Structurally-Sensitive Multi-Scale Deep Neural Network for Low-Dose CT Denoising

You, Chenyu; Yang, Qingsong; Shan, Hongming; Gjesteby, Lars; Li, Guang; Ju, Shenghong; Zhang, Zhuiyang; Zhen, Zhao; Zhang, Yi; Cong, Wenxiang; Wang, Ge

doi:10.1109/access.2018.2858196

Cited by 211 publications

(181 citation statements)

References 59 publications

Supporting

Mentioning

178

Contrasting

Order By: Relevance

“…To evaluate the effectiveness of the proposed method, we compared it with existing state-of-the-art denoising methods, including CNN-L1 (L 1 -net) 11 and WGAN-based CNN. 6 Note that all the parameters of these selected benchmark methods were set to that suggested in the original papers.…”

Section: Resultsmentioning

confidence: 99%

Low-dose CT via deep CNN with skip connection and network-in-network

You

Yang

Zhang

et al. 2019

Developments in X-Ray Tomography XII

Self Cite

View full text Add to dashboard Cite

A major challenge in computed tomography (CT) is how to minimize patient radiation exposure without compromising image quality and diagnostic performance. The use of deep convolutional (Conv) neural networks for noise reduction in Low-Dose CT (LDCT) images has recently shown a great potential in this important application. In this paper, we present a highly efficient and effective neural network model for LDCT image noise reduction. Specifically, to capture local anatomical features we integrate Deep Convolutional Neural Networks (CNNs) and Skip connection layers for feature extraction. Also, we introduce parallelized 1 × 1 CNN, called Network in Network, to lower the dimensionality of the output from the previous layer, achieving faster computational speed at less feature loss. To optimize the performance of the network, we adopt a Wasserstein generative adversarial network (WGAN) framework. Quantitative and qualitative comparisons demonstrate that our proposed network model can produce images with lower noise and more structural details than state-of-the-art noise-reduction methods.The PSNRs, SSIMs, and RMSEs are listed in Table 1. For noise reduction, the performance metrics were significantly improved by our proposed method (DCSCN). This demonstrates that using residual learning steak artifacts and image noise can be largely removed, enhancing the image quality. In this pilot study, DCSCN achieved the best performance in terms of PSNR and SSIM, and preserved anatomical features the most faithfully. However, there still exits blurry effects as shown in Figs. 3 and 5. DCSWGAN obtained the second best results in term of SSIM. It is noted that our method DCSWGAN produced visually pleasant results with sharp edges. CONCLUSIONIn this work, we have proposed a CNN-based network with skip-connection and network in network to capture structural information and suppress image noise. First, both local and global features are cascaded through skip connections before passing to the reconstruction network. Then, multi-channels are introduced for the reconstruction network with different local receptive fields to optimize the reconstruction performance. Also, the network in network technique is applied to lower the computational complexity. Our results have suggested that the proposed method could be generalized to various medical image denoising problems but further efforts are needed for training, validation, testing, and optimization.

show abstract

Section: Resultsmentioning

confidence: 99%

Low-dose CT via deep CNN with skip connection and network-in-network

You

Yang

Zhang

et al. 2019

Developments in X-Ray Tomography XII

Self Cite

View full text Add to dashboard Cite

show abstract

“…For a reconstruction loss, the squared sum of the pixel-wise difference in value (the squared L 2 loss) is often used, but it results in blurry outputs. Alternative loss functions including the perceptual feature loss [33], structure loss [34], and sharpness loss [35] have been investigated. Each of these loss functions is dependent on the accuracy of the alignment of paired images, and aligned paired training images are indispensable for using the pix2pix framework.…”

Section: Noise Reductionmentioning

confidence: 99%

Overview of image-to-image translation by use of deep neural networks: denoising, super-resolution, modality conversion, and reconstruction in medical imaging

Kaji

Kida²

2019

Radiol Phys Technol

119

View full text Add to dashboard Cite

Since the advent of deep convolutional neural networks (DNNs), computer vision has seen an extremely rapid progress that has led to huge advances in medical imaging. Every year, many new methods are reported of in conferences such as the International Conference on Medical Image Computing and Computer Assisted Intervention (MIC-CAI) and Machine Learning for Medical Image Reconstruction (MLMIR), or published online at the preprint server arXiv. There is a plethora of surveys on applications of neural networks in medical imaging (see [1] for a relatively recent comprehensive survey). This article does not aim to cover all aspects of the field but focuses on a particular topic, image-to-image translation. Although the topic may not sound familiar, it turns out that many seemingly irrelevant applications can be understood as instances of image-to-image translation. Such applications include (1) noise reduction, (2) super-resolution, (3) image synthesis, and (4) reconstruction. The same underlying principles and algorithms work for various tasks. Our aim is to introduce some of the key ideas on this topic from a uniform point of view. We introduce core ideas and jargon that are specific to image processing by use of DNNs. Having an intuitive grasp of the core ideas of and a knowledge of technical terms would be of great help to the reader for understanding the existing and future applications.Most of the recent applications which build on imageto-image translation are based on one of two fundamental The two authors contributed equally to this work.

show abstract

“…This technique is implemented in two phases and provides the best results for the MRI. Structurally-Sensitive multi-scale deep Neural network for low-dose CT denoising [3] this paper present the work related to CT denoising. The risk related the ionizing radiation is to be measured.…”

Section: Literature Surveymentioning

confidence: 99%

“…Ali Almuntashri [9] 2009 Noise-resilient edge detection algo A improved edge detection algo based with canny to reduce the noise over edges 3 The improved canned algorithm shows the improved results for the better edge detection Jiang [10] 2019 Synthetic CT generation method using Multitask max Entropy clustering MT-MEC algorithm introduce for the image segmenation with effective results 9 values of the three evaluation indicators of the MT-MEC algorithm is better than the FCM and MEC algorithms with less complex derivative.…”

Section: [8]mentioning

confidence: 99%

Brain Tumor Detection using Image Segmentation Techniques on MRI Images

Sharma*¹,

Marwaha²

2020

IJITEE

View full text Add to dashboard Cite

A refine and absolute detection of the human disease can increase the level for the survival of well being. One of the major diseases can be determine as cancer or brain tumor. Different technique used for the tumor detection such as MRI (Medical Resonance Image), Ultrasound etc. It does not involve the X-ray so there are not side effects of MRI. Claustrophobic sensation can occur with MRI scanning because of the close machine with the suffocation and no movement while the scan can increase the time to complete the scanning process while movement can cause the errors or edge distortion. An MRI or magnetic resonance imaging is a radiology technique scan that uses radio waves and with the help of computer to produce images of different body parts. MRI scanner is a tube surrounded by a giant circular magnet. The patient is placed on a moveable bed that is inserted into the magnet. The magnet creates a strong magnetic field that aligns the protons of hydrogen atoms, which are then exposed to a beam of radio waves. A computer receives the information, which produces an image. MRI is the most accurate and safe method to detect the diseases by detecting the slight change in the body. For the best result various image segmentation techniques and the good edge detection algorithms are used for finding the edge to get the effective images for the examination purpose of medical images.

show abstract

Structurally-Sensitive Multi-Scale Deep Neural Network for Low-Dose CT Denoising

Cited by 211 publications

References 59 publications

Low-dose CT via deep CNN with skip connection and network-in-network

Low-dose CT via deep CNN with skip connection and network-in-network

Overview of image-to-image translation by use of deep neural networks: denoising, super-resolution, modality conversion, and reconstruction in medical imaging

Brain Tumor Detection using Image Segmentation Techniques on MRI Images

Contact Info

Product

Resources

About