Synthesis of Positron Emission Tomography (PET) Images via Multi-channel Generative Adversarial Networks (GANs)

Attention-Like Multimodality Fusion With Data Augmentation for Diagnosis of Mental Disorders Using MRI

The globally rising prevalence of mental disorders leads to shortfalls in timely diagnosis and therapy to reduce patients' suffering. Facing such an urgent public health problem, professional efforts based on symptom criteria are seriously overstretched. Recently, the successful applications of computer-aided diagnosis approaches have provided timely opportunities to relieve the tension in healthcare services. Particularly, multimodal representation learning gains increasing attention thanks to the high temporal and spatial resolution information extracted from neuroimaging fusion. In this work, we propose an efficient multimodality fusion framework to identify multiple mental disorders based on the combination of functional and structural magnetic resonance imaging. A multioutput conditional generative adversarial network (GAN) is developed to address the scarcity of multimodal data for augmentation. Based on the augmented training data, the multiheaded gating fusion model is proposed for classification by extracting the complementary features across different modalities. The experiments demonstrate that the proposed model can achieve robust accuracies of 75.1 ± 1.5 %, 72.9 ± 1.1 %, and 87.2 ± 1.5 % for autism spectrum disorder (ASD), attention deficit/hyperactivity disorder, and schizophrenia, respectively. In addition, the interpretability of our model is expected to enable the identification of remarkable neuropathology diagnostic biomarkers, leading to well-informed therapeutic decisions.

Learning Without Real Data Annotations to Detect Hepatic Lesions in PET Images

OBJECTIVE

Deep neural networks have been recently applied to lesion identification in fluorodeoxyglucose (FDG) positron emission tomography (PET) images, but they typically rely on a large amount of well-annotated data for model training. This is extremely difficult to achieve for neuroendocrine tumors (NETs), because of low incidence of NETs and expensive lesion annotation in PET images. The objective of this study is to design a novel, adaptable deep learning method, which uses no real lesion annotations but instead low-cost, list mode-simulated data, for hepatic lesion detection in real-world clinical NET PET images.

METHODS

We first propose a region-guided generative adversarial network (RG-GAN) for lesion-preserved image-to-image translation. Then, we design a specific data augmentation module for our list-mode simulated data and incorporate this module into the RG-GAN to improve model training. Finally, we combine the RG-GAN, the data augmentation module and a lesion detection neural network into a unified framework for joint-task learning to adaptatively identify lesions in real-world PET data.

RESULTS

The proposed method outperforms recent state-of-the-art lesion detection methods in real clinical 68Ga-DOTATATE PET images, and produces very competitive performance with the target model that is trained with real lesion annotations.

CONCLUSION

With RG-GAN modeling and specific data augmentation, we can obtain good lesion detection performance without using any real data annotations.

SIGNIFICANCE

This study introduces an adaptable deep learning method for hepatic lesion identification in NETs, which can significantly reduce human effort for data annotation and improve model generalizability for lesion detection with PET imaging.

A Survey on Artificial Intelligence in Pulmonary Imaging

Over the last decade, deep learning (DL) has contributed a paradigm shift in computer vision and image recognition creating widespread opportunities of using artificial intelligence in research as well as industrial applications. DL has been extensively studied in medical imaging applications, including those related to pulmonary diseases. Chronic obstructive pulmonary disease, asthma, lung cancer, pneumonia, and, more recently, COVID-19 are common lung diseases affecting nearly 7.4% of world population. Pulmonary imaging has been widely investigated toward improving our understanding of disease etiologies and early diagnosis and assessment of disease progression and clinical outcomes. DL has been broadly applied to solve various pulmonary image processing challenges including classification, recognition, registration, and segmentation. This paper presents a survey of pulmonary diseases, roles of imaging in translational and clinical pulmonary research, and applications of different DL architectures and methods in pulmonary imaging with emphasis on DL-based segmentation of major pulmonary anatomies such as lung volumes, lung lobes, pulmonary vessels, and airways as well as thoracic musculoskeletal anatomies related to pulmonary diseases.

