Three-dimensional self-attention conditional GAN with spectral normalization for multimodal neuroimaging synthesis

To develop a new 3D generative adversarial network that is designed and optimized for the application of multimodal 3D neuroimaging synthesis.

We present a 3D conditional generative adversarial network (GAN) that uses spectral normalization and feature matching to stabilize the training process and ensure optimization convergence (called SC-GAN). A self-attention module was also added to model the relationships between widely separated image voxels. The performance of the network was evaluated on the data set from ADNI-3, in which the proposed network was used to predict PET images, fractional anisotropy, and mean diffusivity maps from multimodal MRI. Then, SC-GAN was applied on a multidimensional diffusion MRI experiment for superresolution application. Experiment results were evaluated by normalized RMS error, peak SNR, and structural similarity.

In general, SC-GAN outperformed other state-of-the-art GAN networks including 3D conditional GAN in all three tasks across all evaluation metrics. Prediction error of the SC-GAN was 18%, 24% and 29% lower compared to 2D conditional GAN for fractional anisotropy, PET and mean diffusivity tasks, respectively. The ablation experiment showed that the major contributors to the improved performance of SC-GAN are the adversarial learning and the self-attention module, followed by the spectral normalization module. In the superresolution multidimensional diffusion experiment, SC-GAN provided superior predication in comparison to 3D Unet and 3D conditional GAN.

In this work, an efficient end-to-end framework for multimodal 3D medical image synthesis (SC-GAN) is presented. The source code is also made available at https://github.com/Haoyulance/SC-GAN.