We demonstrate training of a Generative Adversarial Network (GAN) for prediction of optical property maps (scattering and absorption) using spatial frequency domain imaging (SFDI) image data sets generated synthetically with free open-source 3D modelling and rendering software, Blender. The flexibility of Blender is exploited to simulate 3 models with real-life relevance to clinical SFDI of diseased tissue: flat samples, flat samples with spheroidal tumours and cylindrical samples with spheroidal tumours representing imaging inside a tubular organ e.g. the gastro-intestinal tract. In all 3 scenarios we show the GAN provides accurate reconstruction of optical properties from single SFDI images with mean normalised error ranging from 1-1.2% for absorption and 0.7-1.2% for scattering, resulting in visually improved contrast for tumour spheroid structures. This compares favourably with 25% absorption error and 10% scattering error achieved using GANs on experimental SFDI data. However, some of this improvement is due to lower noise and availability of perfect ground truths so we therefore cross-validate our synthetically-trained GAN with a GAN trained on experimental data and observe visually accurate results with error of <40% for absorption and <25% for scattering, due largely to the presence of spatial frequency mismatch artefacts. Our synthetically trained GAN is therefore highly relevant to real experimental samples, but provides significant added benefits of large training datasets, perfect ground-truths and the ability to test realistic imaging geometries, e.g. inside cylinders, for which no conventional single-shot demodulation algorithms exist. In future we expect that application of techniques such as domain adaptation or training on hybrid real-synthetic datasets will create a powerful tool for fast, accurate production of optical property maps from real clinical imaging systems.
The proposed method, Neural Radiance Projection (NeRP), addresses the three most fundamental shortages of training such a convolutional neural network on X-ray image segmentation: dealing with missing/limited human-annotated datasets; ambiguity on the per-pixel label; and the imbalance across positive- and negative- classes distribution. By harnessing a generative adversarial network, we can synthesize a massive amount of physics-based X-ray images, so-called Variationally Reconstructed Radiographs (VRRs), alongside their segmentation from more accurate labeled 3D Computed Tomography data. As a result, VRRs present more faithfully than other projection methods in terms of photo-realistic metrics. Adding outputs from NeRP also surpasses the vanilla UNet models trained on the same pairs of X-ray images.
In recent years, the machine learning research community has benefited tremendously from the availability of openly accessible benchmark datasets. Clinical data are usually not openly available due to their highly confidential nature. This has hampered the development of reproducible and generalisable machine learning applications in health care. Here we introduce the Health Gym - a growing collection of highly realistic synthetic medical datasets that can be freely accessed to prototype, evaluate, and compare machine learning algorithms, with a specific focus on reinforcement learning. The three synthetic datasets described in this paper present patient cohorts with acute hypotension and sepsis in the intensive care unit, and people with human immunodeficiency virus (HIV) receiving antiretroviral therapy in ambulatory care. The datasets were created using a novel generative adversarial network (GAN). The distributions of variables, and correlations between variables and trends over time in the synthetic datasets mirror those in the real datasets. Furthermore, the risk of sensitive information disclosure associated with the public distribution of the synthetic datasets is estimated to be very low.
Talking head video generation aims to produce a synthetic human face video that contains the identity and pose information respectively from a given source image and a driving video.Existing works for this task heavily rely on 2D representations (e.g. appearance and motion) learned from the input images. However, dense 3D facial geometry (e.g. pixel-wise depth) is extremely important for this task as it is particularly beneficial for us to essentially generate accurate 3D face structures and distinguish noisy information from the possibly cluttered background. Nevertheless, dense 3D geometry annotations are prohibitively costly for videos and are typically not available for this video generation task. In this paper, we first introduce a self-supervised geometry learning method to automatically recover the dense 3D geometry (i.e.depth) from the face videos without the requirement of any expensive 3D annotation data. Based on the learned dense depth maps, we further propose to leverage them to estimate sparse facial keypoints that capture the critical movement of the human head. In a more dense way, the depth is also utilized to learn 3D-aware cross-modal (i.e. appearance and depth) attention to guide the generation of motion fields for warping source image representations. All these contributions compose a novel depth-aware generative adversarial network (DaGAN) for talking head generation. Extensive experiments conducted demonstrate that our proposed method can generate highly realistic faces, and achieve significant results on the unseen human faces.
