Image standardization using generative adversarial networks

1. A method for synthesizing protocol independent magnetic resonance images in a magnetic resonance imaging system, the method comprising:

scanning a patient by the magnetic resonance imaging system to acquire magnetic resonance data;

inputting the magnetic resonance data to a machine learnt generator network trained to extract features from input magnetic resonance data and synthesize protocol independent images using the extracted features;

generating, by the machine learnt generator network, a protocol independent segmented magnetic resonance image from the input magnetic resonance data; and

displaying the protocol independent magnetic resonance image

wherein the machine learnt generator network comprises an independent encoder configured to generate a latent space for the input magnetic resonance data and an independent decoder configured to synthesize the protocol independent image from the modality latent space,

wherein the independent encoder is trained using an adversarial network

wherein the independent encoder comprises a dense fully convolutional network, and

wherein the independent encoder is trained by the adversarial network using concatenated features from each dense block output of the dense fully convolutional network, wherein the adversarial network is trained to correctly classify which protocol training data is from.

2. The method of claim 1 , wherein the independent encoder is trained on magnetic resonance data from acquired using a single protocol from a plurality of institutions.

3. The method of claim 1 wherein the independent decoder is trained to synthesize an image to the same source modality as the magnetic resonance data.

4. The method of claim 1 , wherein the adversarial network is trained using a loss function calculated as a function of a Wasserstein distance.

5. The method of claim 1 , wherein the adversarial network is trained using a loss function calculated as a function of a Wasserstein distance.