Synthetic data generation methods in healthcare: A review on open-source tools and methods.

Synthetic data generation has emerged as a promising solution to overcome the challenges which are posed by data scarcity and privacy concerns, as well as, to address the need for training artificial intelligence (AI) algorithms on unbiased data with sufficient sample size and statistical power. Our review explores the application and efficacy of synthetic data methods in healthcare considering the diversity of medical data. To this end, we systematically searched the PubMed and Scopus databases with a great focus on tabular, imaging, radiomics, time-series, and omics data. Studies involving multi-modal synthetic data generation were also explored. The type of method used for the synthetic data generation process was identified in each study and was categorized into statistical, probabilistic, machine learning, and deep learning. Emphasis was given to the programming languages used for the implementation of each method. Our evaluation revealed that the majority of the studies utilize synthetic data generators to: (i) reduce the cost and time required for clinical trials for rare diseases and conditions, (ii) enhance the predictive power of AI models in personalized medicine, (iii) ensure the delivery of fair treatment recommendations across diverse patient populations, and (iv) enable researchers to access high-quality, representative multimodal datasets without exposing sensitive patient information, among others. We underline the wide use of deep learning based synthetic data generators in 72.6 % of the included studies, with 75.3 % of the generators being implemented in Python. A thorough documentation of open-source repositories is finally provided to accelerate research in the field.

ProLesA-Net: A multi-channel 3D architecture for prostate MRI lesion segmentation with multi-scale channel and spatial attentions.

Prostate cancer diagnosis and treatment relies on precise MRI lesion segmentation, a challenge notably for small (<15 mm) and intermediate (15-30 mm) lesions. Our study introduces ProLesA-Net, a multi-channel 3D deep-learning architecture with multi-scale squeeze and excitation and attention gate mechanisms. Tested against six models across two datasets, ProLesA-Net significantly outperformed in key metrics: Dice score increased by 2.2%, and Hausdorff distance and average surface distance improved by 0.5 mm, with recall and precision also undergoing enhancements. Specifically, for lesions under 15 mm, our model showed a notable increase in five key metrics. In summary, ProLesA-Net consistently ranked at the top, demonstrating enhanced performance and stability. This advancement addresses crucial challenges in prostate lesion segmentation, enhancing clinical decision making and expediting treatment processes.

Multimodal automatic assessment of acute pain through facial videos and heart rate signals utilizing transformer-based architectures.

Accurate and objective pain evaluation is crucial in developing effective pain management protocols, aiming to alleviate distress and prevent patients from experiencing decreased functionality. A multimodal automatic assessment framework for acute pain utilizing video and heart rate signals is introduced in this study. The proposed framework comprises four pivotal modules: the Spatial Module, responsible for extracting embeddings from videos; the Heart Rate Encoder, tasked with mapping heart rate signals into a higher dimensional space; the AugmNet, designed to create learning-based augmentations in the latent space; and the Temporal Module, which utilizes the extracted video and heart rate embeddings for the final assessment. The Spatial-Module undergoes pre-training on a two-stage strategy: first, with a face recognition objective learning universal facial features, and second, with an emotion recognition objective in a multitask learning approach, enabling the extraction of high-quality embeddings for the automatic pain assessment. Experiments with the facial videos and heart rate extracted from electrocardiograms of the BioVid database, along with a direct comparison to 29 studies, demonstrate state-of-the-art performances in unimodal and multimodal settings, maintaining high efficiency. Within the multimodal context, 82.74% and 39.77% accuracy were achieved for the binary and multi-level pain classification task, respectively, utilizing 9.62 million parameters for the entire framework.

Weight Estimation of Mediterranean Food Images using Random Forest Regression Algorithm.

The daily nutrition management is one of the most important issues that concern individuals in the modern lifestyle. Over the years, the development of dietary assessment systems and applications based on food images has assisted experts to manage people's nutritional facts and eating habits. In these systems, the food volume estimation is the most important task for calculating food quantity and nutritional information. In this study, we present a novel methodology for food weight estimation based on a food image, using the Random Forest regression algorithm. The weight estimation model was trained on a unique dataset of 5,177 annotated Mediterranean food images, consisting of 50 different foods with a reference card placed next to the plate. Then, we created a data frame of 6,425 records from the annotated food images with features such as: food area, reference object area, food id, food category and food weight. Finally, using the Random Forest regression algorithm and applying nested cross validation and hyperparameters tuning, we trained the weight estimation model. The proposed model achieves 22.6 grams average difference between predicted and real weight values for each food item record in the data frame and 15.1% mean absolute percentage error for each food item, opening new perspectives in food image-based volume and nutrition estimation models and systems.Clinical Relevance- The proposed methodology is suitable for healthcare systems and applications that monitor an individual's malnutrition, offering the ability to estimate the energy and nutrients consumed using an image of the meal.