Medical Ultrasound (US) is one of the most widely used imaging modalities in clinical practice, but its usage presents unique challenges such as variable imaging quality. Deep Learning (DL) models can serve as advanced medical US image analysis tools, but their performance is greatly limited by the scarcity of large datasets. To solve the common data shortage, we develop GSDA, a Generative Adversarial Network (GAN)-based semi-supervised data augmentation method. GSDA consists of the GAN and Convolutional Neural Network (CNN). The GAN synthesizes and pseudo-labels high-resolution, high-quality US images, and both real and synthesized images are then leveraged to train the CNN. To address the training challenges of both GAN and CNN with limited data, we employ transfer learning techniques during their training. We also introduce a novel evaluation standard that balances classification accuracy with computational time. We evaluate our method on the BUSI dataset and GSDA outperforms existing state-of-the-art methods. With the high-resolution and high-quality images synthesized, GSDA achieves a 97.9% accuracy using merely 780 images. Given these promising results, we believe that GSDA holds potential as an auxiliary tool for medical US analysis.
As Deep Learning algorithms continue to evolve and become more sophisticated, they require massive datasets for model training and efficacy of models. Some of those data requirements can be met with the help of existing datasets within the organizations. Current Machine Learning practices can be leveraged to generate synthetic data from an existing dataset. Further, it is well established that diversity in generated synthetic data relies on (and is perhaps limited by) statistical properties of available dataset within a single organization or entity. The more diverse an existing dataset is, the more expressive and generic synthetic data can be. However, given the scarcity of underlying data, it is challenging to collate big data in one organization. The diverse, non-overlapping dataset across distinct organizations provides an opportunity for them to contribute their limited distinct data to a larger pool that can be leveraged to further synthesize. Unfortunately, this raises data privacy concerns that some institutions may not be comfortable with. This paper proposes a novel approach to generate synthetic data - FedSyn. FedSyn is a collaborative, privacy preserving approach to generate synthetic data among multiple participants in a federated and collaborative network. FedSyn creates a synthetic data generation model, which can generate synthetic data consisting of statistical distribution of almost all the participants in the network. FedSyn does not require access to the data of an individual participant, hence protecting the privacy of participant's data. The proposed technique in this paper leverages federated machine learning and generative adversarial network (GAN) as neural network architecture for synthetic data generation. The proposed method can be extended to many machine learning problem classes in finance, health, governance, technology and many more.
Synthetic digital twins based on medical data accelerate the acquisition, labelling and decision making procedure in digital healthcare. A core part of digital healthcare twins is model-based data synthesis, which permits the generation of realistic medical signals without requiring to cope with the modelling complexity of anatomical and biochemical phenomena producing them in reality. Unfortunately, algorithms for cardiac data synthesis have been so far scarcely studied in the literature. An important imaging modality in the cardiac examination is three-directional CINE multi-slice myocardial velocity mapping (3Dir MVM), which provides a quantitative assessment of cardiac motion in three orthogonal directions of the left ventricle. The long acquisition time and complex acquisition produce make it more urgent to produce synthetic digital twins of this imaging modality. In this study, we propose a hybrid deep learning (HDL) network, especially for synthetic 3Dir MVM data. Our algorithm is featured by a hybrid UNet and a Generative Adversarial Network with a foreground-background generation scheme. The experimental results show that from temporally down-sampled magnitude CINE images (six times), our proposed algorithm can still successfully synthesise high temporal resolution 3Dir MVM CMR data (PSNR=42.32) with precise left ventricle segmentation (DICE=0.92). These performance scores indicate that our proposed HDL algorithm can be implemented in real-world digital twins for myocardial velocity mapping data simulation. To the best of our knowledge, this work is the first one in the literature investigating digital twins of the 3Dir MVM CMR, which has shown great potential for improving the efficiency of clinical studies via synthesised cardiac data.
