BREAST CANCER DIAGNOSIS ON PATHOLOGICAL IMAGE DATA AUGMENTATION METHOD:CYCLE GAN

Abstract

Breast cancer is one of the most common cancers affecting women worldwide, and early and accurate diagnosis is essential for improving patient survival. Histopathological image analysis plays an important role in breast cancer diagnosis, but the development of deep learning models is often limited by insufficient, imbalanced, and diverse pathological image datasets. This project, “Breast Cancer Diagnosis on Pathological Images Using Data Augmentation Method: CycleGAN,” proposes the use of Cycle-Consistent Generative Adversarial Networks (CycleGAN) to generate realistic synthetic pathological images for effective data augmentation. CycleGAN can learn the visual characteristics of breast tissue images and generate additional samples without requiring paired images. The generated images are combined with the original dataset to increase the diversity and quantity of training data. A deep learning-based classification model can then be trained on the augmented dataset to distinguish between benign and malignant breast tissue patterns. The proposed approach aims to reduce overfitting, improve model generalization, and enhance diagnostic classification performance. By incorporating synthetic pathological images into the training process, the system can provide a more robust and efficient computer-aided diagnosis approach. This methodology demonstrates the potential of generative deep learning techniques for addressing limited medical imaging datasets and supporting automated breast cancer diagnosis.