Within-Modality Synthesis and Novel Radiomic Evaluation of Brain MRI Scans

One of the most common challenges in brain MRI scans is to perform different MRI sequences depending on the type and properties of tissues. In this paper, we propose a generative method to translate T2-Weighted (T2W) Magnetic Resonance Imaging (MRI) volume from T2-weight-Fluid-attenuated-Inversion-Recovery (FLAIR) and vice versa using Generative Adversarial Networks (GAN). To evaluate the proposed method, we propose a novel evaluation schema for generative and synthetic approaches based on radiomic features. For the evaluation purpose, we consider 510 pair-slices from 102 patients to train two different GAN-based architectures Cycle GAN and Dual Cycle-Consistent Adversarial network (DC²Anet). The results indicate that generative methods can produce similar results to the original sequence without significant change in the radiometric feature. Therefore, such a method can assist clinics to make decisions based on the generated image when different sequences are not available or there is not enough time to re-perform the MRI scans.

Data Augmentation in High Dimensional Low Sample Size Setting Using a Geometry-Based Variational Autoencoder

In this paper, we propose a new method to perform data augmentation in a reliable way in the High Dimensional Low Sample Size (HDLSS) setting using a geometry-based variational autoencoder (VAE). Our approach combines the proposal of 1) a new VAE model, the latent space of which is modeled as a Riemannian manifold and which combines both Riemannian metric learning and normalizing flows and 2) a new generation scheme which produces more meaningful samples especially in the context of small data sets. The method is tested through a wide experimental study where its robustness to data sets, classifiers and training samples size is stressed. It is also validated on a medical imaging classification task on the challenging ADNI database where a small number of 3D brain magnetic resonance images (MRIs) are considered and augmented using the proposed VAE framework. In each case, the proposed method allows for a significant and reliable gain in the classification metrics. For instance, balanced accuracy jumps from 66.3% to 74.3% for a state-of-the-art convolutional neural network classifier trained with 50 MRIs of cognitively normal (CN) and 50 Alzheimer disease (AD) patients and from 77.7% to 86.3% when trained with 243 CN and 210 AD while improving greatly sensitivity and specificity metrics.

Shifting machine learning for healthcare from development to deployment and from models to data

In the past decade, the application of machine learning (ML) to healthcare has helped drive the automation of physician tasks as well as enhancements in clinical capabilities and access to care. This progress has emphasized that, from model development to model deployment, data play central roles. In this Review, we provide a data-centric view of the innovations and challenges that are defining ML for healthcare. We discuss deep generative models and federated learning as strategies to augment datasets for improved model performance, as well as the use of the more recent transformer models for handling larger datasets and enhancing the modelling of clinical text. We also discuss data-focused problems in the deployment of ML, emphasizing the need to efficiently deliver data to ML models for timely clinical predictions and to account for natural data shifts that can deteriorate model performance.

Self-supervised learning for modal transfer of brain imaging

Today's brain imaging modality migration techniques are transformed from one modality data in one domain to another. In the specific clinical diagnosis, multiple modal data can be obtained in the same scanning field, and it is more beneficial to synthesize missing modal data by using the diversity characteristics of multiple modal data. Therefore, we introduce a self-supervised learning cycle-consistent generative adversarial network (BSL-GAN) for brain imaging modality transfer. The framework constructs multi-branch input, which enables the framework to learn the diversity characteristics of multimodal data. In addition, their supervision information is mined from large-scale unsupervised data by establishing auxiliary tasks, and the network is trained by constructing supervision information, which not only ensures the similarity between the input and output of modal images, but can also learn valuable representations for downstream tasks.

Dynamic Recognition and Analysis of Gait Contour of Dance Movements Based on Generative Adversarial Networks

With the generation of images, videos, and other data, how to identify the gait of the action in the video has gradually become the focus of research. Aiming at the problems of complex and changeable movements, strong coherence, and serious occlusion in dance video images, this paper proposes a dynamic recognition model of gait contour of dance movements based on GAN (generative adversarial networks). GAN method is used to convert the gait diagrams in any state into a group of gait diagrams in normal state with multiple angles, which are arranged in turn. In order to retain as much original feature information as possible, multiple loss strategy is adopted to optimize the network, increase the distance between classes, and reduce the distance within classes. Experimental results show that the average recognition rates of this model at 50°, 90°, and 120°are 93.24, 98.24, and 97.93, respectively, which shows that the recognition accuracy of dance movement recognition method is high. And this method can effectively improve the dynamic recognition of gait contour of dance movements.

