Classification of MR Images pertaining to with artifact / without artifact using Deep Autoencoder Convolutional Neural Network

Today's radiologists are highly concerned with motion artifact available in MR images and expect MR images during diagnosis without motion artifact. Radiologists receive merged data of MR images with and without motion at the time for patient diagnosis. The radiologist may be mislead by these images of motion artifacts. We suggest a classification model in this paper that makes use of a deep autoencoder convolutional neural network (DAE-CNN) and has three levels as (i) input level (ii) hidden level and (iii) output level. We assessed the classification model and the results are analyzed regarding the accuracy in terms of motion/without motion MR Images.