Large Datasets Research Articles

An Unsupervised Learning framework for health diagnosis by incorporating Multiscale Data

Integrating real-time evaluations of health states with Structural Health Monitoring (SHM) data and environmental observations is crucial, particularly under large datasets. Not only is the volume of real-time data analysis substantial, but traditional algorithms like wavelet analysis often require extensive prior work and significant manual input, limiting the application of SHM and big data technologies in structural analysis. To overcome this limitation, this study introduces an unsupervised analytical framework coupling a Denoising Sparse Wavelet Network (DeSpaWN) and a dynamic-inner Principal Component Analysis (DiPCA) algorithm. The DeSpaWN is designed to extract sparse representation from multiscale sensor signals, and the features are subsequently analyzed using the DiPCA to assess health diagnoses. The DeSpaWN is specifically tailored for monitoring high-frequency time series dynamically and adaptively, enabling reconstruction and extraction of signals across various frequency domains. Additionally, the DiPCA effectively handles large-scale time-series data, providing precise evaluations by accounting for dynamic latent correlations. To validate the method, a 5-Degree of Freedom (DoF) structure model is established, equipped with five virtual sensors at each DoF and an anemometer at the top DoF. The sensor signals, obtained under different health states and subjected to artificial excitations, are input into the proposed framework. Validation results demonstrate the effectiveness in synthesizing multi-sensor data and assessing structural integrity. The framework's practical utility is showcased through a real-world case study that assesses the structural integrity of a wind turbine. By integrating real and virtual sensor data collected over a year and cross-referencing this with the fault log, the study validates the algorithm's efficacy in processing multisource heterogeneous data and its precision in structural health diagnosis. The study substantiates real-time tracking of health states for SHM-equipped structures, underscoring its potential to enhance safety and reliability in civil infrastructure.

Enhancing early detection of autistic spectrum disorder in children using machine learning approaches

Diagnosing Autism Spectrum Disorder (ASD) presents a multifaceted challenge, demanding accurate and efficient screening methods. Applying machine learning techniques offers a promising avenue for enhancing diagnostic accuracy and efficiency. This research investigates the efficiency of machine learning in distinguishing individuals with ASD from those without, utilizing a comprehensive dataset comprising screening questions, demographic factors, and ASD related diagnostic classifications. We applied chi-square feature selection technique and also tested Random Forest, Logistic Regression, Gradient Boosting Classifier, and Extra Trees Classifier. Each model showed optimal performance and exhibit high precision, recall, and F1-score for both ASD-positive and ASD-negative instances. Additionally, AUROC curves further validated the models’ exceptional discriminatory abilities, with exceptional results. Our findings highlight the potential of machine learning algorithms for enhancing ASD diagnosis accuracy and efficiency in clinical settings. Further research and validation on larger datasets are required to understand the importance of machine learning methods in ASD diagnosis.

Anthropogenic Climate Change Will Intensify European Explosive Storms Analogous to Alex, Eunice, and Xynthia

Abstract Extratropical storms, particularly explosive storms or “weather bombs” with exceptionally high deepening rates, present substantial risks and are susceptible to climate change. Individual storms may exhibit a complex and hardly detectable response to human-driven climate change because of the atmosphere’s chaotic nature and variability at the regional level. It is thus essential to understand changes in specific storms for building local resilience and advancing our overall comprehension of storm trends. To address this challenge, this study compares analogs—storms with a similar backward track until making landfall—in two climates of three explosive storms impacting different European locations: Alex (October 2020), Eunice (January 2022), and Xynthia (February 2010). We use a large ensemble dataset of 105 members from the Community Earth System Model, version 1 (CESM1). These analogs are identified in two periods: the present-day climate (1991–2001) and a future climate scenario characterized by high anthropogenic greenhouse gas emissions [representative concentration pathway 8.5 (RCP8.5), 2091–2101]. We evaluate future changes in the frequency of occurrence of the storms and intensity, as well as in meteorological hazards and the underlying dynamics. For all storms, our analysis reveals an increase in precipitation and wind severity over land associated with the explosive analogs in the future climate. These findings underscore the potential consequences of explosive storms modified by climate change and their subsequent hazards in various regions of Europe, offering evidence that can be used to prepare and enhance adaptation processes. Significance Statement This study investigates the impact of climate change on explosive storms, or weather bombs, and their potential consequences for European regions. We project future scenarios of three specific storms, Alex, Eunice, and Xynthia, using a state-of-the-art climate model. Our findings reveal an increase in precipitation and wind over land associated with these storms, emphasizing the heightened risks associated with climate change. The significance lies in understanding the local implications of explosive storms, aiding in the development of resilient strategies and adaptation measures.