Boosting the performance of MALT lymphoma classification in patients with primary Sjögren's Syndrome through data augmentation: a case study.

Sjögren's Syndrome (SS) patients with mucosa associated lymphoid tissue lymphomas (MALTLs) and diffuse large B-cell lymphomas (DLBCLs) have 10-year survival rates of 80% and 40%, respectively. This highlights the unique biologic burden of the two histologic forms, as well as, the need for early detection and thorough monitoring of these patients. The lack of MALTL patients and the fact that most studies are single cohort and combine patients with different lymphoma subtypes narrow the understanding of MALTL progression. Here, we propose a data augmentation pipeline that utilizes an advanced synthetic data generator which is trained on a Pan European data hub with primary SS (pSS) patients to yield a high-quality synthetic data pool. The latter is used for the development of an enhanced MALTL classification model. Four scenarios were defined to assess the reliability of augmentation. Our results revealed an overall improvement in the accuracy, sensitivity, specificity, and AUC by 7%, 6.3%, 9%, and 6.3%, respectively. This is the first case study that utilizes data augmentation to reflect the progression of MALTL in pSS.

eSEE-d: Emotional State Estimation Based on Eye-Tracking Dataset.

Affective state estimation is a research field that has gained increased attention from the research community in the last decade. Two of the main catalysts for this are the advancement in the data analysis using artificial intelligence and the availability of high-quality video. Unfortunately, benchmarks and public datasets are limited, thus making the development of new methodologies and the implementation of comparative studies essential. The current work presents the eSEE-d database, which is a resource to be used for emotional State Estimation based on Eye-tracking data. Eye movements of 48 participants were recorded as they watched 10 emotion-evoking videos, each of them followed by a neutral video. Participants rated four emotions (tenderness, anger, disgust, sadness) on a scale from 0 to 10, which was later translated in terms of emotional arousal and valence levels. Furthermore, each participant filled three self-assessment questionnaires. An extensive analysis of the participants' answers to the questionnaires' self-assessment scores as well as their ratings during the experiments is presented. Moreover, eye and gaze features were extracted from the low-level eye-recorded metrics, and their correlations with the participants' ratings are investigated. Finally, we take on the challenge to classify arousal and valence levels based solely on eye and gaze features, leading to promising results. In particular, the Deep Multilayer Perceptron (DMLP) network we developed achieved an accuracy of 92% in distinguishing positive valence from non-positive and 81% in distinguishing low arousal from medium arousal. The dataset is made publicly available.

Cleverballoon: An integrated approach for developing a drug-coated balloon with everolimus.

Coronary artery disease (CAD) is a chronic disease associated with high mortality and morbidity. Although treatment with drug-eluting stents is the most frequent interventional approach for coronary artery disease, drug-coated balloons (DCBs) constitute an innovative alternative, especially in the presence of certain anatomical conditions in the local coronary vasculature. DCBs allow the fast and homogenous transfer of drugs into the arterial wall, during the balloon inflation. Their use has been established for treating in-stent restenosis caused by stent implantation, while recent clinical trials have shown a satisfactory efficacy in de novo small-vessel disease. Several factors affect DCBs performance including the catheter design, the drug dose and formulation. Cleverballoon focuses on the design and development of an innovative DCB with everolimus. For the realization of the development of this new DCB, an integrated approach, including in- vivo, in-vitro studies and in-silico modelling towards the DCB optimization, is presented.Clinical Relevance-The proposed study introduces the integration of in- vivo, in-vitro and in silico approaches in the design and development process of a new DCB, following the principles of 3R's for the replacement, reduction, and refinement of animal and clinical studies.

Bayesian Inference-Based Gaussian Mixture Models With Optimal Components Estimation Towards Large-Scale Synthetic Data Generation for In Silico Clinical Trials.