Applications of Generative Adversarial Networks (GANs) in Positron Emission Tomography (PET) imaging: A review

PURPOSE

This paper reviews recent applications of Generative Adversarial Networks (GANs) in Positron Emission Tomography (PET) imaging. Recent advances in Deep Learning (DL) and GANs catalysed the research of their applications in medical imaging modalities. As a result, several unique GAN topologies have emerged and been assessed in an experimental environment over the last two years.

METHODS

The present work extensively describes GAN architectures and their applications in PET imaging. The identification of relevant publications was performed via approved publication indexing websites and repositories. Web of Science, Scopus, and Google Scholar were the major sources of information.

RESULTS

The research identified a hundred articles that address PET imaging applications such as attenuation correction, de-noising, scatter correction, removal of artefacts, image fusion, high-dose image estimation, super-resolution, segmentation, and cross-modality synthesis. These applications are presented and accompanied by the corresponding research works.

CONCLUSION

GANs are rapidly employed in PET imaging tasks. However, specific limitations must be eliminated to reach their full potential and gain the medical community's trust in everyday clinical practice.

Deep learning in structural and functional lung image analysis

The recent resurgence of deep learning (DL) has dramatically influenced the medical imaging field. Medical image analysis applications have been at the forefront of DL research efforts applied to multiple diseases and organs, including those of the lungs. The aims of this review are twofold: (i) to briefly overview DL theory as it relates to lung image analysis; (ii) to systematically review the DL research literature relating to the lung image analysis applications of segmentation, reconstruction, registration and synthesis. The review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. 479 studies were initially identified from the literature search with 82 studies meeting the eligibility criteria. Segmentation was the most common lung image analysis DL application (65.9% of papers reviewed). DL has shown impressive results when applied to segmentation of the whole lung and other pulmonary structures. DL has also shown great potential for applications in image registration, reconstruction and synthesis. However, the majority of published studies have been limited to structural lung imaging with only 12.9% of reviewed studies employing functional lung imaging modalities, thus highlighting significant opportunities for further research in this field. Although the field of DL in lung image analysis is rapidly expanding, concerns over inconsistent validation and evaluation strategies, intersite generalisability, transparency of methodological detail and interpretability need to be addressed before widespread adoption in clinical lung imaging workflow.

On the proper use of structural similarity for the robust evaluation of medical image synthesis models

PURPOSE

To propose good practices for using the structural similarity metric (SSIM) and reporting its value. SSIM is one of the most popular image quality metrics in use in the medical image synthesis community because of its alleged superiority over voxel-by-voxel measurements like the average error or the peak signal noise ratio (PSNR). It has seen massive adoption since its introduction, but its limitations are often overlooked. Notably, SSIM is designed to work on a strictly positive intensity scale, which is generally not the case in medical imaging. Common intensity scales such as the Houndsfield units (HU) contain negative numbers, and they can also be introduced by image normalization techniques such as the z-normalization.

METHODS

We created a series of experiments to quantify the impact of negative values in the SSIM computation. Specifically, we trained a 3D U-Net to synthesize T2 weighted MRI from T1 weighted MRI using the BRATS 2018 dataset. SSIM was computed on the synthetic images with a shifted dynamic range. Next, to evaluate the suitability of SSIM as a loss function on images with negative values, it was used as a loss function to synthesize z-normalized images. Finally, the difference between 2D SSIM and 3D SSIM was investigated using multiple 2D U-Nets trained on different planes of the images.

RESULTS

The impact of the misuse of the SSIM was quantified; it was established that it introduces a large downward bias in the computed SSIM. It also introduces a small random error that can change the relative ranking of models. The exact values for this bias and error depend on the quality and the intensity histogram of the synthetic images. Although small, the reported error is significant considering the small SSIM difference between state-of-the-art models. It was shown therefore that SSIM cannot be used as a loss function when images contain negative values due to major errors in the gradient calculation, resulting in under-performing models. 2D SSIM was also found to be overestimated in 2D image synthesis models when computed along the plane of synthesis, due to the discontinuities between slices that is typical of 2D synthesis methods.

CONCLUSION

Various types of misuse of the SSIM were identified and their impact was quantified. Based on the findings, this paper proposes good practices when using SSIM, such as reporting the average over the volume of the image containing tissue and appropriately defining the dynamic range. This article is protected by copyright. All rights reserved.

