Fully Convolutional Architecture for Low-Dose CT Image Noise Reduction

2017 International Conference on Artificial Intelligence Applications and Technologies

计算机科学

Badretale, S.^1 ; Shaker, F.^2 ; Babyn, P.^3 ; Alirezaie, J.^1

Department of Electrical and Computer Engineering, Ryerson University, Toronto

ON

M5B 2K3, Canada^1

Department of AI, Faculty of Computer Engineering, University of Isfahan, Isfahan

81746, Iran^2

Department of Medical Imaging, University of Saskatoon, Saskatoon

SK

S7N 0W8, Canada^3

关键词: Deep convolutional neural networks;    High-level features;    Linear units;    Peak signal to noise ratio;    Root mean square errors;    State-of-the-art methods;    Structural similarity indices;    X ray radiation;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/261/1/012012/pdf
DOI  :  10.1088/1757-899X/261/1/012012

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

One of the critical topics in medical low-dose Computed Tomography (CT) imaging is how best to maintain image quality. As the quality of images decreases with lowering the X-ray radiation dose, improving image quality is extremely important and challenging. We have proposed a novel approach to denoise low-dose CT images. Our algorithm learns directly from an end-to-end mapping from the low-dose Computed Tomography images for denoising the normal-dose CT images. Our method is based on a deep convolutional neural network with rectified linear units. By learning various low-level to high-level features from a low-dose image the proposed algorithm is capable of creating a high-quality denoised image. We demonstrate the superiority of our technique by comparing the results with two other state-of-the-art methods in terms of the peak signal to noise ratio, root mean square error, and a structural similarity index.