This paper evaluates synthetically generated healthcare data for biases and investigates the effect of fairness mitigation techniques on utility-fairness. Privacy laws limit access to health data such as Electronic Medical Records (EMRs) to preserve patient privacy. Albeit essential, these laws hinder research reproducibility. Synthetic data is a viable solution that can enable access to data similar to real healthcare data without privacy risks. Healthcare datasets may have biases in which certain protected groups might experience worse outcomes than others. With the real data having biases, the fairness of synthetically generated health data comes into question. In this paper, we evaluate the fairness of models generated on two healthcare datasets for gender and race biases. We generate synthetic versions of the dataset using a Generative Adversarial Network called HealthGAN, and compare the real and synthetic model's balanced accuracy and fairness scores. We find that synthetic data has different fairness properties compared to real data and fairness mitigation techniques perform differently, highlighting that synthetic data is not bias free.
Deep generative models have demonstrated effectiveness in learning compact and expressive design representations that significantly improve geometric design optimization. However, these models do not consider the uncertainty introduced by manufacturing or fabrication. Past work that quantifies such uncertainty often makes simplified assumptions on geometric variations, while the "real-world" uncertainty and its impact on design performance are difficult to quantify due to the high dimensionality. To address this issue, we propose a Generative Adversarial Network-based Design under Uncertainty Framework (GAN-DUF), which contains a deep generative model that simultaneously learns a compact representation of nominal (ideal) designs and the conditional distribution of fabricated designs given any nominal design. We demonstrated the framework on two real-world engineering design examples and showed its capability of finding the solution that possesses better performances after fabrication.
This paper proposes a high-precision self-supervised monocular VO, which is specifically designed for navigation in foggy weather. A cycled generative adversarial network is designed to obtain high-quality self-supervised loss via forcing the forward and backward half-cycle to output consistent estimation. Moreover, gradient-based loss and perceptual loss are introduced to eliminate the interference of complex photometric change on self-supervised loss in foggy weather. To solve the ill-posed problem of depth estimation, a self-supervised multi-task learning aided depth estimation module is designed based on the strong correlation between the depth estimation and transmission map calculation of hazy images in foggy weather. The experimental results on the synthetic foggy KITTI dataset show that the proposed self-supervised monocular VO performs better in depth and pose estimation than other state-of-the-art monocular VO in the literature, indicating the designed method is more suitable for foggy weather.
Modeling biological soft tissue is complex in part due to material heterogeneity. Microstructural patterns, which play a major role in defining the mechanical behavior of these tissues, are both challenging to characterize, and difficult to simulate. Recently, machine learning-based methods to predict the mechanical behavior of heterogeneous materials have made it possible to more thoroughly explore the massive input parameter space associated with heterogeneous blocks of material. Specifically, we can train machine learning (ML) models to closely approximate computationally expensive heterogeneous material simulations where the ML model is trained on a dataset of simulations that capture the range of spatial heterogeneity present in the material of interest. However, when it comes to applying these techniques to biological tissue more broadly, there is a major limitation: the relevant microstructural patterns are both challenging to obtain and difficult to analyze. Consequently, the number of useful examples available to characterize the input domain under study is limited. In this work, we investigate the efficacy of ML-based generative models as well as procedural methods as a tool for augmenting limited input pattern datasets. We find that a Style-based Generative Adversarial Network with adaptive discriminator augmentation is able to successfully leverage just 1,000 example patterns to create the most authentic generated patterns. In general, diverse generated patterns with adequate resemblance to the real patterns can be used as inputs to finite element simulations to meaningfully augment the training dataset. To enable this methodological contribution, we have created an open access dataset of Finite Element Analysis simulations based on Cahn-Hilliard patterns. We anticipate that future researchers will be able to leverage this dataset and build on the work presented here.
Training a generative adversarial network (GAN) with limited data has been a challenging task. A feasible solution is to start with a GAN well-trained on a large scale source domain and adapt it to the target domain with a few samples, termed as few shot generative model adaption. However, existing methods are prone to model overfitting and collapse in extremely few shot setting (less than 10). To solve this problem, we propose a relaxed spatial structural alignment method to calibrate the target generative models during the adaption. We design a cross-domain spatial structural consistency loss comprising the self-correlation and disturbance correlation consistency loss. It helps align the spatial structural information between the synthesis image pairs of the source and target domains. To relax the cross-domain alignment, we compress the original latent space of generative models to a subspace. Image pairs generated from the subspace are pulled closer. Qualitative and quantitative experiments show that our method consistently surpasses the state-of-the-art methods in few shot setting.
