CT-SGAN: Computed Tomography Synthesis GAN

Diversity in data is critical for the successful training of deep learning models. Leveraged by a recurrent generative adversarial network, we propose CT-SGAN model that generates large-scale 3D synthetic CT-scan volumes (\({\ge }224\times 224\times 224\)) when trained on small dataset chest CT-scans. offers an attractive solution to two major challenges facing machine medical imaging: number given i.i.d. data, and restrictions around sharing patient preventing rapidly obtain larger more diverse datasets. We evaluate fidelity generated images qualitatively quantitatively using various metrics including Fréchet Inception Distance Score. further show can significantly improve lung nodule detection accuracy pre-training classifier vast amount data.