Goal: To develop a computationally efficient and unbiased synthetic data generator for large-scale in silico clinical trials (CTs). Methods: We propose the BGMM-OCE, an extension of the conventional BGMM (Bayesian Gaussian Mixture Models) algorithm to provide unbiased estimations regarding the optimal number of Gaussian components and yield high-quality, large-scale synthetic data at reduced computational complexity. Spectral clustering with efficient eigenvalue decomposition is applied to estimate the hyperparameters of the generator. A case study is conducted to compare the performance of BGMM-OCE against four straightforward synthetic data generators for in silico CTs in hypertrophic cardiomyopathy (HCM). Results: The BGMM-OCE generated 30000 virtual patient profiles having the lowest coefficient-of-variation (0.046), inter- and intra-correlation differences (0.017, and 0.016, respectively) with the real ones in reduced execution time. Conclusions: BGMM-OCE overcomes the lack of population size in HCM which obscures the development of targeted therapies and robust risk stratification models.

A "smart" Imputation Approach for Effective Quality Control Across Complex Clinical Data Structures.

The overwhelming need to improve the quality of complex data structures in healthcare is more important than ever. Although data quality has been the point of interest in many studies, none of them has focused on the development of quantitative and explainable methods for data imputation. In this work, we propose a "smart" imputation workflow to address missing data across complex data structures in the context of in silico clinical trials. AI algorithms were utilized to produce high-quality virtual patient profiles. A search algorithm was then developed to extract the best virtual patient profiles through the definition of a profile matching score (PMS). A case study was conducted, where the real dataset was randomly contaminated with multiple missing values (e.g., 10 to 50%). In total, 10000 virtual patient profiles with less than 0.02 Kullback-Leibler (KL) divergence were produced to estimate the PMS distribution. The best generator achieved the lowest average squared absolute difference (0.4) and average correlation difference (0.02) with the real dataset highlighting its increased effectiveness for data imputation across complex clinical data structures.

COLET: A dataset for COgnitive workLoad estimation based on eye-tracking.

BACKGROUND AND OBJECTIVE: The cognitive workload is an important component in performance psychology, ergonomics, and human factors. Publicly available datasets are scarce, making it difficult to establish new approaches and comparative studies. In this work, COLET-COgnitive workLoad estimation based on Eye-Tracking dataset is presented. METHODS: Forty-seven (47) individuals' eye movements were monitored as they solved puzzles involving visual search activities of varying complexity and duration. The participants' cognitive workload level was evaluated with the subjective test of NASA-TLX and this score is used as an annotation of the activity. Extensive data analysis was performed in order to derive eye and gaze features from low-level eye recorded metrics, and a range of machine learning models were evaluated and tested regarding the estimation of the cognitive workload level. RESULTS: The activities induced four different levels of cognitive workload. Multi tasking and time pressure have induced a higher level of cognitive workload than the one induced by single tasking and absence of time pressure. Multi tasking had a significant effect on 17 eye features while time pressure had a significant effect on 7 eye features. Both binary and multi-class identification attempts were performed by testing a variety of well-known classifiers, resulting in encouraging results towards cognitive workload levels estimation, with up to 88% correct predictions between low and high cognitive workload. CONCLUSIONS: Machine learning analysis demonstrated potential in discriminating cognitive workload levels using only eye-tracking characteristics. The proposed dataset includes a much higher sample size and a wider spectrum of eye and gaze metrics than other similar datasets, allowing for the examination of their relations with various cognitive states.

A User-Centric approach for Personalization based on Human Activity Recognition.

The objective of this work focuses on multiple independent user profiles that capture behavioral, emotional, medical, and physical patterns in the working and living environment resulting in one general user profile. Depending on the user's current activity (e.g. walking, eating, etc.), medical history, and other influential factors, the developed framework acts as a supplemental assistant to the user by providing not only the ability to enable supportive functionalities (e.g. image filtering, magnification, etc.) but also informative recommendations (e.g. diet, alcohol, etc.). The personalization of such a profile lies within the user's past preferences using human activity recognition as a base, and it is achieved through a statistical model, the Bayesian belief network. Training and real-time methodological pipelines are introduced and validated. The employed deep learning techniques for identifying human activities are presented and validated in publicly available and in-house datasets. The overall accuracy of human activity recognition reaches up to 86.96 %.

Segmentation of left atrium using CT images and a deep learning model.

The left atrium (LA) is one of the cardiac cavities with the most complex anatomical structures. Its role in the clinical diagnosis and patient's management is critical, as it is responsible for the atrial fibrillation, a condition that promotes the thrombogenesis inside the left atrial appendage. The development of an automated approach for LA segmentation is a demanding task mainly due to its anatomical complexity and the variation of its shape among patients. In this study, we focus to develop an unbiased pipeline capable to segment the atrial cavity from CT images. For evaluation purposes state-of-the-art metrics were used to assess the segmentation results. Particularly, the results indicated the mean values of the dice score 80%, the hausdorff distance 11.78mm, the average surface distance 2.24mm and the rand error index 0.2.