Hybrid real-time wave forecasting model combining Gaussian process regression and neural networks

To safely and effectively conduct offshore operations, a real-time wave forecast model that can accurately and rapidly forecast waves hours or days in advance is highly valuable. Recent studies have investigated models with neural networks (NNs) and have demonstrated their advantages in terms of forecasting speed and accuracy. However, NNs are limited by their requirement for a large training dataset, which can hinder the construction of the model if sufficient training data are unavailable. Therefore, this study constructed a hybrid real-time wave forecast model by combining an NN model and Gaussian process regression (GPR) model. The objective was to obtain reasonably accurate forecasts that were more accurate than those from individual NN models even when sufficient training data for the NN model was unavailable. To derive an optimal prediction, the values predicted by the NN and GPR models were combined by utilizing the standard deviation of the latter and considering the accuracy of each model. Consequently, predictions that were more accurate compared to those of the individual models for an entire year were obtained. This improvement was particularly significant for predicting relatively large wave heights, as indicated by the daily maximum wave heights. Additionally, various accuracy evaluation metrics demonstrated that despite using a small training dataset, the optimal prediction of the hybrid model was still superior to those of the individual models for predicting the daily maximum wave heights.

Phoenix criteria for sepsis: are these enough to guide a clinician?

In conclusion, Phoenix criteria are a step in the right direction to define life-threatening organ dysfunction in sepsis, but clinicians need to be mindful that diagnosis/treatment of less severe sepsis should not be delayed if these criteria are not met. Therefore, local early detection and management tools for sepsis should be followed. • There has always been a quest for a definition for pediatric sepsis. There are limitations to the previous pediatric sepsis criteria which were published in 2005 by the International Pediatric Sepsis Consensus Conference (IPSCC). IPSCC defines sepsis as a suspected or confirmed infection in the presence of systemic inflammatory response syndrome (SIRS). These new Phoenix Pediatric Sepsis (PPS) criteria for sepsis and septic shock are intended to identify children with life-threatening organ dysfunction due to infection, and the score was developed based on a very large pediatric dataset. • Though the intention of Phoenix criteria is to help identify children with life threatening organ dysfunction, unfortunately the crietria will miss signs of early sepis. In this manuscript, we point out some of the drawbacks of these criteria which need to be borne in mind while applying these criteria.

Prevalence of targetable and potentially targetable alterations in metastatic/recurrent cervical cancer specimens in a large genomic dataset

Prevalence of targetable and potentially targetable alterations in metastatic/recurrent cervical cancer specimens in a large genomic dataset

Vector space model, term frequency-inverse document frequency with linear search, and object-relational mapping Django on hadith data search

For Muslims, the Hadith ranks as the secondary legal authority following the Quran. This research leverages hadith data to streamline the search process within the nine imams’ compendium using the vector space model (VSM) approach. The primary objective of this research is to enhance the efficiency and effectiveness of the search process within Hadith collections by implementing pre-filtering techniques. This study aims to demonstrate the potential of linear search and Django object-relational mapping (ORM) filters in reducing search times and improving retrieval performance, thereby facilitating quicker and more accurate access to relevant Hadiths. Prior studies have indicated that VSM is efficient for large data sets because it assigns weights to every term across all documents, regardless of whether they include the search keywords. Consequently, the more documents there are, the more protracted the weighting phase becomes. To address this, the current research pre-filters documents prior to weighting, utilizing linear search and Django ORM as filters. Testing on 62,169 hadiths with 20 keywords revealed that the average VSM search duration was 51 seconds. However, with the implementation of linear and Django ORM filters, the times were reduced to 7.93 and 8.41 seconds, respectively. The recall@10 rates were 79% and 78.5%, with MAP scores of 0.819 and 0.814, accordingly.