Histopathological visualizations are a pillar of modern medicine and biological research. Surgical oncology relies exclusively on post-operative histology to determine definitive surgical success and guide adjuvant treatments. The current histology workflow is based on bright-field microscopic assessment of histochemical stained tissues and has some major limitations. For example, the preparation of stained specimens for brightfield assessment requires lengthy sample processing, delaying interventions for days or even weeks. Hence, there is a pressing need for improved histopathology methods. In this paper, we present a deep-learning-based approach for virtual label-free histochemical staining of total-absorption photoacoustic remote sensing (TA-PARS) images of unstained tissue. TA-PARS provides an array of directly measured label-free contrasts such as scattering and total absorption (radiative and non-radiative), ideal for developing H&E colorizations without the need to infer arbitrary tissue structures. We use a Pix2Pix generative adversarial network (GAN) to develop visualizations analogous to H&E staining from label-free TA-PARS images. Thin sections of human skin tissue were first virtually stained with the TA-PARS, then were chemically stained with H&E producing a one-to-one comparison between the virtual and chemical staining. The one-to-one matched virtually- and chemically- stained images exhibit high concordance validating the digital colorization of the TA-PARS images against the gold standard H&E. TA-PARS images were reviewed by four dermatologic pathologists who confirmed they are of diagnostic quality, and that resolution, contrast, and color permitted interpretation as if they were H&E. The presented approach paves the way for the development of TA-PARS slide-free histology, which promises to dramatically reduce the time from specimen resection to histological imaging.
Automatic detection of parasitic eggs in microscopy images has the potential to increase the efficiency of human experts whilst also providing an objective assessment. The time saved by such a process would both help ensure a prompt treatment to patients, and off-load excessive work from experts' shoulders. Advances in deep learning inspired us to exploit successful architectures for detection, adapting them to tackle a different domain. We propose a framework that exploits two such state-of-the-art models. Specifically, we demonstrate results produced by both a Generative Adversarial Network (GAN) and Faster-RCNN, for image enhancement and object detection respectively, on microscopy images of varying quality. The use of these techniques yields encouraging results, though further improvements are still needed for certain egg types whose detection still proves challenging. As a result, a new dataset has been created and made publicly available, providing an even wider range of classes and variability.
Cyberattacks from within an organization's trusted entities are known as insider threats. Anomaly detection using deep learning requires comprehensive data, but insider threat data is not readily available due to confidentiality concerns of organizations. Therefore, there arises demand to generate synthetic data to explore enhanced approaches for threat analysis. We propose a linear manifold learning-based generative adversarial network, SPCAGAN, that takes input from heterogeneous data sources and adds a novel loss function to train the generator to produce high-quality data that closely resembles the original data distribution. Furthermore, we introduce a deep learning-based hybrid model for insider threat analysis. We provide extensive experiments for data synthesis, anomaly detection, adversarial robustness, and synthetic data quality analysis using benchmark datasets. In this context, empirical comparisons show that GAN-based oversampling is competitive with numerous typical oversampling regimes. For synthetic data generation, our SPCAGAN model overcame the problem of mode collapse and converged faster than previous GAN models. Results demonstrate that our proposed approach has a lower error, is more accurate, and generates substantially superior synthetic insider threat data than previous models.
Spoken language change detection (LCD) refers to detecting language switching points in a multilingual speech signal. Speaker change detection (SCD) refers to locating the speaker change points in a multispeaker speech signal. The objective of this work is to understand the challenges in LCD task by comparing it with SCD task. Human subjective study for change detection is performed for LCD and SCD. This study demonstrates that LCD requires larger duration spectro-temporal information around the change point compared to SCD. Based on this, the work explores automatic distance based and model based LCD approaches. The model based ones include Gaussian mixture model and universal background model (GMM-UBM), attention, and Generative adversarial network (GAN) based approaches. Both the human and automatic LCD tasks infer that the performance of the LCD task improves by incorporating more and more spectro-temporal duration.