CryoETGAN: Cryo-Electron Tomography Image Synthesis via Unpaired Image Translation

Cryo-electron tomography (Cryo-ET) has been regarded as a revolution in structural biology and can reveal molecular sociology. Its unprecedented quality enables it to visualize cellular organelles and macromolecular complexes at nanometer resolution with native conformations. Motivated by developments in nanotechnology and machine learning, establishing machine learning approaches such as classification, detection and averaging for Cryo-ET image analysis has inspired broad interest. Yet, deep learning-based methods for biomedical imaging typically require large labeled datasets for good results, which can be a great challenge due to the expense of obtaining and labeling training data. To deal with this problem, we propose a generative model to simulate Cryo-ET images efficiently and reliably: CryoETGAN. This cycle-consistent and Wasserstein generative adversarial network (GAN) is able to generate images with an appearance similar to the original experimental data. Quantitative and visual grading results on generated images are provided to show that the results of our proposed method achieve better performance compared to the previous state-of-the-art simulation methods. Moreover, CryoETGAN is stable to train and capable of generating plausibly diverse image samples.

When medical images meet generative adversarial network: recent development and research opportunities

Deep learning techniques have promoted the rise of artificial intelligence (AI) and performed well in computer vision. Medical image analysis is an important application of deep learning, which is expected to greatly reduce the workload of doctors, contributing to more sustainable health systems. However, most current AI methods for medical image analysis are based on supervised learning, which requires a lot of annotated data. The number of medical images available is usually small and the acquisition of medical image annotations is an expensive process. Generative adversarial network (GAN), an unsupervised method that has become very popular in recent years, can simulate the distribution of real data and reconstruct approximate real data. GAN opens some exciting new ways for medical image generation, expanding the number of medical images available for deep learning methods. Generated data can solve the problem of insufficient data or imbalanced data categories. Adversarial training is another contribution of GAN to medical imaging that has been applied to many tasks, such as classification, segmentation, or detection. This paper investigates the research status of GAN in medical images and analyzes several GAN methods commonly applied in this area. The study addresses GAN application for both medical image synthesis and adversarial learning for other medical image tasks. The open challenges and future research directions are also discussed.

Self-Attentive Spatial Adaptive Normalization for Cross-Modality Domain Adaptation

Despite the successes of deep neural networks on many challenging vision tasks, they often fail to generalize to new test domains that are not distributed identically to the training data. The domain adaptation becomes more challenging for cross-modality medical data with a notable domain shift. Given that specific annotated imaging modalities may not be accessible nor complete. Our proposed solution is based on the cross-modality synthesis of medical images to reduce the costly annotation burden by radiologists and bridge the domain gap in radiological images. We present a novel approach for image-to-image translation in medical images, capable of supervised or unsupervised (unpaired image data) setups. Built upon adversarial training, we propose a learnable self-attentive spatial normalization of the deep convolutional generator network's intermediate activations. Unlike previous attention-based image-to-image translation approaches, which are either domain-specific or require distortion of the source domain's structures, we unearth the importance of the auxiliary semantic information to handle the geometric changes and preserve anatomical structures during image translation. We achieve superior results for cross-modality segmentation between unpaired MRI and CT data for multi-modality whole heart and multi-modal brain tumor MRI (T1/T2) datasets compared to the state-of-the-art methods. We also observe encouraging results in cross-modality conversion for paired MRI and CT images on a brain dataset. Furthermore, a detailed analysis of the cross-modality image translation, thorough ablation studies confirm our proposed method's efficacy.

