Improved U-Net deep learning using clustering based on firefly and particle swarm optimization in breast cancer detection and classification in mammography images

Document Type : Original Article

Authors

1 Semnan Branch, Islamic Azad University, Semnan, Iran

2 Department of Electrical and Computer Engineering, Semnan Branch, Islamic Azad University, Semnan, Iran.

10.22034/jasp.2023.54109.1209

Abstract

Breast cancer is one of the important diseases of today's age, which correct and early diagnosis can lead to the process of improving the disease and preventing mortality. Deep learning in medical image processing is an effective solution to help doctors diagnose this cancer and tumor conditions, because the high density of breast tissue is one of the limitations in diagnosing this disease in mammography images. In this research, a new method using the U-Net deep network in combination with fuzzy clustering and the hybrid algorithms of firefly and particle swarm optimization is presented to recognize and segment the breast cancer tumor. The proposed method was analyzed on the real dataset of valid and standard DDSM and INbreast mammography images. The results of convolutional neural network method with 97.5% accuracy, U-Net method with 99.5% accuracy, improved method with particle swarm optimization PSO-FCM-U-Net with 99.6% accuracy and the proposed FFPSO-FCM-U-Net method with 99.8% accuracy have been accompanied. In general, the results of breast cancer detection accuracy with the proposed improved U-Net model yielded promising results.

Keywords

Main Subjects