Low Dose CT Denoising by ResNet With Fused Attention Modules and Integrated Loss Functions

Marcos, Luella; Alirezaie, Javad; Babyn, Paul

doi:10.3389/frsip.2021.812193

Cited by 7 publications

(3 citation statements)

References 34 publications

(41 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Wang et al [29] found that some images with the same MSE value had obvious visual differences. In order to alleviate the problems caused by MSE, Marcos et al [30]introduced the perceptual loss into VGG-16 to compared the similarity between denoised CT and NDCT images in terms of high-level features. Yin et al [31]used the multi-perceptual loss to compare the feature differences from the VGG-19 network.…”

Section: Loss Functionmentioning

confidence: 99%

Irregular Feature Enhancer for Low-dose CT Denoising

Deng,

Hu,

et al. 2024

Preprint

View full text Add to dashboard Cite

So far, deep learning based networks have been wildly applied in Low-Dose Computed Tomography (LDCT) image denoising. However, they usually adopt symmetric convolution to achieve regular feature extraction, but cannot effectively extract irregular features. Therefore, in this paper, an Irregular Feature Enhancer (IFE) focusing on effectively extracting irregular features is proposed by combining Symmetric-Asymmetric-Synergy Convolution Module (SASCM) with a hybrid loss module. Rather than simply stacking symmetric convolution layers used in traditional deep learning based networks, SASCM jointly utilizes symmetric and asymmetric convolution layers so as to effectively extract irregular tissue information of the image. In addition, the hybrid loss module is proposed to guide IFE to further mine the intrinsic feature information of the image from three perspectives: pixel point, high-level feature space, and gradient. The ablation experiments demonstrate the effectiveness and feasibility of SASCM and the hybrid loss. The quantitative experimental results also show that compared with several related LDCT denoising methods, the proposed IFE performs the best in terms of PSNR and SSIM. Furthermore, it can be observed from the qualitative visualization that the proposed IFE can recover the best image detail structure information among the compared methods.

show abstract

Section: Loss Functionmentioning

confidence: 99%

Irregular Feature Enhancer for Low-dose CT Denoising

Deng,

Hu,

et al. 2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Compared to traditional filtering methods, this approach achieves higher values of SSIM and PSNR. The study [15] suggests a Residual Network (ResNet) as a deep neural network architecture suitable for enhancing Low-Dose Computed Tomography (LDCT) images. There are many cutting-edge methods implemented in the network, such as dilated convolution, batch normalization, ReLU, and combined spatial and channel attention modules.…”

Section: Literature Surveymentioning

confidence: 99%

“…The below equation can be used to get the cutoff 𝜉 𝐴𝑉𝐷𝐷𝐹 . [15] In the aforementioned equation, 𝜉 𝐴𝑉𝐷𝐷𝐹 stands for the approximated variance, and _𝜆 𝐴𝑉𝐷𝐷𝐹 is used to adjust the AVDDF filter's smoothing properties. The computation of the multivariate variance of the vectors within the window 𝑊 is denoted by the term 𝜓 𝐴𝑉𝐷𝐷𝐹 , which is derived using below equation.…”

Section: 𝐴(𝑥 𝑖 𝑥mentioning

confidence: 99%

FPGA-Based Image Denoising Using Adaptive Vector Directional Distance Filter: Performance Analysis and Comparison

et al. 2023

Preprint

View full text Add to dashboard Cite

Image processing is essential in a wide range of applications, like medical imaging and remote sensing, where precise visual scene interpretation is critical. However, noise is typically introduced during the process of creating, editing, transmitting, and storing color images, diminishing their visual quality and the efficacy of later image processing operations. Digital image processing relies heavily on noise reduction, also known as denoising, which seeks to get rid of unwanted noise without altering the quality of the image's essential details. In this research, we show how to use a hardware/software codesign strategy to efficiently construct the Adaptive Vector Directional Distance Filter (AVDDF) on a Field-Programmable Gate Array (FPGA). To streamline the development process, we use a high-level synthesis (HLS) pipeline with the Xilinx Vivado HLS tool. The software module is written in C/C++ and runs on a high-performance Reduced Instruction Set Computer (RISC). Comparisons are done with well-established denoising techniques like the Vector Directional Distance Filter (VDDF) and the Vector Directional Filter (VDF) to evaluate the effectiveness of the AVDDF technique. For performance comparisons, evaluation criteria such as the Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Mean Squared Error (MSE) are used. In addition, the performance of the FPGA is measured in terms of its Look-Up Tables (LUTs), Flip-Flops (FFs), BRAM_18K (Block RAMs), and DSP48E (Digital Signal Processing Blocks) using Verilog. The area, power, and delay performance of the suggested ASIC implementation is evaluated as well. The findings of this study prove that the AVDDF denoising method is effective in enhancing image quality while keeping crucial image properties unchanged.

show abstract