Investigation of crimping effects on the stent deployment through in silico modeling.

Atherosclerosis is one of the most mortal diseases that affects the arterial vessels, due to accumulation of plaque, altering the hemodynamic environment of the artery by preventing the sufficient delivery of blood to other organs. Stents are expandable tubular wires, used as a treatment option. In silico studies have been extensively exploited towards examining the performance of such devices by employing Finite Element Modeling. This study models the crimping stage during stent implantation to examine the effect of inclusion of pre-stress state of the stent. The results show that modeling of the crimping stress state of the stent prior to the deployment results in under-expansion of the stent, due to the indirect inclusion of strain-induced hardening effects. As a result, it is evident that the compressive stent stress configuration is important to be considered in the computational modeling approaches of stent deployment.

Disease Progression of Hypertrophic Cardiomyopathy: Modeling Using Machine Learning.

BACKGROUND: Cardiovascular disorders in general are responsible for 30% of deaths worldwide. Among them, hypertrophic cardiomyopathy (HCM) is a genetic cardiac disease that is present in about 1 of 500 young adults and can cause sudden cardiac death (SCD). OBJECTIVE: Although the current state-of-the-art methods model the risk of SCD for patients, to the best of our knowledge, no methods are available for modeling the patient's clinical status up to 10 years ahead. In this paper, we propose a novel machine learning (ML)-based tool for predicting disease progression for patients diagnosed with HCM in terms of adverse remodeling of the heart during a 10-year period. METHODS: The method consisted of 6 predictive regression models that independently predict future values of 6 clinical characteristics: left atrial size, left atrial volume, left ventricular ejection fraction, New York Heart Association functional classification, left ventricular internal diastolic diameter, and left ventricular internal systolic diameter. We supplemented each prediction with the explanation that is generated using the Shapley additive explanation method. RESULTS: The final experiments showed that predictive error is lower on 5 of the 6 constructed models in comparison to experts (on average, by 0.34) or a consortium of experts (on average, by 0.22). The experiments revealed that semisupervised learning and the artificial data from virtual patients help improve predictive accuracies. The best-performing random forest model improved R2 from 0.3 to 0.6. CONCLUSIONS: By engaging medical experts to provide interpretation and validation of the results, we determined the models' favorable performance compared to the performance of experts for 5 of 6 targets.

Real Time Human Activity Recognition Using Acceleration and First-Person Camera data.

The aim of this work is to present an automated method, working in real time, for human activity recognition based on acceleration and first-person camera data. A Long-Short-Term-Memory (LSTM) model has been built for recognizing locomotive activities (i.e. walking, sitting, standing, going upstairs, going downstairs) from acceleration data, while a ResNet model is employed for the recognition of stationary activities (i.e. eating, reading, writing, watching TV working on PC). The outcomes of the two models are fused in order for the final decision, regarding the performed activity, to be made. For the training, testing and evaluation of the proposed models, a publicly available dataset and an "in-house" dataset are utilized. The overall accuracy of the proposed algorithmic pipeline reaches 87.8%.

Gait and balance patterns related to Free-Walking and TUG tests in Parkinson's Disease based on plantar pressure data.

Continuous monitoring of patients with Parkinson's Disease (PD) is critical for their effective management, as early detection of improvement or degradation signs play an important role on pharmaceutical and/or interventional plans. Within this work, a group of seven PD patients and a group of ten controls performed a set of exercises related to the evaluation of PD gait. Plantar pressure signals were collected and used to derive a set of analytics. Statistical tests and feature selection approaches revealed that the spatial distribution of the Center of Pressure during a static balance exercise is the most discriminative analytic and may be used for every-day monitoring of the patients. Results have revealed that out of the 28 features extracted from the collected signals, 10 were statistically significant (p < 0.05) and can be used to machine learning algorithms and/or similar approaches.

Variational Gaussian Mixture Models with robust Dirichlet concentration priors for virtual population generation in hypertrophic cardiomyopathy: a comparison study.

