DLF: A Deep Learning Framework Using Convolution Neural Network Algorithm for Breast Cancer Detection and Classification

• Published in: Traitement du signal Vol. 41; no. 3; pp. 1101 - 1114
• Main Authors: Govindarajan, Kalpana, Narayanasamy, Deepa
• Format: Journal Article
• Language: English
• Published: Edmonton International Information and Engineering Technology Association (IIETA) 01.06.2024
• Subjects: Algorithms; Artificial neural networks; Breast cancer; Classification; Convolution; Datasets; Deep learning; Feature extraction; Feature recognition; Image coding; Image processing; Image quality; Machine learning; Medical imaging; Medical research; Mortality; Neural networks; Task complexity
• ISSN: 0765-0019; 1958-5608; 1958-5608
• DOI: 10.18280/ts.410302

Breast cancer is one of the most frequently affecting second types of cancer in men and women worldwide. Of the overall types of cancer, 25% of them are breast cancer in women. Erratic development of breast cells results in breast cancer. The growth of cancer increases the metastasizing of the tissues, spreads fast to the other parts of the body, and results in death. The medical industry requires an efficient algorithm to detect and classify the severity level of breast cancers with the metastasis of the affected tissues. Several earlier research works have focused on constructing a computer algorithm to diagnose breast cancer images to detect and classify cancer. The earlier algorithms involved more sub-functions or procedures in completing individual tasks separately, thus increasing the computational and time complexity. This paper introduces a Deep Learning Framework (DLF) to diagnose breast images automatically and speedily with less complexity. The proposed DLF includes a few image processing tasks to improve the quality of the input image and increase classification accuracy. Recently, Convolution Neural Network has been used as an extraordinary class of models for image recognition processes. CNN is one of the deep learning models that can extract the entire set of image features and use them for analysis and classification. Thus, this paper implements a deep CNN for diagnosing and classifying benign and malignant cancers from input datasets with Python coding—the deep form of the CNN obtained by increasing the number of hidden layers and epochs. The experiment proves that CNN is highly reliable compared to the existing algorithm.