1. Deep Learning RN-BCNN Model for Breast Cancer BI-RADS Classification
- Author
-
Jiyun Li, Arslan Manzoor, Shahbaz Siddeeq, Umar Subhan Malhi, and Hafiz Muhammad Ali Bhatti
- Subjects
medicine.diagnostic_test ,Artificial neural network ,Computer science ,business.industry ,Deep learning ,05 social sciences ,BI-RADS ,010501 environmental sciences ,Machine learning ,computer.software_genre ,medicine.disease ,01 natural sciences ,Convolutional neural network ,Breast cancer ,0502 economics and business ,Medical imaging ,medicine ,Mammography ,Pyramid (image processing) ,Artificial intelligence ,050207 economics ,business ,computer ,0105 earth and related environmental sciences - Abstract
The most efficient and ordinarily used early detection method of breast cancer is screening mammography and deep learning is widely employed in the medical imaging domain. But In the medical circumstances, large data size for training is a significant difficulty and the second thing is very less work on six levels of breast cancer BI-RADS classification. In this research, for the purpose of six levels of BI-RADS classification, we proposed the methodology with a ResNet-based customized neural network (RN-BCNN) as compared to the traditional ConvNet model, using the data augmentation and pyramid of scales techniques on an imbalanced dataset. We obtained the outcomes from the INbreast dataset of mammograms. Therefore, we used the elastic deformation technique for increasing the training dataset size which aid in the improvement of outcomes. Moreover, the proposed methodology improves the accuracy of up to 85.9% because the customized model and elastic deformation took important roles in the efficiency of the proposed strategy.
- Published
- 2021