Deep neural network for CT metal artifact reduction

An apparatus for computed tomography (CT) metal artifact reduction (MAR), the apparatus comprising:

• a projection completion circuitry configured to generate an intermediate CT image data based, at least in part, on input CT projection data, the intermediate CT image data configured to include fewer artifacts than an uncorrected CT image reconstructed from the input CT projection data, wherein generating the intermediate CT image data comprises interpolation and normalization with prior CT projection data;
• detail image circuitry configured to generate detail CT image data based, at least in part, on input CT image data;
• and
• an artificial neural network (ANN) configured to generate CT output image data based, at least in part, on the intermediate CT image data and the detail CT image data, the CT output image data configured to include fewer artifacts compared to the intermediate CT image data, the ANN comprising an intermediate CT image data stream, a detail CT image data stream, and a CT output image data stream, the intermediate CT image data stream and the detail CT image data stream each comprising an initial convolution stage and a sequence of residual units, the CT output image data stream comprising a sequence of convolution stages and an output stage, each convolution stage comprising a convolution layer, a batch normalization layer, and a rectified linear unit, and the output stage comprising a convolution layer and a rectified linear unit.