JBFnet - Low Dose CT Denoising by Trainable Joint Bilateral Filtering

Patwari, Mayank; Gutjahr, Ralf; Raupach, Rainer; Maier, Andreas

doi:10.1007/978-3-030-59713-9_49

Cited by 19 publications

(25 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The quadratic autoencoder architecture (QAE) proposed by Fan et al 13 , as well as the JBFnet by Patwari et al 12 , both achieve remarkable results on the CT denoising task, outperforming many architectures proposed during the 2016 Low Dose CT Grand Challenge.…”

Section: Iid Experimental Setupmentioning

confidence: 99%

“…The Adam optimizer with learning rates 4 • 10 −4 (decaying), 1 • 10 −4 (constant), 1 • 10 −4 (decaying), and 1 • 10 −5 (constant) is used for QAE, JBFnet, RED-CNN, and WGAN respectively. Both QAE and RED-CNN are trained using the MSE loss, whereas JBFnet makes use of a custom loss described in their publication 12 as well as pre-training of its prior module. Generator and discriminator of WGAN are trained alternately using Wasserstein loss for stable convergence and a perceptual loss as presented in 14 .…”

Section: Iid Experimental Setupmentioning

confidence: 99%

“…Whereas such approaches usually require hand-tuned hyperparameters, cannot abstract complex features, or are known to be computationally expensive, purely data-driven, end-to-end trainable, approaches have been proposed in recent years 11,12,13,14,15,16,17 fueled by the emergence of deep learning and, in particular, convolutional neural networks. Most of these models achieve competitive denoising performance but are built on deep neural networks with multiple layers and contain hundreds of thousands of trainable parameters.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Ultra Low-Parameter Denoising: Trainable Bilateral Filter Layers in Computed Tomography

Wagner¹,

Thies²,

Gu³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Background: Computed tomography is widely used as an imaging tool to visualize three-dimensional structures with expressive bone-soft tissue contrast. However, CT resolution and radiation dose are tightly entangled, highlighting the importance of lowdose CT combined with sophisticated denoising algorithms. Purpose: Most data-driven denoising techniques are based on deep neural networks and, therefore, contain hundreds of thousands of trainable parameters, making them incomprehensible and prone to prediction failures. Developing understandable and robust denoising algorithms achieving state-of-the-art performance helps to minimize radiation dose while maintaining data integrity. Methods: This work presents an open-source CT denoising framework based on the idea of bilateral filtering. We propose a bilateral filter that can be incorporated into a deep learning pipeline and optimized in a purely data-driven way by calculating the gradient flow toward its hyperparameters and its input. Denoising in pure image-to-image pipelines and across different domains such as raw detector data and reconstructed volume, using a differentiable backprojection layer, is demonstrated. Results:Although only using three spatial parameters and one range parameter per filter layer, the proposed denoising pipelines can compete with deep state-of-the-art denoising architectures with several hundred thousand parameters. Competitive denoising performance is achieved on x-ray microscope bone data (0.7053 and 33.10) and the 2016 Low Dose CT Grand Challenge dataset (0.9674 and 43.07) in terms of SSIM and PSNR.

show abstract

Section: Iid Experimental Setupmentioning

confidence: 99%

Section: Iid Experimental Setupmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Ultra Low-Parameter Denoising: Trainable Bilateral Filter Layers in Computed Tomography

Wagner¹,

Thies²,

Gu³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Also, JBFnet was split into four filtering blocks, each of which performs filtering offering improved performance than state of the art. 36…”

Section: Bilateral Filtermentioning

confidence: 99%

Denoising of computed tomography using bilateral median based autoencoder network

Juneja

Joshi

Singla

et al. 2021

Int J Imaging Syst Tech

View full text Add to dashboard Cite

Denoising of Computed tomography (CT) images is a critical aspect of image processing that is expected to improve the performance of Computer‐aided diagnosis (CAD) systems. However, the use of complex imaging modalities such as CT imaging to ascertain pancreatic cancer is vulnerable to gaussian and poisson noises, making image denoising an imperative step for the accurate performance of CAD systems. This paper presents a Bilateral median based autoencoder network (BMAuto‐Net) constructed with intermediate batch normalization layers and dropout factors to eliminate gaussian noise from the CT images. The skip connections adjoining the network, prevent performance degradation that generally occurs in most autoencoder architectures. Based on the presented study, BMAuto‐Net is reckoned to outperform other traditional filters and autoencoders. The performance measurement of the proposed architecture is performed using the peak signal‐to‐noise ratio (PSNR), mean squared error (MSE), and structured similarity index (SSIM) metric values. The Cancer imaging archive (TCIA) dataset consisting of 19 000 CT images is used to validate the performance of the architecture with average PSNR values of 30.01, 30.53, and 30.52, MSE values of 98.23, 98.87, and 98.94, and SSIM of values of 0.67, 0.60, and 0.57 for noise factors (NFs) of 0.1, 0.3, and 0.5 respectively.

show abstract

“…Deep learning approaches have been successfully applied to the CT denoising problem 22,23,24 . Most deep learning approaches for CT denoising are formulated in the form of image translation tasks 25,26,27,28,29,30,31,32 . CNNs learn a mapping from a noisy CT volume to a clean CT volume.…”

Section: Introductionmentioning

confidence: 99%

Limited Parameter Denoising for Low-dose X-ray Computed Tomography Using Deep Reinforcement Learning

Patwari,

Gutjahr,

Raupach

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

The use of deep learning has successfully solved several problems in the field of medical imaging. Deep learning has been applied to the CT denoising problem successfully. However, the use of deep learning requires large amounts of data to train deep convolutional networks (CNNs). Moreover, due to large parameter count, such deep CNNs may cause unexpected results. In this study, we introduce a novel CT denoising framework, which has interpretable behaviour, and provides useful results with limited data. Methods: We employ bilateral filtering in both the projection and volume domains to remove noise. To account for non-stationary noise, we tune the σ parameters of the volume for every projection view, and for every volume pixel. The tuning is carried out by two deep CNNs. Due to impracticality of labelling, the two deep CNNs are trained via a Deep-Q reinforcement learning task. The reward for the task is generated by using a custom reward function represented by a neural network. Our experiments were carried out on abdominal scans for the Mayo Clinic TCIA dataset, and the AAPM Low Dose CT Grand Challenge. Results: Our denoising framework has excellent denoising performance increasing the PSNR from 28.53 to 28.93, and increasing the SSIM from 0.8952 to 0.9204. We outperform several state-of-the-art deep CNNs, which have several orders of magnitude higher number of parameters (p-value (PSNR) = 0.000, p-value (SSIM) = 0.000). Our method does not introduce any blurring, which is introduced by MSE loss based methods, or any deep learning artifacts, which are introduced by WGAN based models. Our ablation studies show that parameter tuning and using our reward network results in the best possible results. Conclusions: We present a novel CT denoising framework, which focuses on interpretability to deliver good denoising performance, especially with limited data. Our method outperforms state-of-the-art deep neural networks. Future work will be focused on accelerating our method, and generalizing to different geometries and body parts.

show abstract

JBFnet - Low Dose CT Denoising by Trainable Joint Bilateral Filtering

Cited by 19 publications

References 19 publications

Ultra Low-Parameter Denoising: Trainable Bilateral Filter Layers in Computed Tomography

Ultra Low-Parameter Denoising: Trainable Bilateral Filter Layers in Computed Tomography

Denoising of computed tomography using bilateral median based autoencoder network

Limited Parameter Denoising for Low-dose X-ray Computed Tomography Using Deep Reinforcement Learning

Contact Info

Product

Resources

About