A Framework for Generating Realistic Synthetic Tabular Data in a Randomized Controlled Trial Setting.

Preserving logical and functional dependencies in synthetic tabular data

High-quality tabular data is a crucial requirement for developing data-driven applications, especially healthcare-related ones, because most of the data nowadays collected in this context is in tabular form. However, strict data protection laws complicates the access to medical datasets. Thus, synthetic data has become an ideal alternative for data scientists and healthcare professionals to circumvent such hurdles. Although many healthcare institutions still use the classical de-identification and anonymization techniques for generating synthetic data, deep learning-based generative models such as generative adversarial networks (GANs) have shown a remarkable performance in generating tabular datasets with complex structures. This paper examines the GANs’ potential and applicability within the healthcare industry, which often faces serious challenges with insufficient training data and patient records sensitivity. We investigate several state-of-the-art GAN-based models proposed for tabular synthetic data generation. Healthcare datasets with different sizes, numbers of variables, column data types, feature distributions, and inter-variable correlations are examined. Moreover, a comprehensive evaluation framework is defined to evaluate the quality of the synthetic records and the viability of each model in preserving the patients’ privacy. The results indicate that the proposed models can generate synthetic datasets that maintain the statistical characteristics, model compatibility and privacy of the original data. Moreover, synthetic tabular healthcare datasets can be a viable option in many data-driven applications. However, there is still room for further improvements in designing a perfect architecture for generating synthetic tabular data.

<p><strong>Because of its wide range of applications, generative artificial intelligence (Generative AI) has received a lot of attention in academia and industry. Despite significant advances in computer vision and natural language processing, the use of Generative AI in tabular data is still underexplored. This gap is especially significant given the prevalence of tabular data as the primary data modality. This thesis seeks to close this gap by focusing on the efficient generation, evaluation, refinement, and application of tabular data, while addressing the challenges inherent in its heterogeneous nature.</strong></p><p>The primary goal of this thesis is to develop efficient algorithms for tabular data synthesis, advancing the field of Generative AI in the tabular domain. To achieve this goal, four specific research objectives have been outlined.</p><p>First, this thesis addresses gaps in evaluation metrics by unifying a framework for the comprehensive and consistent assessment of synthetic data. The proposed framework is designed to meet diverse business requirements across various downstream tasks by incorporating a wide range of advanced metrics. These metrics cover various types of evaluations, including univariate, bivariate, multivariate, cluster, and record-level evaluations. Additionally, standardized visualizations are provided to facilitate qualitative assessments. The results indicate that the proposed evaluation framework not only enables consistent comparisons, rankings, and the selection of different data synthesis approaches but also acts as a valuable tool to promptly assess the reliability of results obtained from synthetic data.</p><p>Second, this thesis presents innovative algorithms for the generation of tabular data, aiming to significantly enhance the quality of data synthesis. One of the contributions is the development of a reversible feature engineering pipeline designed to automatically represent tabular data in an efficient format while also ensuring that the transformed data can be easily converted back to its original format. Additionally, the thesis proposes novel deep learning-based tabular data generation models that are capable of learning the joint distribution of multivariate datasets without relying on predefined distribution assumptions. Experimental results highlight the effectiveness of the proposed algorithms in handling heterogeneous data, demonstrating superior performance in most scenarios within the same training duration when compared to similar alternatives.</p><p>Third, this thesis introduces innovative synthetic data prototype selection algorithms with the goal of refining the generated samples. This approach aims to leverage the advantages of Deep Generative Models, which, once trained, can generate unlimited and diverse synthetic data. By carefully identifying and selecting high-quality samples or removing unrealistic ones, the quality of the synthetic data can be enhanced from a post-processing perspective. Building on this hypothesis, we recognize the iterative nature inherent in the data synthesis procedure, where the processes of data generation, evaluation, and refinement should operate repeatedly in a cyclical flow. During the data generation phase, the evaluation framework assists in identifying potential issues and risks based on use cases, while the refinement (post-processing) step iteratively improves synthetic data in alignment with the evaluation outcomes.</p><p>Last, this thesis applies and validates the proposed models across various domains, including business, healthcare, and government, each requiring distinct downstream tasks. Specifically, we assess our algorithms from the standpoint of data balancing using churn data, evaluate our models focusing on data augmentation with health data, and test our algorithms from the perspective of data representation considering data privacy concerns within sensitive citizen data. These applications serve as practical illustrations, demonstrating the effectiveness and utility of the proposed Generative AI models in real-world scenarios.</p><p>In summary, this thesis makes a substantial contribution to the advancement of Generative AI within the domain of tabular data. It not only presents innovative algorithms and evaluation methods but also introduces practical frameworks applicable to real-world scenarios. The findings demonstrate the considerable capacity of Generative AI to fundamentally transform tabular data in various fields, ultimately leading to improved data availability, quality, and quantity.</p>