Multi-modal MRI Image Synthesis via GAN with Multi-scale Gate Mergence

Multi-modal magnetic resonance imaging (MRI) plays a critical role in clinical diagnosis and treatment nowadays. Each modality of MRI presents its own specific anatomical features which serve as complementary information to other modalities and can provide rich diagnostic information. However, due to the limitations of time consuming and expensive cost, some image sequences of patients may be lost or corrupted, posing an obstacle for accurate diagnosis. Although current multi-modal image synthesis approaches are able to alleviate the issues to some extent, they are still far short of fusing modalities effectively. In light of this, we propose a multi-scale gate mergence based generative adversarial network model, namely MGM-GAN, to synthesize one modality of MRI from others. Notably, we have multiple down-sampling branches corresponding to input modalities to specifically extract their unique features. In contrast to the generic multi-modal fusion approach of averaging or maximizing operations, we introduce a gate mergence (GM) mechanism to automatically learn the weights of different modalities across locations, enhancing the task-related information while suppressing the irrelative information. As such, the feature maps of all the input modalities at each down-sampling level, i.e., multi-scale levels, are integrated via GM module. In addition, both the adversarial loss and the pixel-wise loss, as well as gradient difference loss (GDL) are applied to train the network to produce the desired modality accurately. Extensive experiments demonstrate that the proposed method outperforms the state-of-the-art multi-modal image synthesis methods.

Bidirectional Mapping of Brain MRI and PET With 3D Reversible GAN for the Diagnosis of Alzheimer's Disease

Combining multi-modality data for brain disease diagnosis such as Alzheimer's disease (AD) commonly leads to improved performance than those using a single modality. However, it is still challenging to train a multi-modality model since it is difficult in clinical practice to obtain complete data that includes all modality data. Generally speaking, it is difficult to obtain both magnetic resonance images (MRI) and positron emission tomography (PET) images of a single patient. PET is expensive and requires the injection of radioactive substances into the patient's body, while MR images are cheaper, safer, and more widely used in practice. Discarding samples without PET data is a common method in previous studies, but the reduction in the number of samples will result in a decrease in model performance. To take advantage of multi-modal complementary information, we first adopt the Reversible Generative Adversarial Network (RevGAN) model to reconstruct the missing data. After that, a 3D convolutional neural network (CNN) classification model with multi-modality input was proposed to perform AD diagnosis. We have evaluated our method on the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, and compared the performance of the proposed method with those using state-of-the-art methods. The experimental results show that the structural and functional information of brain tissue can be mapped well and that the image synthesized by our method is close to the real image. In addition, the use of synthetic data is beneficial for the diagnosis and prediction of Alzheimer's disease, demonstrating the effectiveness of the proposed framework.

A Tour of Unsupervised Deep Learning for Medical Image Analysis

BACKGROUND

Interpretation of medical images for the diagnosis and treatment of complex diseases from high-dimensional and heterogeneous data remains a key challenge in transforming healthcare. In the last few years, both supervised and unsupervised deep learning achieved promising results in the area of medical image analysis. Several reviews on supervised deep learning are published, but hardly any rigorous review on unsupervised deep learning for medical image analysis is available.

OBJECTIVES

The objective of this review is to systematically present various unsupervised deep learning models, tools, and benchmark datasets applied to medical image analysis. Some of the discussed models are autoencoders and its other variants, Restricted Boltzmann machines (RBM), Deep belief networks (DBN), Deep Boltzmann machine (DBM), and Generative adversarial network (GAN). Further, future research opportunities and challenges of unsupervised deep learning techniques for medical image analysis are also discussed.

CONCLUSION

Currently, interpretation of medical images for diagnostic purposes is usually performed by human experts that may be replaced by computer-aided diagnosis due to advancement in machine learning techniques, including deep learning, and the availability of cheap computing infrastructure through cloud computing. Both supervised and unsupervised machine learning approaches are widely applied in medical image analysis, each of them having certain pros and cons. Since human supervisions are not always available or inadequate or biased, therefore, unsupervised learning algorithms give a big hope with lots of advantages for biomedical image analysis.

Longitudinal Prediction of Infant MR Images With Multi-Contrast Perceptual Adversarial Learning

