Processing of digital mammogram images using optimized ELM with deep transfer learning for breast cancer diagnosis

• Published in: Multimedia tools and applications Vol. 82; no. 30; pp. 47585 - 47609
• Main Authors: Chakravarthy, S. R. Sannasi, Bharanidharan, N., Rajaguru, Harikumar
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.12.2023; Springer Nature B.V
• Subjects: Artificial neural networks; Breast cancer; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Datasets; Diagnosis; Digital imaging; Flight; Image analysis; Iterative methods; Lung cancer; Machine learning; Mammography; Medical imaging; Multimedia Information Systems; Parameters; Particle swarm optimization; Performance enhancement; Search algorithms; Special Purpose and Application-Based Systems; Support vector machines; Track 2: Medical Applications of Multimedia; Crow-Search; Breast cancer; Transfer-Learning; Extreme Learning Machine; Mammograms; Median filtering
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-023-15265-5

The mortality of breast cancer is more among women besides lung cancer. However, the survival rates of breast cancer can be increased when there is a promising computer-aided diagnosis tool available for earlier detection and timely diagnosis. To tackle this, several research works are emerging with different methodologies but still accuracy and robustness are the key issues. Hence, a robust framework that incorporates the concept of Extreme Learning Machine (ELM) and Deep Transfer Learning is proposed and the performance of ELM is improved using an Iterative Flight-Length-Based Crow-Search Algorithm (iFLCSA) in this research work. Performance of ELM heavily depends on its parameters and to provide enhanced performance, the optimum parameters of ELM are found through the iFLCSA. When compared to the existing Crow Search Algorithm(CSA), the flight length parameter will be updated iteratively using an appropriate equation in iFLCSA to provide better balance between exploration and exploitation. Digital & full-field digital mammograms from the Mammographic Image Analysis Society (MIAS) and INbreast datasets are used for evaluation. The results obtained are then compared with the existing Support Vector Machine, ELM, Particle Swarm Optimization and CSA optimized ELM algorithms. The proposed iFLCSA-ELM provides a maximum classification accuracy of 98.292% and 98.171% for MIAS & INbreast datasets respectively.