Segmentation of breast lesion in DCE-MRI by multi-level thresholding using sine cosine algorithm with quasi opposition-based learning

• Published in: Pattern analysis and applications : PAA 2023-02, Vol.26 (1), p.201-216
• Main Authors: Si, Tapas, Patra, Dipak Kumar, Mondal, Sukumar, Mukherjee, Prakash
• Format: Article
• Language: English
• Subjects: Algorithms; Breast cancer; Computer Science; Decision analysis; Feature extraction; Fields (mathematics); Image segmentation; Lesions; Machine learning; Magnetic resonance imaging; Medical imaging; Multiple criterion; Optimization; Pattern Recognition; Swarm intelligence; Theoretical Advances; Trigonometric functions
• ISSN: 1433-7541; 1433-755X
• DOI: 10.1007/s10044-022-01099-8

In recent times, the high prevalence of breast cancer in women has increased significantly. Breast cancer diagnosis and detection employing computerized algorithms for feature extraction and segmentation can be aided by a physician’s expertise in the field. To separate breast lesions from other tissue types in Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) for segmentation and lesion detection in breast DCE-MRI, radiologists think that multi-level thresholding optimization is efficient. In this article, a lesion segmentation method for breast DCE-MRI using the opposition-based Sine Cosine Algorithm (SCA) is proposed. For breast DCE-MRI segmentation utilizing multilevel thresholding, this work provides an upgraded version of the SCA with Quasi Opposition-based Learning (QOBL). SCAQOBL is the name given to the suggested method in this paper. The Anisotropic Diffusion Filter (ADF) is used to de-noise MR images, and subsequently, Intensity Inhomogeneities (IIHs) are corrected in the preprocessing stage. The lesions are then retrieved from the segmented images and located in MR images. On 100 sagittal T2-weighted fat-suppressed DCE-MRI images, the proposed approach is examined. The proposed method is compared to Opposition-based SCA (OBSCA), SCA, Particle Swarm Optimizer (PSO), Slime Mould Algorithm (SMA), Hidden Markov Random Field (HMRF), and Improved Markov Random Field (IMRF) algorithms. The proposed technique achieves a high accuracy of 99.11 percent, sensitivity of 97.78 percent, and Dice Similarity Coefficient (DSC) of 95.42 percent. The analysis of results is conducted using a one-way ANOVA test followed by a Tukey-HSD test, and Multi-Criteria Decision Analysis (MCDA). The proposed strategy surpasses other examined methods in both quantitative and qualitative findings.