The infant brain undergoes a remarkable period of neural development that is crucial for the development of cognitive and behavioral capacities (Hasegawa et al., 2018). Longitudinal magnetic resonance imaging (MRI) is able to characterize the developmental trajectories and is critical in neuroimaging studies of early brain development. However, missing data at different time points is an unavoidable occurrence in longitudinal studies owing to participant attrition and scan failure. Compared to dropping incomplete data, data imputation is considered a better solution to address such missing data in order to preserve all available samples. In this paper, we adapt generative adversarial networks (GAN) to a new application: longitudinal image prediction of structural MRI in the first year of life. In contrast to existing medical image-to-image translation applications of GANs, where inputs and outputs share a very close anatomical structure, our task is more challenging as brain size, shape and tissue contrast vary significantly between the input data and the predicted data. Several improvements over existing GAN approaches are proposed to address these challenges in our task. To enhance the realism, crispness, and accuracy of the predicted images, we incorporate both a traditional voxel-wise reconstruction loss as well as a perceptual loss term into the adversarial learning scheme. As the differing contrast changes in T1w and T2w MR images in the first year of life, we incorporate multi-contrast images leading to our proposed 3D multi-contrast perceptual adversarial network (MPGAN). Extensive evaluations are performed to assess the qualityand fidelity of the predicted images, including qualitative and quantitative assessments of the image appearance, as well as quantitative assessment on two segmentation tasks. Our experimental results show that our MPGAN is an effective solution for longitudinal MR image data imputation in the infant brain. We further apply our predicted/imputed images to two practical tasks, a regression task and a classification task, in order to highlight the enhanced task-related performance following image imputation. The results show that the model performance in both tasks is improved by including the additional imputed data, demonstrating the usability of the predicted images generated from our approach.

Simultaneous super-resolution and contrast synthesis of routine clinical magnetic resonance images of the knee for improving automatic segmentation of joint cartilage: data from the Osteoarthritis Initiative

PURPOSE

High resolution three-dimensional (3D) magnetic resonance (MR) images are well suited for automated cartilage segmentation in the human knee joint. However, volumetric scans such as 3D Double-Echo Steady-State (DESS) images are not routinely acquired in clinical practice which limits opportunities for reliable cartilage segmentation using (fully) automated algorithms. In this work, a method for generating synthetic 3D MR (syn3D-DESS) images with better contrast and higher spatial resolution from routine, low resolution, two-dimensional (2D) Turbo-Spin Echo (TSE) clinical knee scans is proposed.

METHODS

A UNet convolutional neural network is employed for synthesizing enhanced artificial MR images suitable for automated knee cartilage segmentation. Training of the model was performed on a large, publically available dataset from the OAI, consisting of 578 MR examinations of knee joints from 102 healthy individuals and patients with knee osteoarthritis.

RESULTS

The generated synthetic images have higher spatial resolution and better tissue contrast than the original 2D TSE, which allow high quality automated 3D segmentations of the cartilage. The proposed approach was evaluated on a separate set of MR images from 88 subjects with manual cartilage segmentations. It provided a significant improvement in automated segmentation of knee cartilages when using the syn3D-DESS images compared to the original 2D TSE images.

CONCLUSION

The proposed method can successfully synthesize 3D DESS images from 2D TSE images to provide images suitable for automated cartilage segmentation.

Sample-Adaptive GANs: Linking Global and Local Mappings for Cross-Modality MR Image Synthesis

Generative adversarial network (GAN) has been widely explored for cross-modality medical image synthesis. The existing GAN models usually adversarially learn a global sample space mapping from the source-modality to the target-modality and then indiscriminately apply this mapping to all samples in the whole space for prediction. However, due to the scarcity of training samples in contrast to the complicated nature of medical image synthesis, learning a single global sample space mapping that is "optimal" to all samples is very challenging, if not intractable. To address this issue, this paper proposes sample-adaptive GAN models, which not only cater for the global sample space mapping between the source- and the target-modalities but also explore the local space around each given sample to extract its unique characteristic. Specifically, the proposed sample-adaptive GANs decompose the entire learning model into two cooperative paths. The baseline path learns a common GAN model by fitting all the training samples as usual for the global sample space mapping. The new sample-adaptive path additionally models each sample by learning its relationship with its neighboring training samples and using the target-modality features of these training samples as auxiliary information for synthesis. Enhanced by this sample-adaptive path, the proposed sample-adaptive GANs are able to flexibly adjust themselves to different samples, and therefore optimize the synthesis performance. Our models have been verified on three cross-modality MR image synthesis tasks from two public datasets, and they significantly outperform the state-of-the-art methods in comparison. Moreover, the experiment also indicates that our sample-adaptive strategy could be utilized to improve various backbone GAN models. It complements the existing GANs models and can be readily integrated when needed.