A Systematic Literature Review on The Role of LSTM Networks in Capturing Temporal Dependencies in Data Mining Algorithms

Data mining plays a crucial role in extracting meaningful insights from large and complex data sets, with broad applications in sectors like finance, health care, and market analysis. Traditional techniques—such as classification, clustering, association rule mining, regression, and anomaly detection—are effective for analyzing structured data but struggle with sequential data due to the challenges of modeling temporal dependencies. Long Short Term Memory (LSTM) networks, a specialized form of Recurrent Neural Networks (RNNs), provide a solution to these challenges. By incorporating memory cells and gating mechanisms, LSTM effectively manage long-term dependencies and address issues like vanishing and exploding gradients. This paper reviews the impact of LSTM networks on data mining, analyzing over 60 key publications. By synthesizing concepts and recent advancements, the review underscores how LSTMs enhance the ability of data mining algorithms to capture and predict temporal patterns, reflecting current research trends and innovations

Machine learning techniques for diagnosis of rare diseases from medical images

Introduction/Background: Rare diseases are life-threatening or chronically debilitating conditions that affect a small percentage of the population. Early and accurate diagnosis of rare diseases is challenging due to their complexity and limited understanding. Conventional diagnostic methods are often inadequate, time-consuming, and expensive. Machine learning (ML) has emerged as a promising technique for the analysis of medical images to support the diagnosis of various diseases including rare disorders. ML algorithms can learn complex patterns from large medical imaging datasets to aid clinicians in disease diagnosis, prognosis, and detection of complications. Materials and Methods: A structured search was conducted in electronic databases, including PubMed, Scopus, Web of Science, and Google Scholar to identify peer-reviewed articles published between 2013 to 2023 related to ML-based diagnosis of rare diseases from medical images. A total of 187 articles were identified after removing duplicates. The titles and abstracts of retrieved articles were screened to determine their relevance based on the research objective. Finally, 51 full-text articles were selected for the final review. The key data extracted from selected studies included diseases, imaging modalities, ML algorithms, performance metrics, datasets, and limitations. Results: Most studies evaluated deep learning-based ML techniques, with convolutional neural networks (CNN) being the most applied algorithm. CNNs were mainly used to classify rare diseases based on specific imaging features in X-rays, CT, and MRI scans. Diseases frequently investigated included cardiovascular, neurological, and rare genetic disorders. Publicly available datasets such as Deep Lesion, MSD, and ChestX-ray14 were commonly utilized. Most studies reported high classification accuracy ranging from 80-95% on test datasets. However, limited generalizability due to small private datasets and lack of external validation were limitations. Discussion: The results demonstrate the potential of deep learning for the image-based diagnosis of rare diseases. Features learned from large imaging data sets enabled CNNs to distinguish subtle abnormalities. Hybrid models combining CNNs with other techniques like recurrent neural networks further improved performance. Overall, ML showed promise as a decision support tool. However, more multi-center validation studies are needed using larger and diverse datasets before clinical adoption. Standardization of performance metrics and reporting is also required to establish reliability and generalizability. Conclusion: This comprehensive review provided insights into ongoing research applying ML to medical images for rare disease diagnosis. While initial results are encouraging, further work is still necessary before ML can be reliably used in clinical decision making. Larger collaborative efforts and open-source datasets are needed to advance the field. Standardization of ML frameworks can also promote reproducible research and comparison of different methodologies.

Utilization of big data analytics to identify population health trends and optimize healthcare delivery system efficiency

