Breast image registration using a textural transformation

Abstract

(c)Permission h u betn granted to the Library of The Univenity of Manitoba to leud o r se11 copies of this thesidpracticum, to the National Libmry of Canada to micronlm thu thesis and to lend or sel1 copies of the film, and to Dissertations Abstracts International to publisb an abstract of t h h thesWpnicticum.The author reserves other publication rights, and neither t h b thesis/prncticum nor extensive estmcts from it may be p n n t d o r otherwise reproduced without the author's d t t e n permission. AbstractBreast cancer codd be cured 99.4% if ail tumors are detected by the time they.a reach 0.4 cm. However, since many normal structures in the breast will be of this magnitude, to detect small tumors we need to aiign two breast images taken some time apart and subtract them. Hence image registration is important for breast imaging.Image registratbn consiats in identifjbg correspondhg points in two images and interpolating between them. The low amttast and noise in bresst images permit few contrd points. to 8, for a 9 x 9 neighborhood. i -p: Starbyte trsnsformed images of Figiire 2.2b, uing protocols 1 to 8, for a 9 x 9 neighborhood. . . . . . 28 2.7 a -h: Starbyte transfonned images of Figure 2.k, using protocols 1 to 8, for a 9 x 9 neighborhood. i -p: Starbytr t r a n s f o d images of .24b, using profocois 1 to 8, for a 9 x 9 neighborhood. . . . . a -h:Starbyte transformed images of Figure 2.24c, using protocols 1 to 8, for a 9 x 9 neighborhood i -p: Starbyte transformed images of Figure 2.24d, using protocois 1 to 8, for a 9 x 9 neighborhood. . . . . 2.27a -h: Starbyte t r a n s f o 4 images of Figure 2.24e, using protocoIs 1 to 8, for a 9 x 9 neighborhood i -p: Starbyte transfonned images of Figure 2.24f, using protocols 1 to 8, for a 9 x 9 neighborhood. . . . . 3.4 a to p: Density slices in Figure 3.2 after OC operation with a "plus" SE. 72 3.5 a to p: Density slices in Figure 3.3 after OC operation with a '$lus" SE. 73 a -h:3.6 a to p: Density slioes in Figure 3.2 after OC operation with a 3 x 3 SE. 74 3.7 a to p: Density slioes in Figure 3.3 a h r OC operaion with a 3 x 3 SE. . . . . . . . .Registration achieved after using UICP with protocol 1. 138. . . . (2.2) 3. geometric transformation of the referenoe image by means of the mapping function to obtain the unwarped image with gray ievel resampling. Control point selectionThe problem of control point selection has been extensively studied in the literature. Intrinsic control points are markers which are pl& on the object king irnaged just for the purpose of registration and can be easily identified in both im- (Figure 2. la). Each sector is filled in with black or white, depending on whether the average value in the sector is below or above the picture density at the center of this neighborhood (or the average density over the whole neighborhood).Some such patterns will look like "stars", with bladr taking the value 0, and white taking the d u e 1. The bits assigned to the bladr and white sectors, when concatenated into a...

Breast image registration techniques: a survey

“…In practice, the extraction of the anatomical features, both at the surface [40,70,100,115,150] and internal [25,74,81,113,131,137,139], helps to locate corresponding control points in both images that can be used for registration. Still other features, like texton (a basic texture element) [56,93] and texture map [119], can also be found in the literature. On the other hand, intensitybased registration methods are getting more attention [31,106,147].…”

“…Petroudi et al [93] constructed a texton directory to help the location of control points. Sivaramakrishna et al [119] proposed a textual transformation to register mammograms non-rigidly. First, the textual transformation is used to convert mammograms into texture maps where control points are selected.…”

Hybrid Registration of Corresponding Mammogram Images for Automatic Detection of Breast Cancer

“…Although much work has been done on mammogram registration using either feature-based methods or intensity-based method, [15][16][17][18] only a few studies have been done on combining the aforementioned two techniques. In this research, a hybrid mammogram registration algorithm, which used both methods to bring mammograms into alignment, has been developed.…”

A New Method for the Registration of Mammograms

“…As a result, many non-rigid-based mammogram registration techniques were proposed [5,6]. Although much work has been done on mammogram registration using either feature-based methods [3,4] or intensity-based method [7,8], few studies have been done on combining the aforementioned two techniques. In this research, a hybrid mammogram registration algorithm, which used both methods to bring mammograms into alignment, has been developed.…”