Deep Learning for Low-Dose CT Denoising Using Perceptual Loss and Edge Detection Layer

Abstract

Low-dose CT denoising is a challenging task that has been studied by many researchers. Some studies have used deep neural networks to improve the quality of low-dose CT images and achieved fruitful results. In this paper, we propose a deep neural network that uses dilated convolutions with different dilation rates instead of standard convolution helping to capture more contextual information in fewer layers. Also, we have employed residual learning by creating shortcut connections to transmit image information from the early layers to later ones. To further improve the performance of the network, we have introduced a non-trainable edge detection layer that extracts edges in horizontal, vertical, and diagonal directions. Finally, we demonstrate that optimizing the network by a combination of mean-square error loss and perceptual loss preserves many structural details in the CT image. This objective function do not suffer from over smoothing and blurring effects caused by per-pixel loss and grid-like artifacts resulting from perceptual loss. The experiments show that each modification to the network improves the outcome while only minimally changing the complexity of the network.

A review on Deep Learning approaches for low-dose Computed Tomography restoration

“…This study revealed that some of the LDCT restoration applications reported in [ 20 , 23 , 71 , 73 ] followed the same ResNet model published by Zhang et al [ 86 ]. Accordingly, a cascaded ResNet-based LDCT restoration model published by Wu et al [ 73 ] has the strength to restore the noise patterns that would rarely encounter in the training datasets via iterative cascaded learning.…”

“…Residual Network (ResNet): Stacking more layers in the CNN model is one of the basic techniques for improving the performance of the CNN model. However, increasing the depth of the network will always not influence CNN positively due to the issue called gradient diffusion [ 20 , 50 ]. Also, gradient diffusion might result in failures in network training.…”

“…Hence, the proposed GRCNN has gained the training effectiveness and ability to preserve the sharpness of features of the processed LDCT images. In addition to these applications, the ResNet published by Gholizadeh-Ansari et al [ 20 ] (DRL-E-MP) has some unique features compared to other applications that followed the model in Zhang et al [ 86 ]. Those are edge-detection-based image restoration and the application of dilated convolution operations.…”

Efficiency of Artificial Intelligence in Detecting COVID-19 Pneumonia and Other Pneumonia Causes by Quantum Fourier Transform Method

“…Asa solution to this problem, many methods have been designed for denoising LDCT images. These methods generally fall into three categories: techniques based on sinogram filtering, iterative reconstruction, and post-reconstruction image processing [8][9][10]. In addition, denoising algorithms using artificial deep neural networks have demonstrated impressive performance capabilities in this context [8][9][10][11][12][13][14][15][16][17].…”

Image quality improvement of single-shot turbo spin-echo magnetic resonance imaging of female pelvis using a convolutional neural network

“…[ 9 – 14 ] Noise reduction techniques using CNN have also been proposed in the field of medical imaging. [ 15 ] Several studies have shown that deep learning-based super-resolution or denoise approaches were successfully applied to low-quality MR images to shorten imaging time. [ 16 , 17 ] However, few studies have shown the impact of the deep learning-based approaches that improve the image quality of SSTSE image.…”