A new automated segmentation and classification of mammogram images

• Published in: Multimedia tools and applications 2022-03, Vol.81 (6), p.7783-7816
• Main Authors: Patil, Rajeshwari S., Biradar, Nagashettappa, Pawar, Rashmi
• Format: Article
• Language: English
• Subjects: Algorithms; Breast cancer; Classification; Clustering; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Deep learning; Feature extraction; Image classification; Image enhancement; Image segmentation; Machine learning; Mammography; Medical imaging; Model accuracy; Multimedia Information Systems; Object recognition; Optimization; Recurrent neural networks; Special Purpose and Application-Based Systems; Tumors
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-022-11932-1

Early-stage recognition of lesions is the better probable manner for fighting against breast cancer to find a disease with the highest ratio of malignancy around women. Existing approaches are generally based on deep learning that has been designed for the segmentation of tumors, however, it is complex because of the false positives and the inaccurate detection of boundaries for segmentation, as the existing models incorrectly predict the positive classes, thus affecting the overall classification. In this paper, an enhanced mammogram image classification is proposed by introducing novel segmentation and classification approaches. The initial process of the proposed model is pre-processing, which is performed by the median filtering that tends to remove the noise from the images. The preprocessed images are subjected to segment the tumor from the mammogram images by a new segmentation approach termed Region growing with Adaptive Fuzzy C-Means Clustering (RG-AFCM). Once the segmentation of the tumor is done, feature extraction is performed, where the features are extracted using Gray-Level Run-Length Matrix (GLRM) and Grey Level Co-occurrence Matrix (GLCM) approaches. Furthermore, the extracted features are classified using optimal trained Recurrent Neural Networks (RNN). Here, a new algorithm named Average Fitness New Updating-based Grasshopper Optimization Algorithm (AFU-GOA) is proposed for enhancing both the segmentation and classification phases. Finally, the performance of RG-AFCM-based segmentation is compared over the stat-of-the-art segmentation approaches, and optimal trained RNN is compared over the existing classifiers and deep learning models to prove the reliability of the proposed model. The accuracy of the developed AFU-GOA-RNN is 1%, 2%, 1%, and 3% enhanced than PSO-RNN, GWO-RNN, FF-RNN, and GOA-RNN. Hence, the proposed classification using AFU-GOA-based trained RNN establishes a better performance than existing models.