A novel deep learning framework designated as COBRC-MOS achieved a 98.35 percent classification accuracy in evaluating mammographic images for breast cancer, according to research published in the International Journal of Computational Intelligence Systems. Designed to overcome persistent challenges in computer-aided diagnosis, the proposed computational pipeline integrates automated deep feature extraction, optimization-driven feature selection, machine learning classification, and image segmentation to identify and differentiate benign and malignant breast cancer conditions from mammographic scans.
In establishing the research context, the authors highlighted that breast cancer represents one of the most common and deadly diseases, necessitating early identification and effective diagnosis to increase patient survival chances. However, traditional diagnostic approaches suffer from notable practical limitations, including elevated false-positive and false-negative rates. Additionally, conventional diagnostic procedures face limited accessibility within resource-constrained medical environments, underscoring the ongoing requirement for more reliable and accessible computational diagnostic solutions.
To develop and assess the COBRC-MOS framework, the investigators initially gathered input mammography images from the MIAS dataset. To prepare these clinical scans for downstream computational processing, the input images were preprocessed using contrast-limited Adaptive Histogram Equalization, widely known as CLAHE.
For the analytical pipeline, the deep learning model EfficientNet is used to perform feature extraction directly from the preprocessed mammograms. Rather than utilizing all extracted parameters indiscriminately, the framework implements the Crayfish Optimization Algorithm, designated as CFO, to select the most discriminative and optimal features. The optimization algorithm efficiently explores the complex search space to identify essential attributes, thereby enhancing model performance while simultaneously improving overall model interpretability.
Following this feature selection phase, a Support Vector Machine classifier, or SVM, is used to categorize the mammographic data into eight distinct breast cancer types. The eight targeted conditions encompass four malignant diagnoses—malignant papillary carcinoma, malignant mucinous carcinoma, malignant ductal carcinoma, and malignant lobular carcinoma—as well as four benign diagnoses, specifically benign tubular adenoma, benign phyllodes tumor, benign fibroadenoma, and benign adenosis.
Finally, the integrated framework incorporates U-Net segmentation to identify affected tissue regions accurately across the examined mammographic images. In comparative performance assessments, the proposed COBRC-MOS model demonstrated clear diagnostic advantages over alternative baseline architectures. The model improved total classification accuracy by 2.99 percent over BC2NetRF, by 5.79 percent over BC2-MultiRad, and by 5.37 percent over BCD-XGBoost, establishing documented performance gains across all three comparative models.
Subject of Research: Deep learning and crayfish optimization for mammographic breast cancer classification
Article Title: COBRC-MOS: Crayfish Optimization Algorithm for Breast Cancer Classification using Mammography Images
Article References: Thrishul, K., Priya, M. B., Vimaladevi, M., Suganthi, N. M., Abbas, S. H., Jaison, B., & Ahilan, A. (2026). COBRC-MOS: Crayfish Optimization Algorithm for Breast Cancer Classification using Mammography Images. International Journal of Computational Intelligence Systems. https://doi.org/10.1007/s44196-026-01532-4
Image Credits: AI Generated
DOI: 10.1007/s44196-026-01532-4
Keywords: Breast cancer classification, Mammography images, Crayfish Optimization Algorithm, EfficientNet, Support Vector Machine, U-Net segmentation, Contrast-limited Adaptive Histogram Equalization, MIAS dataset, Malignant papillary carcinoma, Malignant mucinous carcinoma, Malignant ductal carcinoma, Malignant lobular carcinoma, Benign tubular adenoma, Benign phyllodes tumor, Benign fibroadenoma, Benign adenosis, Feature extraction, Feature selection
News Source: Nathaniel Bowman. (November 19, 2026). Crayfish Optimization Model Classifies Mammography Breast Cancer with 98.35 Percent Accuracy. Scienmag.



