The study explores using generative models like CycleGAN and Neural Style Transfer to generate synthetic cytology images from publicly available breast histopathology samples. The scarcity of annotated datasets in medical imaging poses a challenge, prompting the use of generative models for data augmentation. The research focuses on unpaired image-to-image translation, specifically shifting from breast histopathology to cytology images. By measuring FID and KID scores, the study confirms that the generated cytology images closely resemble real samples. Various techniques and models are discussed, highlighting the importance of realistic synthetic data generation for improved classification performance.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Soumyajyoti ... at arxiv.org 03-19-2024
https://arxiv.org/pdf/2403.10885.pdfDeeper Inquiries