Developing advanced techniques for denoising Low-dose CT Images

Computed Tomography (CT) is one of the most widespread non-invasive imaging modalities in medical diagnostics. Recent concerns regarding radiation induced cancer, has drawn a lot of attention to reduce the radiation dose used during CT scanning. However, the signal to noise ratio of scans taken at lower radiation dose is considerably lower than at higher dosages, resulting in poorer diagnostic accuracy. Hence post processing of low-dose scans has become a major concern in medical image processing. In this project two denoising approaches based on sparse representations will be proposed to address the problem of Low-dose CT image denoising. In the first approach, we will use an enhanced version of analytical Discrete Cosine Transform dictionary which leads to more efficient representation of input images. Furthermore, to speed up the process of finding the sparsest representation of an image, a new efficient sparse coding method will be introduced. The second approach is based on adaptive dictionaries. Here, we will propose a novel approach called Adjustable length K-SVD to learn a dictionary with sufficient number of atoms. Finally, it is anticipated that the proposed techniques could be used to reduce the radiation dose needed on CT to acquire the images in clinical environments which has benefits for patients especially for pediatrics.

Faculty Supervisor:

Dr. Javad Alirezaie

Student:

Azar Tolouee & Samira Ghadrdan

Partner:

Dr. Paul Babyn Professional Medical Corporation

Discipline:

Engineering - computer / electrical

Sector:

Life sciences

University:

Ryerson University