Deep frequency-recurrent priors for inverse imaging reconstruction

He, Zhuonan; Hong, Kai; Zhou, Jiazhen; Liang, Dong; Wang, Yuhao; Liu, Qiegen

doi:10.1016/j.sigpro.2021.108320

Cited by 8 publications

(11 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Firstly, we evaluated the proposed StruNet on the CT dataset and compared it with several popular low‐dose CT reconstruction algorithms including BM3D, 49 K‐SVD, 6 HF‐DAEP, 9 REDAEP, 10 CNN200 7 and RED‐CNN 8 …”

Section: Resultsmentioning

confidence: 99%

StruNet: Perceptual and low‐rank regularized transformer for medical image denoising

Yan

Liu

et al. 2023

Medical Physics

View full text Add to dashboard Cite

BackgroundVarious types of noise artifacts inevitably exist in some medical imaging modalities due to limitations of imaging techniques, which impair either clinical diagnosis or subsequent analysis. Recently, deep learning approaches have been rapidly developed and applied on medical images for noise removal or image quality enhancement. Nevertheless, due to complexity and diversity of noise distribution representations in different medical imaging modalities, most of the existing deep learning frameworks are incapable to flexibly remove noise artifacts while retaining detailed information. As a result, it remains challenging to design an effective and unified medical image denoising method that will work across a variety of noise artifacts for different imaging modalities without requiring specialized knowledge in performing the task.PurposeIn this paper, we propose a novel encoder‐decoder architecture called Swin transformer‐based residual u‐shape Network (StruNet), for medical image denoising.MethodsOur StruNet adopts a well‐designed block as the backbone of the encoder‐decoder architecture, which integrates Swin Transformer modules with residual block in parallel connection. Swin Transformer modules could effectively learn hierarchical representations of noise artifacts via self‐attention mechanism in non‐overlapping shifted windows and cross‐window connection, while residual block is advantageous to compensate loss of detailed information via shortcut connection. Furthermore, perceptual loss and low‐rank regularization are incorporated into loss function respectively in order to constrain the denoising results on feature‐level consistency and low‐rank characteristics.ResultsTo evaluate the performance of the proposed method, we have conducted experiments on three medical imaging modalities including computed tomography (CT), optical coherence tomography (OCT) and optical coherence tomography angiography (OCTA).ConclusionsThe results demonstrate that the proposed architecture yields a promising performance of suppressing multiform noise artifacts existing in different imaging modalities.

show abstract

Section: Resultsmentioning

confidence: 99%

StruNet: Perceptual and low‐rank regularized transformer for medical image denoising

Yan

Liu

et al. 2023

Medical Physics

View full text Add to dashboard Cite

show abstract

“…Following the correction of earlier acquired data, deep-learning-based image auto-segmentation or reconstruction may see notable improvements, particularly for models that rely heavily on higher frequency components. 54,55 Recent studies have shown that the prediction of tumor contours on cine MRI frames using a convolutional LSTM is challenging. 56 Our intra-frame motion compensation model works in 2D, providing an efficient time latency offset for anatomical structure changes, and the network can be trained to estimate the complete intra-frame motion trajectory by generating output images not only at the final position but also at intermediate positions, offering potential information for 2D motion prediction algorithms.…”

Section: Discussionmentioning

confidence: 99%

“…Further research is necessary to explore the potential benefits of intra‐frame motion compensation for improving the accuracy of the downstream tasks in MRgRT such as beam gating, motion prediction 53 or real‐time tumor tracking. Following the correction of earlier acquired data, deep‐learning‐based image auto‐segmentation or reconstruction may see notable improvements, particularly for models that rely heavily on higher frequency components 54,55 . Recent studies have shown that the prediction of tumor contours on cine MRI frames using a convolutional LSTM is challenging 56 .…”

Section: Discussionmentioning

confidence: 99%

Intra‐frame motion deterioration effects and deep‐learning‐based compensation in MR‐guided radiotherapy