Nowadays, there is a growing need for the development of computationally efficient virtual population generators for large-scale in-silico clinical trials. In this work, we utilize the Gaussian Mixture Models (GMM) with variational Bayesian inference (BGMM) using robust estimations of Dirichlet concentration priors for the generation of virtual populations. The estimations were based on an exponential transformation of the number of Gaussian components. The proposed method was compared against state-of-the-art virtual data generators, such as, the Bayesian networks, the supervised tree ensembles (STE), the unsupervised tree ensembles (UTE), and the artificial neural networks (ANN) towards the generation of 20000 virtual patients with hypertrophic cardiomyopathy (HCM). Our results suggest that the proposed BGMM can yield virtual distributions with small inter- and intra-correlation difference (0.013 and 0.012), in lower execution time (4.321 sec) than STE which achieved the second-best performance.

Comparison of three U-Net family architectures for left ventricular myocardial wall automatic segmentation.

Left ventricular (LV) segmentation is an important process which can provide quantitative clinical measurements such as volume, wall thickness and ejection fraction. The development of an automatic LV segmentation procedure is a challenging and complicated task mainly due to the variation of the heart shape from patient to patient, especially for those with pathological and physiological changes. In this study, we focus on the implementation, evaluation and comparison of three different Deep Learning architectures of the U-Net family: the custom 2-D U-Net, the ResU-Net++ and the DenseU-Net, in order to segment the LV myocardial wall. Our approach was applied to cardiac CT datasets specifically derived from patients with hypertrophic cardiomyopathy. The results of the models demonstrated high performance in the segmentation process with minor losses. The model revealed a dice score for U-Net, Res-U-net++ and Dense U-Net, 0.81, 0.82 and 0.84, respectively.

A machine learning-based risk stratification model for ventricular tachycardia and heart failure in hypertrophic cardiomyopathy.

BACKGROUND: Machine learning (ML) and artificial intelligence are emerging as important components of precision medicine that enhance diagnosis and risk stratification. Risk stratification tools for hypertrophic cardiomyopathy (HCM) exist, but they are based on traditional statistical methods. The aim was to develop a novel machine learning risk stratification tool for the prediction of 5-year risk in HCM. The goal was to determine if its predictive accuracy is higher than the accuracy of the state-of-the-art tools. METHOD: Data from a total of 2302 patients were used. The data were comprised of demographic characteristics, genetic data, clinical investigations, medications, and disease-related events. Four classification models were applied to model the risk level, and their decisions were explained using the SHAP (SHapley Additive exPlanations) method. Unwanted cardiac events were defined as sustained ventricular tachycardia occurrence (VT), heart failure (HF), ICD activation, sudden cardiac death (SCD), cardiac death, and all-cause death. RESULTS: The proposed machine learning approach outperformed the similar existing risk-stratification models for SCD, cardiac death, and all-cause death risk-stratification: it achieved higher AUC by 17%, 9%, and 1%, respectively. The boosted trees achieved the best performing AUC of 0.82. The resulting model most accurately predicts VT, HF, and ICD with AUCs of 0.90, 0.88, and 0.87, respectively. CONCLUSIONS: The proposed risk-stratification model demonstrates high accuracy in predicting events in patients with hypertrophic cardiomyopathy. The use of a machine-learning risk stratification model may improve patient management, clinical practice, and outcomes in general.

A computational pipeline for data augmentation towards the improvement of disease classification and risk stratification models: A case study in two clinical domains.

Virtual population generation is an emerging field in data science with numerous applications in healthcare towards the augmentation of clinical research databases with significant lack of population size. However, the impact of data augmentation on the development of AI (artificial intelligence) models to address clinical unmet needs has not yet been investigated. In this work, we assess whether the aggregation of real with virtual patient data can improve the performance of the existing risk stratification and disease classification models in two rare clinical domains, namely the primary Sjögren's Syndrome (pSS) and the hypertrophic cardiomyopathy (HCM), for the first time in the literature. To do so, multivariate approaches, such as, the multivariate normal distribution (MVND), and straightforward ones, such as, the Bayesian networks, the artificial neural networks (ANNs), and the tree ensembles are compared against their performance towards the generation of high-quality virtual data. Both boosting and bagging algorithms, such as, the Gradient boosting trees (XGBoost), the AdaBoost and the Random Forests (RFs) were trained on the augmented data to evaluate the performance improvement for lymphoma classification and HCM risk stratification. Our results revealed the favorable performance of the tree ensemble generators, in both domains, yielding virtual data with goodness-of-fit 0.021 and KL-divergence 0.029 in pSS and 0.029, 0.027 in HCM, respectively. The application of the XGBoost on the augmented data revealed an increase by 10.9% in accuracy, 10.7% in sensitivity, 11.5% in specificity for lymphoma classification and 16.1% in accuracy, 16.9% in sensitivity, 13.7% in specificity in HCM risk stratification.