An Efficient Image Reconstruction Method for Low-Dose Computed Tomography

Abstract

The Computed Tomography (CT) has made a tremendous growth in medical imaging field for for the different applications like: clinical diagnosis, detection, and interventions over the past few decades. Recent researches in medical science have signi cantly improvised the perception of scientist and Physician's to better understand the cause of the disease that a ffects the human system and how the treatment process works behind it. However, excessive use of X-ray radiation may have some harmful e ffects such as genetic disorder, cancer syndrome and other harmful diseases. So, minimizing the radiations from CT has been a major concern. Therefore, low-dose CT is gaining the interest of researchers. There are many methods proposed over the past few decades, still many problems are faced. In this work, we have presented two new statistical image reconstruction algorithm by proposing a suitable regularisation method. The proposed framework is the combination of two basic terms namely data fidelity and regularisation. Maximising the log likelihood gives the data fi delity term, which represents the distribution of noise in low-dose CT images. Maximum likelihood expectation maximization algorithm (MLEM) is introduced as a data- fidelity term in both the frameworks. In the former framework, Complex Di usion (CD) is introduced as a regularisation term into the proposed framework that minimizes the noise without blurring edges. Whereas, in the latter strategy a new hybrid regularisation mixture of Complex Di ffusion and Shock Filter is used as regularization term. The complex diff usion filter helps in denoising the image and the shock lter helps to retain the basic fi ne structures and deblurring of the edges. Both the frameworks have been evaluated on both simulated and real standard thorax phantoms. The final results are compared with the other standard methods and it is analyzed that the proposed frameworks have many desirable properties such as better noise robustness, less computational cost, enhanced denoising e ffect.