The estimation of loss distributions for dynamic portfolios requires the simulation of scenarios representing realistic joint dynamics of their components. We propose a novel data-driven approach for simulating realistic, high-dimensional multi-asset scenarios, focusing on accurately representing tail risk for a class of static and dynamic trading strategies. We exploit the joint elicitability property of Value-at-Risk (VaR) and Expected Shortfall (ES) to design a Generative Adversarial Network (GAN) that learns to simulate price scenarios preserving these tail risk features. We demonstrate the performance of our algorithm on synthetic and market data sets through detailed numerical experiments. In contrast to previously proposed data-driven scenario generators, our proposed method correctly captures tail risk for a broad class of trading strategies and demonstrates strong generalization capabilities. In addition, combining our method with principal component analysis of the input data enhances its scalability to large-dimensional multi-asset time series, setting our framework apart from the univariate settings commonly considered in the literature.
A powerful approach, and one of the most common ones in structural health monitoring (SHM), is to use data-driven models to make predictions and inferences about structures and their condition. Such methods almost exclusively rely on the quality of the data. Within the SHM discipline, data do not always suffice to build models with satisfactory accuracy for given tasks. Even worse, data may be completely missing from one's dataset, regarding the behaviour of a structure under different environmental conditions. In the current work, with a view to confronting such issues, the generation of artificial data using a variation of the generative adversarial network (GAN) algorithm, is used. The aforementioned variation is that of the conditional GAN or cGAN. The algorithm is not only used to generate artificial data, but also to learn transformations of manifolds according to some known parameters. Assuming that the structure's response is represented by points in a manifold, part of the space will be formed due to variations in external conditions affecting the structure. This idea proves efficient in SHM, as it is exploited to generate structural data for specific values of environmental coefficients. The scheme is applied here on a simulated structure which operates under different temperature and humidity conditions. The cGAN is trained on data for some discrete values of the temperature within some range, and is able to generate data for every temperature in this range with satisfactory accuracy. The novelty, compared to classic regression in similar problems, is that the cGAN allows unknown environmental parameters to affect the structure and can generate whole manifolds of data for every value of the known parameters, while the unknown ones vary within the generated manifolds.
Generative adversarial network (GAN) is widely used for generalized and robust learning on graph data. However, for non-Euclidean graph data, the existing GAN-based graph representation methods generate negative samples by random walk or traverse in discrete space, leading to the information loss of topological properties (e.g. hierarchy and circularity). Moreover, due to the topological heterogeneity (i.e., different densities across the graph structure) of graph data, they suffer from serious topological distortion problems. In this paper, we proposed a novel Curvature Graph Generative Adversarial Networks method, named \textbf{\modelname}, which is the first GAN-based graph representation method in the Riemannian geometric manifold. To better preserve the topological properties, we approximate the discrete structure as a continuous Riemannian geometric manifold and generate negative samples efficiently from the wrapped normal distribution. To deal with the topological heterogeneity, we leverage the Ricci curvature for local structures with different topological properties, obtaining to low-distortion representations. Extensive experiments show that CurvGAN consistently and significantly outperforms the state-of-the-art methods across multiple tasks and shows superior robustness and generalization.
Linear modal analysis is a useful and effective tool for the design and analysis of structures. However, a comprehensive basis for nonlinear modal analysis remains to be developed. In the current work, a machine learning scheme is proposed with a view to performing nonlinear modal analysis. The scheme is focussed on defining a one-to-one mapping from a latent `modal' space to the natural coordinate space, whilst also imposing orthogonality of the mode shapes. The mapping is achieved via the use of the recently-developed cycle-consistent generative adversarial network (cycle-GAN) and an assembly of neural networks targeted on maintaining the desired orthogonality. The method is tested on simulated data from structures with cubic nonlinearities and different numbers of degrees of freedom, and also on data from an experimental three-degree-of-freedom set-up with a column-bumper nonlinearity. The results reveal the method's efficiency in separating the `modes'. The method also provides a nonlinear superposition function, which in most cases has very good accuracy.