Ascertaining Abnormal Regions in Mammogram Images Using Gravitational Search Local Map View Technique
M.P. Sukassini1, T. Velmurugan2

1M.P. Sukassini, Research Scholar, Bharathiar University, Coimbatore, India.
2T. Velmurugan, Associate Professor, PG and Research Department of Computer Science, D. G. Vaishnav College, Arumbakkam, Chennai-600106, India.

Manuscript received on 28 June 2019 | Revised Manuscript received on 05 July 2019 | Manuscript published on 30 July 2019 | PP: 1861-1868 | Volume-8 Issue-9, July 2019 | Retrieval Number: I8416078919/19©BEIESP | DOI: 10.35940/ijitee.I8416.078919

Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Segmentation of mammogram images has gained importance in many medical treatments and diagnostic processes. Mammogram image segmentation aims at correctly separating different tissues, organs, or pathologies in volumetric image data. Most of the existing algorithms for image segmentation have a “scattered” cluster problem (disconnected clusters) happened in many clustering techniques (agglomerative, k-means, Dbscan) . Above algorithms not taken into account of both quality value and connectivity of points and region varying shapes. Two methods are proposed in this paper. The first technique is LMV (Local Map View) and the second technique is GSLMV (Gravitational Search Local Map View). LMV concentrates on determination of local quality for each point in all instances of the region in the comparative similarity view by applying the initial cluster technique. This view allows the user to choose instances for detailed analysis and filter the outlier instances from the input, next specific feature selection process identifies regions with systematic characteristics across the images. In this research work, pixels in groups with high intensity are assumed to be abnormal regions in cancer and non-cancer images. Fuzzy clustering is used to cluster the pixels. The optimal threshold from GS are initialized as cluster centres. This increases the speed of GSLMV algorithm. Performances of GSA, LMV and GSLMV methods are measured using False Rejection Rate, pixel count, Peak signal to noise ratio and runtime metrics. GSLMV showed better results based on pixel count, PSNR and runtime.
Index Terms: Mammogram Image, Segmentation, Local Map View, Peak to Signal Noise Ratio, Gravitational Search Algorithm.

Scope of the Article: Local-area and Metropolitan-Area Networks