Sui,

Palaniappan,

Brenner

et al. 2023

Medical Physics

View full text Add to dashboard Cite

BackgroundCurrent commercially available hybrid magnetic resonance linear accelerators (MR‐Linac) use 2D+t cine MR imaging to provide intra‐fractional motion monitoring. However, given the limited temporal resolution of cine MR imaging, target intra‐frame motion deterioration effects, resulting in effective time latency and motion artifacts in the image domain, can be appreciable, especially in the case of fast breathing.PurposeThe aim of this work is to investigate intra‐frame motion deterioration effects in MR‐guided radiotherapy (MRgRT) by simulating the motion‐corrupted image acquisition, and to explore the feasibility of deep‐learning‐based compensation approaches, relying on the intra‐frame motion information which is spatially and temporally encoded in the raw data (k‐space).MethodsAn intra‐frame motion model was defined to simulate motion‐corrupted MR images, with 4D anthropomorphic digital phantoms being exploited to provide ground truth 2D+t cine MR sequences. A total number of 10 digital phantoms were generated for lung cancer patients, with randomly selected eight patients for training or validation and the remaining two for testing. The simulation code served as the data generator, and a dedicated motion pattern perturbation scheme was proposed to build the intra‐frame motion database, where three degrees of freedom were designed to guarantee the diversity of intra‐frame motion trajectories, enabling a thorough exploration in the domain of the potential anatomical structure positions. U‐Nets with three types of loss functions: L1 or L2 loss defined in image or Fourier domain, referred to as NNImgLoss‐L1, NNFloss‐L1 and NNL2‐Loss were trained to extract information from the motion‐corrupted image and used to estimate the ground truth final‐position image, corresponding to the end of the acquisition. Images before and after compensation were evaluated in terms of (i) image mean‐squared error (MSE) and mean absolute error (MAE), and (ii) accuracy of gross tumor volume (GTV) contouring, based on optical‐flow image registration.ResultsImage degradation caused by intra‐frame motion was observed: for a linearly and fully acquired Cartesian readout k‐space trajectory, intra‐frame motion resulted in an imaging latency of approximately 50% of the acquisition time; in comparison, the motion artifacts exhibited only a negligible contribution to the overall geometric errors. All three compensation models led to a decrease in image MSE/MAE and GTV position offset compared to the motion‐corrupted image. In the investigated testing dataset for GTV contouring, the average dice similarity coefficients (DSC) improved from 88% to 96%, and the 95th percentile Hausdorff distance (HD95) dropped from 4.8 mm to 2.1 mm. Different models showed slight performance variations across different intra‐frame motion amplitude categories: NNImgLoss‐L1 excelled for small/medium amplitudes, whereas NNFloss‐L1 demonstrated higher DSC median values at larger amplitudes. The saliency maps of the motion‐corrupted image highlighted the major contribution of the later acquired k‐space data, as well as the edges of the moving anatomical structures at their final positions, during the model inference stage.ConclusionsOur results demonstrate the deep‐learning‐based approaches have the potential to compensate for intra‐frame motion by utilizing the later acquired data to drive the convergence of the earlier acquired k‐space components.

show abstract

“…Os AE's podem ser utilizados em diversas categorias de problemas. Alguns autores optam por sua utilizac ¸ão em detecc ¸ão de anomalias [Zhang et al 2022, Ko et al 2021, outros preferem trabalhos na área de visão computacional para recuperac ¸ão de informac ¸ão e retirada de ruídos nas imagens [Yilmaz et al 2022, Xu et al 2022, He et al 2022, Vijayalakshmi and Shanthakumar 2019. Neste último caso o modelo recebe um conjunto de entrada em que há falhas ou falta de informac ¸ão nas imagens e o objetivo é recriar/retirar ruído das fotos.…”

Section: Codificadores Automáticos -(Autoencoders)unclassified

O Uso de AutoEncoders como Fomentador de Medidas em Sistemas de Distribuição

Silva¹,

Souza²,

Filho³

2022

Anais Do XLIX Seminário Integrado De Software E Hardware (SEMISH 2022)

View full text Add to dashboard Cite

O sucesso de funções de automação e controle vislumbradas para as redes de distribuição inteligentes depende da supervisão confiável da rede em tempo real. Esta tarefa é realizada pelo estimador de estado da distribuição, responsável por processar um conjunto de medidas recebidas através do sistema de aquisição de dados. Em redes inteligentes, a infraestrutura avançada de medição (AMI) permite a leitura regular de medidas de tensão e potência dos consumidores, as quais podem complementar as poucas medidas usualmente presentes na rede de distribuição (medidas SCADA) e beneficiar o processo de estimação de estado. Entretanto, devido a limitações na infraestrutura de comunicação, medidas SCADA e AMI possuem taxas de amostragem e aquisição diferentes. Este trabalho apresenta uma metodologia que emprega uma rede neural do tipo AutoEncoder para a geração de pseudomedidas AMI para complementar medidas SCADA em instantes de tempo em que apenas estas últimas estão disponíveis para processamento. Simulações realizadas com um sistema de distribuição de 34 barras ilustram a metodologia proposta e os resultados obtidos confirmam o seu potencial de aplicação para a geração de pseudomedidas.

show abstract

Deep frequency-recurrent priors for inverse imaging reconstruction

Cited by 8 publications

References 42 publications

StruNet: Perceptual and low‐rank regularized transformer for medical image denoising

StruNet: Perceptual and low‐rank regularized transformer for medical image denoising

Intra‐frame motion deterioration effects and deep‐learning‐based compensation in MR‐guided radiotherapy

O Uso de AutoEncoders como Fomentador de Medidas em Sistemas de Distribuição

Contact Info

Product

Resources

About