Bioinformatics-Inspired Algorithms for 2D-Image Analysis­­—Application to Medical Images Part II: Images in Circular Format

Bioinformatics-Inspired Algorithms for 2D-Image Analysis­­—Application to Medical Images Part II: Images in Circular Format

Perambur S. Neelakanta (Department of Computer and Electrical Engineering & Computer Science, Florida Atlantic University, Boca Raton, FL, USA), Edward M. Bertot (Department of Computer and Electrical Engineering & Computer Science, Florida Atlantic University, Boca Raton, FL, USA) and Deepti Pappusetty (Department of Computer and Electrical Engineering & Computer Science, Florida Atlantic University, Boca Raton, FL, USA)
Copyright: © 2012 |Pages: 10
DOI: 10.4018/ijbce.2012010104

Abstract

This paper describes a new method of comparing images of circular/near-circular symmetry so as to elucidate the similarity details between them. If one such image is a test-entity and the other is a reference template, the comparison in question will lead to find the unique features (and their locations) in the test-image vis-à-vis the template. The method of comparison and similarity assessment indicated thereof is to use the so-called Needleman-Wunsch (NW) and Smith-Waterman (SW) algorithms commonly adopted in bioinformatic contexts of comparing two linear sequences (like DNA chains). Relevant procedure is extended in this study to address 2D-patterns. It involves first transforming the test-image (of circular symmetry) from polar-plane to a rectangular format. Next, the transformed test-image is digitised and compared against a template (also in digital rectangular format) on row-to-row and column-to-column basis. The resulting alignment of pixel bits in the test-image versus the template leads to an optimal score-of-similarity on the comparisons made. Biomedical applications of the proposed strategy are explored with reference to typical and circular/quasi-circular MRI images, and the associated image recognition, interpretation, and locating of the artefacts are discussed.
Article Preview

Application Of Nw/Sw Techniques To 2D-Images: An Overview

The method of logically extending traditional 1D-sequence analysis (via NW/SW algorithms) to compare a pair of 2D-images can be accomplished as follows: First, the test-image (specified in rectangular format) is discretized into a set of rows and a set of columns; and, the image pixel-values across the resulting matrix (of discretized patterns) are designated as ‘0s” and ‘1s” based on, gray-scale levels of pixel intensity. Thus, the 2D-patterns of images being compared would eventually be in digitised (binary) matrix formats.

Next, a row-to-row and a column-to-column comparison between digitised patterns is performed by considering each row or a column (of bits) as a linear sequence. Therefore, alignment and scoring between row-to-row or column-to-column basis are exercised using either the NW or the SW algorithm; hence, the relative similarity between the patterns is scored in terms of the net result due to the prevailing matches (yielding positive score-values), mismatches and indels (of negative score values). Such estimated scores of binary correlation at the pixels depict implicitly the characteristic (relative) feature of the image at those locales. That is, with the pixel-by-pixel scored details obtained on image-features, the local artefacts that may prevail can be located and the overall image features can be displayed globally.

The methodology as above can be advocated in medical-image applications, where anomalies (like tumours) in an image can be revealed by comparing the test-image vis-à-vis a template (reference) image, both formatted in binary matrix as described above. However, the test-image as well as the template should be specified first in a matrix of rectangular symmetry for analysis via NW/SW algorithms. That is, the proposed methodology is compatible only when the images are specified in rectangular formats. However, in several medical contexts there are possibilities that the test-images in question may have circular symmetry as for example, in breast or retinal scans. In such cases of circular and/or quasi-circular images, an initial transformation of circular-to-rectangular frame of matrix of pixels is required. This would then allow a corresponding image similarity evaluation via the said NW/SW algorithms. The overall considerations as above are illustrated in Figure 1 by means of a flowchart.

Figure 1.

Flowchart to implement binary correlation via NW/SW algorithms pertinent to 2D-images of circular symmetry

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 8: 2 Issues (2019): 1 Released, 1 Forthcoming
Volume 7: 2 Issues (2018)
Volume 6: 2 Issues (2017)
Volume 5: 2 Issues (2016)
Volume 4: 2 Issues (2015)
Volume 3: 2 Issues (2014)
Volume 2: 2 Issues (2013)
Volume 1: 2 Issues (2012)
View Complete Journal Contents Listing