Medical Image Synthesis via Deep Learning

Medical images have been widely used in clinics, providing visual representations of under-skin tissues in human body. By applying different imaging protocols, diverse modalities of medical images with unique characteristics of visualization can be produced. Considering the cost of scanning high-quality single modality images or homogeneous multiple modalities of images, medical image synthesis methods have been extensively explored for clinical applications. Among them, deep learning approaches, especially convolutional neural networks (CNNs) and generative adversarial networks (GANs), have rapidly become dominating for medical image synthesis in recent years. In this chapter, based on a general review of the medical image synthesis methods, we will focus on introducing typical CNNs and GANs models for medical image synthesis. Especially, we will elaborate our recent work about low-dose to high-dose PET image synthesis, and cross-modality MR image synthesis, using these models.

CycleGAN for style transfer in X-ray angiography

PURPOSE

We aim to perform generation of angiograms for various vascular structures as a mean of data augmentation in learning tasks. The task is to enhance the realism of vessels images generated from an anatomically realistic cardiorespiratory simulator to make them look like real angiographies.

METHODS

The enhancement is performed by applying the CycleGAN deep network for transferring the style of real angiograms acquired during percutaneous interventions into a data set composed of realistically simulated arteries.

RESULTS

The cycle consistency was evaluated by comparing an input simulated image with the one obtained after two cycles of image translation. An average structural similarity (SSIM) of 0.948 on our data sets has been obtained. The vessel preservation was measured by comparing segmentations of an input image and its corresponding enhanced image using Dice coefficient.

CONCLUSIONS

We proposed an application of the CycleGAN deep network for enhancing the artificial data as an alternative to classical data augmentation techniques for medical applications, particularly focused on angiogram generation. We discussed success and failure cases, explaining conditions for the realistic data augmentation which respects both the complex physiology of arteries and the various patterns and textures generated by X-ray angiography.

Robust Histopathology Image Analysis: to Label or to Synthesize?

Detection, segmentation and classification of nuclei are fundamental analysis operations in digital pathology. Existing state-of-the-art approaches demand extensive amount of supervised training data from pathologists and may still perform poorly in images from unseen tissue types. We propose an unsupervised approach for histopathology image segmentation that synthesizes heterogeneous sets of training image patches, of every tissue type. Although our synthetic patches are not always of high quality, we harness the motley crew of generated samples through a generally applicable importance sampling method. This proposed approach, for the first time, re-weighs the training loss over synthetic data so that the ideal (unbiased) generalization loss over the true data distribution is minimized. This enables us to use a random polygon generator to synthesize approximate cellular structures (i.e., nuclear masks) for which no real examples are given in many tissue types, and hence, GAN-based methods are not suited. In addition, we propose a hybrid synthesis pipeline that utilizes textures in real histopathology patches and GAN models, to tackle heterogeneity in tissue textures. Compared with existing state-of-the-art supervised models, our approach generalizes significantly better on cancer types without training data. Even in cancer types with training data, our approach achieves the same performance without supervision cost. We release code and segmentation results on over 5000 Whole Slide Images (WSI) in The Cancer Genome Atlas (TCGA) repository, a dataset that would be orders of magnitude larger than what is available today.

3D Auto-Context-Based Locality Adaptive Multi-Modality GANs for PET Synthesis

Positron emission tomography (PET) has been substantially used recently. To minimize the potential health risk caused by the tracer radiation inherent to PET scans, it is of great interest to synthesize the high-quality PET image from the low-dose one to reduce the radiation exposure. In this paper, we propose a 3D auto-context-based locality adaptive multi-modality generative adversarial networks model (LA-GANs) to synthesize the high-quality FDG PET image from the low-dose one with the accompanying MRI images that provide anatomical information. Our work has four contributions. First, different from the traditional methods that treat each image modality as an input channel and apply the same kernel to convolve the whole image, we argue that the contributions of different modalities could vary at different image locations, and therefore a unified kernel for a whole image is not optimal. To address this issue, we propose a locality adaptive strategy for multi-modality fusion. Second, we utilize 1 ×1 ×1 kernel to learn this locality adaptive fusion so that the number of additional parameters incurred by our method is kept minimum. Third, the proposed locality adaptive fusion mechanism is learned jointly with the PET image synthesis in a 3D conditional GANs model, which generates high-quality PET images by employing large-sized image patches and hierarchical features. Fourth, we apply the auto-context strategy to our scheme and propose an auto-context LA-GANs model to further refine the quality of synthesized images. Experimental results show that our method outperforms the traditional multi-modality fusion methods used in deep networks, as well as the state-of-the-art PET estimation approaches.