Infrared (IR) images are essential to improve the visibility of dark or camouflaged objects. Object recognition and segmentation based on a neural network using IR images provide more accuracy and insight than color visible images. But the bottleneck is the amount of relevant IR images for training. It is difficult to collect real-world IR images for special purposes, including space exploration, military and fire-fighting applications. To solve this problem, we created color visible and IR images using a Unity-based 3D game editor. These synthetically generated color visible and IR images were used to train cycle consistent adversarial networks (CycleGAN) to convert visible images to IR images. CycleGAN has the advantage that it does not require precisely matching visible and IR pairs for transformation training. In this study, we discovered that additional synthetic data can help improve CycleGAN performance. Neural network training using real data (N = 20) performed more accurate transformations than training using real (N = 10) and synthetic (N = 10) data combinations. The result indicates that the synthetic data cannot exceed the quality of the real data. Neural network training using real (N = 10) and synthetic (N = 100) data combinations showed almost the same performance as training using real data (N = 20). At least 10 times more synthetic data than real data is required to achieve the same performance. In summary, CycleGAN is used with synthetic data to improve the IR image conversion performance of visible images.

Accurately modelled computer-generated data can be used in place of real-world signals for the design, test and validation of signal processing techniques in situations where real data is difficult to obtain. Bio-signal processing researchers interested in working with fNIRS data are restricted due to the lack of freely available fNIRS data and by the prohibitively expensive cost of fNIRS systems. We present a simplified mathematical description and associated MATLAB implementation of model-based synthetic fNIRS data which could be used by researchers to develop fNIRS signal processing techniques. The software, which is freely available, allows users to generate fNIRS data with control over a wide range of parameters and allows for fine-tuning of the synthetic data. We demonstrate how the model can be used to generate raw fNIRS data similar to recorded fNIRS signals. Signal processing steps were then applied to both the real and synthetic data. Visual comparisons between the temporal and spectral properties of the real and synthetic data show similarity. This paper demonstrates that our model for generating synthetic fNIRS data can replicate real fNIRS recordings.

Deep learning currently struggles with tabular data, but it can benefit from multimodal learning. SAINT is a deep learning model for tabular data on which we base our presented developments. In this study, we introduce SAINTENS as a new deep learning method, specifically for the in healthcare predominant tabular real world data. For this purpose, we compare SAINTENS with SAINT and the State of the Art Machine Learning methods for tabular data. We use tabular data from geriatrics to predict four different targets (dysphagia, pressure ulcers, decompensated heart failure and delirium). We determine the relevant feature sets and train the models on these sets. Both SAINTENS and SAINT models are at least on the same performance level as the current State of the Art (Gradient Boosting Decision Trees). In combination with multimodal learning SAINTENS and SAINT may be used on real world data comprising tabular, text and image data, for discovery and development of new digital biomarkers.

The purpose of the European Stroke Organisation-Karolinska Stroke Update Conference is to provide updates on recent stroke therapy research and to give an opportunity for the participants to discuss how these results may be implemented into clinical routine. The meeting started 22 years ago as Karolinska Stroke Update, but since 2014 it is a joint conference with European Stroke Organisation. Importantly, it provides a platform for discussion on the European Stroke Organisation guidelines process and on recommendations to the European Stroke Organisation guidelines committee on specific topics. By this, it adds a direct influence from stroke professionals otherwise not involved in committees and work groups on the guideline procedure. The discussions at the conference may also inspire new guidelines when motivated. The topics raised at the meeting are selected by the scientific programme committee mainly based on recent important scientific publications. This year's European Stroke Organisation-Karolinska Stroke Update Meeting was held in Stockholm on 11-13 November 2018. There were 11 scientific sessions discussed in the meeting including two short sessions. Each session except the short sessions produced a consensus statement (Full version with background, issues, conclusions and references are published as web-material and at www.eso-karolinska.org and http://eso-stroke.org) and recommendations which were prepared by a writing committee consisting of session chair(s), scientific secretary and speakers. These statements were presented to the 250 participants of the meeting. In the open meeting, general participants commented on the consensus statement and recommendations and the final document were adjusted based on the discussion from the general participants Recommendations (grade of evidence) were graded according to the 1998 Karolinska Stroke Update meeting with regard to the strength of evidence. Grade A Evidence: Strong support from randomised controlled trials and statistical reviews (at least one randomised controlled trial plus one statistical review). Grade B Evidence: Support from randomised controlled trials and statistical reviews (one randomised controlled trial or one statistical review). Grade C Evidence: No reasonable support from randomised controlled trials, recommendations based on small randomised and/or non-randomised controlled trials evidence.

Deep learning (DL) applications of seismic reservoir characterization often require the generation of synthetic data to augment available sparse labeled data. An approach for generating synthetic training data consists of specifying probability distributions modeling prior geologic uncertainty on reservoir properties and forward modeling the seismic data. A prior falsification approach is critical to establish the consistency of the synthetic training data distribution with real seismic data. With the help of a real case study of facies classification with convolutional neural networks (CNNs) from an offshore deltaic reservoir, we have highlighted several practical nuances associated with training DL models on synthetic seismic data. We highlight the issue of overfitting of CNNs to the synthetic training data distribution and propose regularization strategies to address it. We demonstrate the efficacy of our proposed strategies by training the CNN on synthetic data and making robust predictions with real 3D partial stack seismic data.

