissertation develops varied visualization techniques for application to the detection and localization of these three prevalent cancers. All techniques presented herein were developed and tested using real patient data acquired from CT or MRI scanning modalities. There has been substantial research performed in the development of clinical virtual colonoscopy (VC) systems based on CT scans, but methods to meld multiple sources of diagnostic information are still under active development. Presented are beginning methods to merge VC with the traditional optical colonoscopy, such that VC information can be better utilized on a patient referred for the optical procedure. This work includes a method for removing the radial distortion introduced by the fisheye lens on the endoscope and a method for correlating the path of the physical colonoscope with the VC path. Techniques are also presented here which can enhance a traditional VC environment. For CT scans of the colon acquired with the patient in different orientations, a method is presented to register the two scans together in a one-to-one and onto manner. This direct mapping between the two surfaces allows for the corresponding visualization of the same location within each of the two colon models. Mesh models are often used as a map to show the user's location within an object. In the case of VC, this 3D model can often have occlusions due to the twisted shape of the colon. A method is presented to create a planar map which preserves the global shape of the colon and does not contain any overlapping sections. This allows for the user to observe the entire colon surface at once, ensuring that there are no occlusions which could lead to an ambiguity in determining one's location. Similar problems exist for exploring the lungs in virtual bronchoscopy applications, and can be even worse, due to the large number of bifurcations. The idea of a context preserving map is expanded from tubular structures to treelike structures, being applicable not only to the bronchi, but other branching structures such as blood vessels. As there are different aspects to be addressed compared to a simple tubular structure, this work is not simply an extension, but presents new methods which have been developed to deal with the challenges inherent in treelike structures. For the prostate, multiple MRI modes are typically used for the detection of cancer. Compared to colon and lung cancer, the prostate has so far seen relatively little in terms of visualization research, and work is presented here on how to combine the multiple MR modes to assist in the detection of prostate cancer. This includes an upsampling and analysis of the slices to identify regions of interest and the display of these regions within the prostate along with the surrounding anatomy using multi-volume rendering. Also developed is a method of visibility persistence to allow for easy viewing of occluded regions of interest, the ability of the user to paint custom regions into the data, and an extension of the visibility persistence to these user painted regions.
dcterms.available2013-05-22T17:35:14Z
dcterms.available2015-04-24T14:47:07Z
dcterms.creatorMarino, Joseph
dcterms.date2012
dcterms.dateAccepted2013-05-22T17:35:14Z
dcterms.dateAccepted2015-04-24T14:47:07Z
dcterms.dateSubmitted2013-05-22T17:35:14Z
dcterms.dateSubmitted2015-04-24T14:47:07Z
dcterms.descriptionDepartment of Computer Science
dcterms.descriptionAdvisor: Kaufman, Arie
dcterms.descriptionCommittee members: Mueller, Klaus; Gu, Xianfeng; Li, Wei
dcterms.descriptionDepartment of Computer Science
dcterms.extent145 pages
dcterms.formatapplication/pdf
dcterms.issued2012-12
dcterms.languageen
dcterms.provenanceMade available in DSpace on 2013-05-22T17:35:14Z (GMT). No. of bitstreams: 1 Marino_grad.sunysb_0771E_11178.pdf: 14227827 bytes, checksum: c1311a569f912efb1c725cc9389d8a7e (MD5) Previous issue date: 1
dcterms.provenanceMade available in DSpace on 2015-04-24T14:47:07Z (GMT). No. of bitstreams: 3 Marino_grad.sunysb_0771E_11178.pdf.jpg: 1894 bytes, checksum: a6009c46e6ec8251b348085684cba80d (MD5) Marino_grad.sunysb_0771E_11178.pdf.txt: 261772 bytes, checksum: 0a5fe5f31b7c6f6820d94c0d41f6b4e9 (MD5) Marino_grad.sunysb_0771E_11178.pdf: 14227827 bytes, checksum: c1311a569f912efb1c725cc9389d8a7e (MD5) Previous issue date: 1
dcterms.publisherStony Brook University
dcterms.subjectComputer science
dcterms.typeText


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record