ROI Selection Criteria for Finding the Abnormal Tissues from Breast Magnetic Resonance Imaging
Poonam Jaglan1, Rajeshwar Dass2, Manoj Duhan3
1Poonam Jaglan*, Pursuing PhD, Department of Electronics & Communication Engineering, DCRUST, Murthal, Sonepat, India.
2Rajeshwar Dass, Assistant Professor, Department of Electronics & Communication Engineering, DCRUST, Murthal, Sonepat, India.
3Manoj Duhan, Professor, Department of Electronics & Communication Engineering, DCRUST, Murthal, Sonepat, India.
Manuscript received on November 22, 2019. | Revised Manuscript received on December 15, 2019. | Manuscript published on December 30, 2019. | PP: 2744-2749 | Volume-9 Issue-2, December, 2019. | Retrieval Number: A2262109119/2019©BEIESP | DOI: 10.35940/ijeat.A2262.129219
Open Access | Ethics and Policies | Cite | Mendeley
© 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: The imaging methods in breast diagnostics play a preeminent role in the early detection and finding out the exact location & area of the suspicious breast tissues for malignancy. The further treatment significantly depends on the tumour-tobreast size relationship. The tumor size considered as the most influential factors for pathological/clinical assessment of breast cancer. In general, localization of the tumor’s location and also the selection of a region of interest (ROI) were performed manually by an experienced radiologist. The objective of this paper is to propose an effective criterion for selection of ROI for abnormal tissues detection from breast MRI. This paper implements an efficacious ROI selection criterion for finding the exact location & area of the breast abnormal tissues from magnetic resonance imaging automatically. The proposed algorithm integrates the simple techniques like filtering, edge detection and morphological operations for inner segmentation. Outer breast region segmentation is performed by selecting the peak and valley points and then connects the selected points by applying fit to circle function which makes the MR image rotation invariant. The method is implemented on the 80 images contained in S1 dataset i.e. multi-parametric breast MRI dataset and the evaluation is done through comparative analysis of predicted image with manually segmented images. The experimental results in terms of evaluation matrices i.e. Precision, Recall and Score depict the efficacy of the proposed work.
Keywords: Breast MRI, Image segmentation, Region of interest.