Study On Medical Image Reconstruction Algorithm Based On Adaptive Dictionary Learning

Medical imaging technology, as a non-intrusive way medical diagnosis reference, is playing a more and more important role in modern medicine. But the mainstream of medical imaging system exist some deficiencies, for example: The scan time of MRI is long and the image is easily influenced by motion artifacts; CT have a certain amount of radiation damage for patients and so on. How to reduce the amount of data collection as well as to achieve high quality of image reconstruction is the main research in the field of the current medical imaging, this paper revolves around two kinds of dictionary learning models in compression perception theory and the adaptive medical image reconstruction algorithm, the specific contents are as follows:Firstly, based on adaptive orthogonal dictionary learning propose the adaptive medical image reconstruction algorithm, according to the advantage of orthogonal dictionary which atoms are no correlation that can improve the performance of image sparse coding. In the algorithm, the process of orthogonal dictionary learning achieves by SVD method, which greatly reduces the computation complexity as well as improves the reconstruction speed, and at the same time achieves the better quality of image reconstruction.Secondly, considering the condition that under the same dimension analysis dictionary is better than synthesis dictionary for more flexible and stronger image sparse representation ability, taking advantage of tight frame learning, proposes the medical image reconstruction algorithm base on adaptive tight frame learning. The algorithm can only use undersampling measurements to achieve tight frame learning and image reconstruction. In sparse coding steps, tight frame learning uses simple threshold method for sparse coding. So the algorithm running time is less than the CSMRI algorithm based on redundancy comprehensive dictionary. In addition, adaptive tight frame can fully capture the details of the image information, the image reconstruction quality based on adaptive tight frame learning medical image reconstruction algorithm gets obviously improved than other previous algorithms.Finally, considering that the MRI images is easy effected by motion artifacts in the process of image data scan, by analyzing the principle of motion artifacts produced, based on adaptive tight frame learning proposes the phase correction of medical reconstruction algorithm. The algorithm can achieve the better correction motion artifact, and has strong robustness to the additive noise and phase noise.