Big data analytics has emerged as an incredibly valuable tool in understanding population health trends and improving the effectiveness of healthcare delivery systems. By leveraging big data sources from various domains, including the patients’ electronic health records, claims data, wearables, and social media accounts, healthcare organizations can obtain novel and rich insights regarding population health characteristics, predisposing factors, and disease progression. The arrival of big data has transformed healthcare organizations by providing a scientific population health management and health system enhancement tool. Using superior analytical tools, healthcare entrepreneurs and policymakers can discover relations, rates, and patterns that are concealed in large data sets. Such knowledge can be used for early detection of possible interventions, management of resources, and prevention measures, thus leading to better health and less spending on health issues. Moreover, big data analytics helps in early diagnostics and the development of management strategies for high-risk groups, which in turn improves the functioning and effectiveness of healthcare systems. It is also important to note that big data solutions in healthcare are not limited to population health management, but also include functional aspects of healthcare organizations. Additionally, using data about patient movements, resource consumption, and clinical activity, it is possible to determine inefficiencies and opportunities to improve processes in healthcare organizations. This approach of collecting and analyzing data helps in decision making thus reducing time and improving patient flow and experience. However, incorporating real-time data into the clinical decision support systems can improve diagnostic capabilities, treatments offered, and patient tracking resulting in improved quality of services delivered.

Enhancing financial fraud detection using adaptive machine learning models and business analytics

Financial fraud continues to be a critical threat to businesses and economies worldwide, necessitating advanced detection techniques. This paper reviews the role of adaptive machine learning (ML) models and business analytics in enhancing fraud detection systems. Traditional fraud detection methods often fall short in addressing the complexity and evolving nature of fraudulent activities, making adaptive ML models, such as decision trees and neural networks, more effective in identifying subtle patterns in large datasets. Organizations can refine ML models by integrating business analytics, ensuring real-time detection and continuous improvement of fraud detection systems. The paper also explores the challenges in deploying these technologies, including data privacy concerns and model accuracy, as well as the potential impact of emerging technologies such as blockchain and quantum computing. Future directions emphasize improving the interpretability of ML models and leveraging multi-modal data for a more holistic approach to fraud prevention. The synergy between adaptive machine learning and business analytics promises a more resilient and effective framework for combating financial fraud.

A Comprehensive Study of Datasets for Research Works in IoT Networks

Internet of Things is a trending subject in the field of Computer Science and Information Technology. When a research scholar tries to find a problem statement in the arena of IoT, he or she probably Google the phrase “Security Issues in Internet of Things”. On this day August 22, 2024, when we searched the Google with this phrase, we got about 48,40,00,000 results in 0.46 seconds. When we searched Google Scholar with the same phrase, we got about 40,70,000 results (0.52 sec). This shows that lakhs of researchers are toiling to find solutions for Security Issues in IoT. There is a new and trending way of research that is finding solutions for issues in IoT using Machine Learning and Deep Learning Models. ML and DL algorithms need large sets of Data to train and test the algorithms. This made us to take up a study on available Datasets for this research topic. When we made review of literature, in this arena of study, we found that researchers are working on only a few datasets, though numerous datasets are available freely on the internet for academic purposes. Hence in this paper we made a comprehensive study of Datasets available for research works in Internet of Things.

What do Gen Z really want from a workplace?

Purpose This paper aims to offer a deeper understanding of Gen Z professional attitudes and expectations in the workplace through a large data set. Design/methodology/approach The authors provide reference to a unique survey of 1,234 adults who work full- or part-time across different age groups. Gen Z data were compared to three other generational cohorts, namely, Millennials, Gen X and Baby Boomers. Findings Broad findings reveal that Gen Z have lower satisfaction levels with their job compared to other generations. Furthermore, the findings show that Gen Z craves progress, challenge and better mental health and diversity commitments from employers. Related to that, mental health is reported as significantly lower amongst Gen Z participants. Originality/value The research reinforces that there is often a negative view of “other” generations. Indeed, the findings here indicate there are often negative assumptions made about Gen Z. Importantly, these are not aligned with how Gen Z perceive themselves. The authors offer some reflections and recommendations for strategic HR approaches.