Synthesizing Chest X-Ray Pathology for Training Deep Convolutional Neural Networks

Medical datasets are often highly imbalanced with over-representation of prevalent conditions and poor representation of rare medical conditions. Due to privacy concerns, it is challenging to aggregate large datasets between health care institutions. We propose synthesizing pathology in medical images as a means to overcome these challenges. We implement a deep convolutional generative adversarial network (DCGAN) to create synthesized chest X-rays based upon a modest sized labeled dataset. We used a combination of real and synthesized images to train deep convolutional neural networks (DCNNs) to detect pathology across five classes of chest X-rays. The comparative study of DCNNs trained with the combination of real and synthesized images showed that these networks can outperform similar networks trained solely with real images in pathology classification. This improved performance is largely attributable to the balancing of the dataset using DCGAN synthesized images, where classes that are lacking in example images are preferentially augmented.

Automated Pulmonary Nodule Classification in Computed Tomography Images Using a Deep Convolutional Neural Network Trained by Generative Adversarial Networks

Lung cancer is a leading cause of death worldwide. Although computed tomography (CT) examinations are frequently used for lung cancer diagnosis, it can be difficult to distinguish between benign and malignant pulmonary nodules on the basis of CT images alone. Therefore, a bronchoscopic biopsy may be conducted if malignancy is suspected following CT examinations. However, biopsies are highly invasive, and patients with benign nodules may undergo many unnecessary biopsies. To prevent this, an imaging diagnosis with high classification accuracy is essential. In this study, we investigate the automated classification of pulmonary nodules in CT images using a deep convolutional neural network (DCNN). We use generative adversarial networks (GANs) to generate additional images when only small amounts of data are available, which is a common problem in medical research, and evaluate whether the classification accuracy is improved by generating a large amount of new pulmonary nodule images using the GAN. Using the proposed method, CT images of 60 cases with confirmed pathological diagnosis by biopsy are analyzed. The benign nodules assessed in this study are difficult for radiologists to differentiate because they cannot be rejected as being malignant. A volume of interest centered on the pulmonary nodule is extracted from the CT images, and further images are created using axial sections and augmented data. The DCNN is trained using nodule images generated by the GAN and then fine-tuned using the actual nodule images to allow the DCNN to distinguish between benign and malignant nodules. This pretraining and fine-tuning process makes it possible to distinguish 66.7% of benign nodules and 93.9% of malignant nodules. These results indicate that the proposed method improves the classification accuracy by approximately 20% in comparison with training using only the original images.

Locality Adaptive Multi-modality GANs for High-Quality PET Image Synthesis

Positron emission topography (PET) has been substantially used in recent years. To minimize the potential health risks caused by the tracer radiation inherent to PET scans, it is of great interest to synthesize the high-quality full-dose PET image from the low-dose one to reduce the radiation exposure while maintaining the image quality. In this paper, we propose a locality adaptive multi-modality generative adversarial networks model (LA-GANs) to synthesize the full-dose PET image from both the low-dose one and the accompanying T1-weighted MRI to incorporate anatomical information for better PET image synthesis. This paper has the following contributions. First, we propose a new mechanism to fuse multi-modality information in deep neural networks. Different from the traditional methods that treat each image modality as an input channel and apply the same kernel to convolute the whole image, we argue that the contributions of different modalities could vary at different image locations, and therefore a unified kernel for a whole image is not appropriate. To address this issue, we propose a method that is locality adaptive for multimodality fusion. Second, to learn this locality adaptive fusion, we utilize 1 × 1 × 1 kernel so that the number of additional parameters incurred by our method is kept minimum. This also naturally produces a fused image which acts as a pseudo input for the subsequent learning stages. Third, the proposed locality adaptive fusion mechanism is learned jointly with the PET image synthesis in an end-to-end trained 3D conditional GANs model developed by us. Our 3D GANs model generates high quality PET images by employing large-sized image patches and hierarchical features. Experimental results show that our method outperforms the traditional multi-modality fusion methods used in deep networks, as well as the state-of-the-art PET estimation approaches.