Cross-modality Synthesis from MRI to PET Using Adversarial U-Net with Different Normalization

Hu, Shengye; Yuan, Jianpeng; Wang, Shuqiang

doi:10.1109/icmipe47306.2019.9098219

Cited by 44 publications

(11 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the optimization theory [30] and deep learning technology developed in recent years, the medical images computing technology has made great success. For example, automatic diagnosis [32, 34-38, 50, 52], disease prediction [19,20,44], medical image synthesis [8][9][10][11]51].…”

Section: Related Work 21 Medical Image Computingmentioning

confidence: 99%

Brain MR Images Super-Resolution with the Consistent Features

You

Shen

et al. 2022

2022 14th International Conference on Machine Learning and Computing (ICMLC)

Self Cite

View full text Add to dashboard Cite

Magnetic resonance imaging plays an important role in auxiliary diagnosis and brain exploration. However, limited by hardware, scanning time and cost, it's challenging to acquire high-resolution (HR) magnetic resonance (MR) image clinically. In this paper, consistent feature generative adversarial network (CFGAN) is proposed to produce HR MR images from the low-resolution counterparts. Specifically, a consistent-features encoder is employed to extract the multi-scales features and encode them into latent codes. Then, a progressive generator is utilized to decode the latent codes from high-level to low-level features. With the encoder and generator, the shared consistent features between low-resolution and highresolution can be fully extracted and recovered. Experiments on ADNI dataset demonstrate that CFGAN outperforms the competing methods quantitatively and qualitatively. CCS CONCEPTS• Computing methodologies → Machine Learning.

show abstract

Section: Related Work 21 Medical Image Computingmentioning

confidence: 99%

Brain MR Images Super-Resolution with the Consistent Features

You

Shen

et al. 2022

2022 14th International Conference on Machine Learning and Computing (ICMLC)

Self Cite

View full text Add to dashboard Cite

show abstract

“…For example, Kao et al [16] apply GN for brain tumor segmentation; Isensee et al [15] apply IN in a self-adaptive framework for various segmentation tasks; Chen et al [3] apply LN in Transformers for abdominal multi-organ segmentation. A detailed comparison among different normalization for biomedical semantic segmentation and cross-modality synthesis have been summarized in [38] and [12]. To offer stronger affine transformation, later methods [6,13] utilize external data to denormalize the features.…”

Section: Related Workmentioning

confidence: 99%

CateNorm: Categorical Normalization for Robust Medical Image Segmentation

Xiao¹,

Yu²,

Zhou³

et al. 2021

Preprint

View full text Add to dashboard Cite

Batch Normalization (BN) is one of the key components for accelerating network training, and has been widely adopted in the medical image analysis field. However, BN only calculates the global statistics at the batch level, and applies the same affine transformation uniformly across all spatial coordinates, which would suppress the image contrast of different semantic structures. In this paper, we propose to incorporate the semantic class information into normalization layers, so that the activations corresponding to different regions (i.e., classes) can be modulated differently. We thus develop a novel DualNorm-UNet, to concurrently incorporate both global image-level statistics and local region-wise statistics for network normalization. Specifically, the local statistics are integrated by adaptively modulating the activations along different class regions via the learned semantic masks in the normalization layer. Compared with existing methods, our approach exploits semantic knowledge at normalization and yields more discriminative features for robust segmentation results. More importantly, our network demonstrates superior ability in capturing domain-invariant information from multiple domains (institutions) of medical data. Extensive experiments show that our proposed DualNorm-UNet consistently improves the performance on various segmentation tasks, even in the face of more complex and variable data distributions. Code is available at https://github.com/lambert-x/ DualNorm-Unet.

show abstract

“…For the normalization layers, we use the instance normalization rather than batch normalization to largely ease the optimization and benefit the generalization of deep networks. It has been proved to have a better performance in image synthetic tasks [33], [34]. In the output layer, the normalization layer is removed and the Tanh function is adopted to obtain the synthetic PET images.…”

Section: Basic Ideas Of Bmganmentioning

confidence: 99%

Bidirectional Mapping Generative Adversarial Networks for Brain MR to PET Synthesis

Hu¹,

Lei²,

Wang³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Fusing multi-modality medical images, such as MR and PET, can provide various anatomical or functional information about human body. But PET data is always unavailable due to different reasons such as cost, radiation, or other limitations. In this paper, we propose a 3D end-to-end synthesis network, called Bidirectional Mapping Generative Adversarial Networks (BMGAN), where image contexts and latent vector are effectively used and jointly optimized for brain MR-to-PET synthesis. Concretely, a bidirectional mapping mechanism is designed to embed the semantic information of PET images into the highdimensional latent space. And the 3D DenseU-Net generator architecture and the extensive objective functions are further utilized to improve the visual quality of synthetic results. The most appealing part is that the proposed method can synthesize the perceptually realistic PET images while preserving the diverse brain structures of different subjects. Experimental results demonstrate that the performance of the proposed method outperforms other competitive cross-modality synthesis methods in terms of quantitative measures, qualitative displays, and classification evaluation.

show abstract

Cross-modality Synthesis from MRI to PET Using Adversarial U-Net with Different Normalization

Cited by 44 publications

References 11 publications

Brain MR Images Super-Resolution with the Consistent Features

Brain MR Images Super-Resolution with the Consistent Features

CateNorm: Categorical Normalization for Robust Medical Image Segmentation

Bidirectional Mapping Generative Adversarial Networks for Brain MR to PET Synthesis

Contact Info

Product

Resources

About