A robust approach to Gaussian process implementation

Abstract. Gaussian process (GP) regression is a flexible modeling technique used to predict outputs and to capture uncertainty in the predictions. However, the GP regression process becomes computationally intensive when the training spatial dataset has a large number of observations. To address this challenge, we introduce a scalable GP algorithm, termed MuyGPs, which incorporates nearest-neighbor and leave-one-out cross-validation during training. This approach enables the evaluation of large spatial datasets with state-of-the-art accuracy and speed in certain spatial problems. Despite these advantages, conventional quadratic loss functions used in the MuyGPs optimization, such as root mean squared error (RMSE), are highly influenced by outliers. We explore the behavior of MuyGPs in cases involving outlying observations and, subsequently, develop a robust approach to handle and mitigate their impact. Specifically, we introduce a novel leave-one-out loss function based on the pseudo-Huber function (LOOPH) that effectively accounts for outliers in large spatial datasets within the MuyGPs framework. Our simulation study shows that the LOOPH loss method maintains accuracy despite outlying observations, establishing MuyGPs as a powerful tool for mitigating unusual observation impacts in the large data regime. In the analysis of US ozone data, MuyGPs provides accurate predictions and uncertainty quantification, demonstrating its utility in managing data anomalies. Through these efforts, we advance the understanding of GP regression in spatial contexts.

Alterations in Hurler-Scheie Syndrome Revealed by Mass Spectrometry-Based Proteomics and Phosphoproteomics Analysis.

Hurler-Scheie syndrome (MPS IH/S), also known as mucopolysaccharidosis type I-H/S (MPS IH/S), is a lysosomal storage disorder caused by deficiency of the enzyme alpha-L-iduronidase (IDUA) leading to the accumulation of glycosaminoglycans (GAGs) in various tissues, resulting in a wide range of symptoms affecting different organ systems. Postgenomic omics technologies offer the promise to understand the changes in proteome, phosphoproteome, and phosphorylation-based signaling in MPS IH/S. Accordingly, we report here a large dataset and the proteomic and phosphoproteomic analyses of fibroblasts derived from patients with MPS IH/S (n = 8) and healthy individuals (n = 8). We found that protein levels of key lysosomal enzymes such as cathepsin D, prosaposin, arylsulfatases (arylsulfatase A and arylsulfatase B), and IDUA were downregulated. We identified 16,693 unique phosphopeptides, corresponding to 4,605 proteins, in patients with MPS IH/S. We found that proteins related to the cell cycle, mitotic spindle assembly, apoptosis, and cytoskeletal organization were differentially phosphorylated in MPS IH/S. We identified 12 kinases that were differentially phosphorylated, including hyperphosphorylation of cyclin-dependent kinases 1 and 2, hypophosphorylation of myosin light chain kinase, and calcium/calmodulin-dependent protein kinases. Taken together, the findings of the present study indicate significant alterations in proteins involved in cytoskeletal changes, cellular dysfunction, and apoptosis. These new observations significantly contribute to the current understanding of the pathophysiology of MPS IH/S specifically, and the molecular mechanisms involved in the storage of GAGs in MPS more generally. Further translational clinical omics studies are called for to pave the way for diagnostics and therapeutics innovation for patients with MPS IH/S.

Processing Next-Generation Mass Spectrometry Imaging Data: Principal Component Analysis at Scale.

Mass spectrometry imaging (MSI) is constantly improving in spatial resolving power, throughput and mass resolution. Although beneficial, these improvements increase data set size and content. The larger data requires correspondingly fast computer-based analyses. However, these analyses often do not scale well with increased data size. Principal component analysis (PCA) is an important analytical tool commonly used with MSI data; however, most PCA algorithms load and process the entire data set within random access memory (RAM) which is most often insufficient for large data sets. PCA algorithms that use less RAM than the data set exist but are usually much slower or sacrifice precision and are rarely used for MSI data processing. Incremental PCA (IPCA) is an alternative algorithm that avoids large RAM allocations while also preserving speed and analytical precision. Here, we demonstrate and benchmark the use of differing implementations of IPCA, PCA, and commercial software on large and often complex MSI data sets. We show that using an already-published Python-based IPCA algorithm, IPCA can be successfully applied to MSI data sets too large to fit with RAM. Furthermore, our benchmarks demonstrate that, contrary to expectations, IPCA is faster than all other tested PCA implementations on all large data sets that can be directly compared.

