Magnetic resonance imaging quality classification based on deep machine-learning to account for less training data

A method for training of classification of image quality, the method comprising:

• simulating artifacts in input images, the simulating resulting in artifact images from the input images;
• deep learning, by a machine, to predict a degree of similarity of the artifact images to the respective input images;
• machine training, by the machine, a classifier from training data with ground truth annotations of the image quality and input features learned from the deep learning;
• and
• storing the machine-trained classifier and a machine-learned network from the deep learning.

A system for magnetic resonance image quality assessment, the system comprising:

• a magnetic resonance scanner configured to generate data representing a patient;
• an image processor configured to generate values of features of the data with a machine-learned network of deep learned features learned without quality annotation, to determine the image quality with a machine-learned classifier based on the values;
• and
• a display configured to display an indication of the image quality.