Generation of synthetic full-scale burst test data for corroded pipelines using the tabular generative adversarial network

Synthetic data, generated using generative AI techniques, closely mimics the characteristics of real data while enhancing privacy for sensitive health data. This study evaluates synthetic tabular data based on fidelity and utility for predictive models. Fidelity is measured through univariate distribution and bivariate differential pairwise correlations, while utility is measured by comparing machine learning model performance trained on synthetic and real data. Results show highly similar model performance on synthetic and real data. We also explore the potential of using synthetic data for hyperparameter tuning. Our findings reveal a strong correlation between prediction accuracy on synthetic and real data, suggesting that hyperparameters optimized using synthetic data can be effectively applied to models trained on real datasets for optimal results.

Generation of synthetic data with low-dimensional features for condition monitoring utilizing Generative Adversarial Networks

The use of electronic health record (EHR) tabular data for clustering has garnered global attention for its potential to uncover meaningful clusters, improve patient care and advance healthcare research. Applications of EHR data extend beyond healthcare to include financial modeling, fraud detection, differential diagnosis and rare disease identification. However, critical challenges persist, including privacy concerns, limited access to real EHR data, underexplored generative methods and inadequate validation frameworks. This study aims to generate and validate synthetic EHR tabular data and demonstrate its utility in clustering, addressing existing gaps in privacy, data quality and validation. Advanced deep learning architectures, including Generative Adversarial Networks (GANs), were developed to generate high-quality synthetic datasets. We implemented robust validation frameworks that incorporate fidelity, utility and privacy-preserving mechanisms. Dirichlet Process Mixture Models (DPMMs) used synthetic data to group similar items and measured the differences between real and synthetic data distributions by estimating divergence. GANs effectively replicated complex datasets while balancing fidelity, utility and privacy. Validation confirmed the robustness of DPMMs in identifying latent structures. Cluster analysis revealed significant multimorbidity patterns, facilitating personalized treatment strategies, resource allocation and preventive care. This study demonstrates the transformative potential of synthetic EHR data in clustering, advancing healthcare delivery, planning and research.

Genome-wide association studies (GWAS) have served as primary methods for the past decade for identifying associations between genetic variants and traits or diseases. Many software packages have been developed for GWAS analysis based on different statistical models. One key factor influencing the statistical reliability of GWAS is the amount of input data used. Few studies have been conducted to investigate this effect by comparing the performance of GWAS programs using varied amounts of experimental data, especially in the context of plants and plant genomes. In this paper, we investigate how input data quantity influences output of four widely used GWAS programs, PLINK, TASSEL, GAPIT, and FaST-LMM. Both synthetic and real data are used. Standard GWAS output includes single nucleotide polymorphisms (SNPs) and their p-values. To evaluate the programs, p-values and q-values of SNPs, and Kendall rank correlation between output SNP lists, are used. Results show that with the same GWAS program, different Arabidopsis thaliana datasets demonstrate similar trends of rank correlation with varied input quantity, but differentiate on the numbers of SNPs passing a given p- or q-value threshold. In practice, experimental datasets may have samples containing varied numbers of biological replicates. We show that this variation in replicates influences the p-values of SNPs, but does not strongly affect the rank correlation. When comparing synthetic and real data, the output SNPs from synthetic data have similar rank correlation trends across all four GWAS programs, but the same measure from real data is diverse across the programs. In addition, the real data results in a linear-like increase in the numbers of significant SNPs with more input data, but the synthetic data does not follow this trend. This study provides guidance on selecting GWAS programs when varied experimental data is present and on selecting significant SNPs for subsequent study. It contributes to understanding how much input data is necessary to yield satisfying GWAS results.

Recently, an increasing number of Artificial Intelligence services have been developed for a variety of domains. Machine Learning and especially Deep Learning services require a large amount of data to provide their functionality. Since data collection is typically complex and difficult, there is often not enough data available. Machine learning services such as anomaly detection or disaggregation algorithms are also being developed in the smart living domain. In practice, however, only a few energy datasets are publicly available, as the collection of such data is expensive and time-consuming due to the equipment required. One way to generate more smart meter data is to use a simulation. Developing such a simulation that is capable of generating meaningful data is a complex task. Therefore, in this paper, we present the Synthetic Time Series Data Generator (SynTiSeD), a multi-agent-based simulation tool that generates meaningful synthetic energy data based on real-world data. Furthermore, SynTiSeD allows generating data of critical situations, which are important for the development of such services, but which cannot be provoked in the real world. For transferability, we demonstrate that Nonintrusive Load Monitoring algorithms trained on synthetic data generated by SynTiSeD provide meaningful results that are even better than those of models trained on real data.

Does simulation really increase gynecologic surgical skill?