6.M. Scientific session: Risk And Resilience Factors Of Mental Health During COVID: European Multi-country Data Findings

Abstract The COVID-19 pandemic significantly affected mental health. It worsened the mental health and well-being of those already vulnerable or exposed to excess pandemic-related stress. In addition, lockdowns and other measures created additional obstacles to mental health support access. In response, the European Commission funded the RESPOND project in 2020, aimed at enhancing the preparedness of mental health systems for pandemics and global crises. As part of the RESPOND project, (meta-)analyses were conducted on prospective cohort datasets, national health register data, and multi-country individual participant data (IPD). The aim was to examine whether there was a change in prevalence rates of mental health conditions (e.g., anxiety and depression) in the general adult population of the EU during, compared to before and in between, the COVID-19 pandemic waves. Analyses on risk factors (e.g. sociodemographic- and economic factors, severity of COVID-19 infection, (lack of) social activity and support, and loneliness) associated with psychological distress and mental disorders as well as resilience and well-being (coping strategies and core value orientations) in the general population or specific vulnerable groups were conducted as well. This workshop will feature five talks covering various aspects of the RESPOND epidemiological analyses. First, an introduction and rationale of the RESPOND study on short- and long-term impacts of the COVID-19 pandemic and outbreak response, will be provided. Subsequently, findings from the multi-country longitudinal COvid MEntal healTh Survey (COMET) on forecasting psychological distress throughout the pandemic will be presented. Second, findings from the French TEMPO (Trajectoires EpideMiologiques en POpulation) cohort and merged datasets will be presented indicating associations and interactions between socioeconomic factors, mental health, and public health behaviours during the pandemic. Third, using data of a large Belgian prospective cohort, fluctuations in risk of psychological distress across study waves and the factors associated with individual trajectories of psychological distress during the COVID-19 pandemic will be presented. Fourth, the risk of SARS-CoV-2 infection, severe COVID-19 outcomes, and death among individuals with various mental disorders, based on Spanish health records, will be presented with specific attention to sex differences. Finally, based on the IPD meta-analyses of 9 European cohorts, findings on stressor reactivity, resilience trajectories, and factors associated with resilient outcomes and vulnerability are presented. By presenting a selection of findings from large datasets together, we aim to facilitate discussions on critical topics such as vulnerability and resilience to mitigate mental health impacts in future public health crises. Key messages • Targeting specific individual vulnerabilities appears to be more effective in reducing the pandemics’ mental health impact than solely responding to pandemic-related fluctuations. • Research and harmonised data on vulnerability and resilience factors are essential for guiding the development of targeted interventions and support systems for vulnerable populations during crises.

Deep Learning in Finance: A Survey of Applications and Techniques

Machine learning (ML) has transformed the financial industry by enabling advanced applications such as credit scoring, fraud detection, and market forecasting. At the core of this transformation is deep learning (DL), a subset of ML that is robust in processing and analyzing complex and large datasets. This paper provides a comprehensive overview of key deep learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), Deep Belief Networks (DBNs), Transformers, Generative Adversarial Networks (GANs), and Deep Reinforcement Learning (Deep RL). Beyond summarizing their mathematical foundations and learning processes, this study offers new insights into how these models are applied in real-world financial contexts, highlighting their specific advantages and limitations in tasks such as algorithmic trading, risk management, and portfolio optimization. It also examines recent advances and emerging trends in the financial industry alongside critical challenges such as data quality, model interpretability, and computational complexity. These insights can guide future research directions toward developing more efficient, robust, and explainable financial models that address the evolving needs of the financial sector.

