Dictionary Learning based Low-dose X-ray Computed Tomography
主 题: Dictionary Learning based Low-dose X-ray Computed Tomography
报告人: Prof. Dr. Hengyong Yu （Department of Radiology VT-WFU School of Biomedical Engineering and Sciences Wake Forest University Health Sciences）
时 间: 2012-11- 21 9:30-11.00
地 点: 理科一号楼1114
Worldwide there are growing concerns on radiation induced genetic, cancerous and other diseases. Facing the increasing radiation risk, how to reduce radiation dose while maintaining the diagnostic performance is a major challenge in the computed tomography (CT) field. Inspired by the state-of-the-art compressive sensing (CS) theory, very recently we proposed a total variation (TV) minimization based statistical reconstruction method for low-dose x-ray CT, which provided very promising results. Noticing that dictionary learning and sparse representation have been successfully applied in the image processing field, in this talk we will present an improved statistical reconstruction method incorporating the dictionary learning and sparse representation techniques. In this method, a sparse constraint on a redundant dictionary is firstly incorporated into a novel object function in a statistical iterative reconstruction (SIR) framework. And the dictionary can be pre-determined before image reconstruction or adaptively updated during the reconstruction process. Then, an alternating minimization algorithm is developed to minimize the objective function. Finally, our method is validated by low-dose x-ray projections from both animal and patient CT studies and evaluated quantitatively by simulation study. Our results show that the proposed method can produce better image quality than the popular TV based method in terms of suppressing noise and preserve structure information.