Breast lesion identification and categorization using mammography screening based on combined convolutional recursive neural network framework with parameters optimized using multi‐objective seagull optimization algorithm. (30th October 2022)

• Record Type: Journal Article
• Authors: Sakthivel, N. K.; Subasree, S.; Alias Priya M, Pachhaiammal; Tyagi, Amit Kumar
• Is Part Of: Concurrency and computation. Volume 34:Number 28(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-10-30
• Subjects: altered phase preserving dynamic range compression -- breast cancer -- convolutional neural networks -- multi‐objective seagull optimization algorithm -- recursive neural networks; Parallel processing (Electronic computers) -- Periodicals; Parallel computers -- Periodicals; 004.35
• DOI: 10.1002/cpe.7348
• Languages: English
• ISSNs: 1532-0626

Abstract

Summary: In recent years, a number of learning methods have been adopted for classifying the mammogram images, which helps the early detection and diagnosis of breast cancer. The breast lesion identification and categorization using mammography screening based on combined convolutional neural network and recursive neural network (CRNN) framework with parameters optimized using multi‐objective seagull optimization algorithm (BLIC‐CRNN‐MOSOA) is proposed in this article. Initially, the unnecessary noise components are taken away from the mammogram images and the quality of the images are enhanced based on altered phase preserving dynamic range compression filtering approach. Then, the deep CRNN model with weight parameters optimized using multi‐objective seagull optimization algorithm is adopted for classifying the mammogram images into three categories: (i) normal, (ii) benign, and (iii) malignant masses. The proposed BLIC‐CRNN‐MOSOA approach is executed in MATLAB platform, and its performance is compared with other deep learning classification approaches. Then the simulation performance of the proposed BLIC‐CRNN‐MOSOA method attains higher accuracy 99.67%, 98.38%, and 97.45%, higher sensitivity 98.33%, 89.34%, and 88.96%, higher specificity 93.15%, 91.25%, and 92.88% compared with existing methods, like BLIC‐FrCN, BLIC‐ICS‐ELM, and BLIC‐DCNN‐BO. By this, the proposed method achieves higher classification accuracy with less misclassified error. Finally, the simulation results show that the proposed method is more efficient than the other classification methods.