Siamese neural networks can identify subjects from anonymised ECGs

Abstract Background/Introduction Many research databases contain anonymised electrocardiograms (ECGs) linked to other sensitive information. ECGs hold features unique to individuals, potentially enabling subject identification from anonymised ECGs. Purpose We assessed if artificial intelligence approaches to output ECG pair similarity can re-identify individuals from anonymised ECGs. Additionally, we aimed to explore clinical risk prediction using ECG similarity over time. Methods We used a convolutional Siamese neural network (SNN), with a triplet loss function, to train a deep learning model determining if two ECGs belong to the same individual (Figure 1). The model aims to encode ECG inputs from same subjects closer than ECG inputs from different subjects. An ECG similarity score is output, corresponding to the probability two ECGs belong to the same subject. This continuous metric can be used in a binary manner. Below a threshold, two ECGs are classed as from the same individual. Our dataset comprised 72,455 secondary care subjects from a USA cohort with 4 to 75 ECGs each, totalling 864,283 ECGs. These were split 50:10:40% at the subject level for training, validation and testing. Results In 2,689,124 same-subject pairs and 2,689,124 different-subject pairs from the test set, the model achieves an accuracy of 91.68% against the binary threshold. This improves to 93.61% and 95.97% in outpatient and normal ECG subsets, respectively. We tested the model using a subject’s ECG to identify them from a group where only one ECG also belongs to that subject. 89.4% success rates occur in groups of 100 normal ECGs (if random, success would be 1%). 91.4%, 95.1%, and 96.3% success occurs when supplemented with subject gender, age (decade-bracket), and both, respectively. We calculated a model certainty score for this task. 99% success occurs when >95% certain. This is in 65% of group size 100 trials. In groups of 1000 normal ECGs, the success rate is 72.5% with no additional information, and 75.9%, 84.2% and 87.5% when supplemented as before (if random, success would be 0.1%). >95% certainty occurs in 33% of group size 1000 trials. For a given subject, ECG pair similarity proved clinically informative. The model accounts for slight variations within a subject’s ECGs, even when far apart in time (Figure 2A). However, it incorrectly predicts two ECGs as being from different individuals for substantial ECG morphology changes (Figure 2B), which could be used to identify clinical deterioration. The temporal evolution of ECG pair similarity reflects clinical trajectory, with greater dissimilarity associating with all-cause mortality (Hazard ratio, 1.22 per 1 standard deviation change, p < 0.0001). Conclusion(s) Anonymised ECGs retain information useful for subject re-identification. Whilst this implies possible ethical and data protection issues with large datasets, such approaches offer potential for continuous monitoring and identifying clinical deterioration.Figure 1Figure 2

Comparison of pulmonary arterial risk assessment tools to predict mortality or morbidity in treatment naive and previously diagnosed patients

Abstract Background Pulmonary arterial hypertension (PAH) risk stratification (1-year prediction of risk) is crucial and should be the most accurate to ensure appropriate aggressiveness of the treatment choices. According to the European guidelines, several risk tools can be used both at baseline and/or follow-up assessment. Yet, a comparison of risk tools performances only in treatment-naïve or only previously diagnosed patients is still missing. Aims To compare the available risk assessment tools performances to predict 1-year mortality but also 1-year clinical worsening in treatment-naïve and previously diagnosed PAH patients. Methods Eight PAH trials (AMBITION, ARIES1, ARIES2, FREEDOM, GRIPHON, PATTENT, PHIRST and SERAPHIN) provided by the FDA were harmonized and treatment-naïve patients were separated from previously diagnosed patients. Outcomes were 1-year mortality and 1-year clinical worsening (i.e., morbidity) either from baseline assessment or at follow-up. Kaplan-Meier and log-rank tests were performed for each of the following risk tools: European 4-strata, REVEAL2.0, REVEAL Lite, noninvasive FRENCH. In each subgroup, their C-indices were calculated and compared using a total of 100 bootstrap samples from the original data, providing a robust method for comparing across the different tools. Results Among the 4,122 patients included, 1,701 (41%) were treatment-naïve and 2,411 (59%) were previously diagnosed patients. Compared to treatment naïve patients, previously diagnosed patients were younger (p<0.001), with lower level of natriuretic peptide (p=0.013), higher 6MWD (p<0.001) and higher mPAP (p<0.001). At 1-year, the mortality rate was the same in the 2 subgroups (5.6% 5.1%, p=0.05) but clinical worsening occurred more often among previously diagnosed patients (19% vs 8.4%, p<0.001). Each of the risk tools were able to predict both 1-year mortality or morbidity in each subgroup, especially among previously diagnosed or at follow-up, with fair to good C-index (Figure 1). Forrest plots of 95%CI for the C-index differences among the 100 bootstrap samples showed that REVEAL 2.0 was the best performer, followed by REVEAL Lite (Figure 2). Conclusions In this large dataset, we showed that all PAH risk tools were able to predict not only 1-year mortality but also 1-year morbidity, both in treatment naïve and previously diagnosed patients. REVEAL 2.0 depicted the higher C-index.Figure 1.C-index for PAH risk toolsFigure 2.Forrest plots of 95%CI