Random Forest-based Algorithm Research Articles

A Random Forest-based Prediction for Liver Cirrhosis

Prediction of cirrhosis through timely screening and diagnosis with effective medical and lifestyle interventions can be taken to delay the disease progression and improve treatment effectiveness. Therefore, early cirrhosis prediction helps reduce associated complications and improve quality of life. This study established a random forest-based algorithm and compared its performance with other machine learning models based on an Indian Liver Patient Records. The dataset was divided 30% of the test set and 70% of the training set. The results showed that the Accuracy, Area under the curve (AUC) and Recall values of Random Forest (RF) model were 0.73, 0.76 and 0.88, respectively, which demonstrated the proposed model can predict the results better than other models. The proposed model for predicting liver cirrhosis in the future will continue to be improved and perfected to improve the accuracy of liver cirrhosis prediction and treatment effect, thereby reducing the health risks and burdens brought by the disease.

The Dynamics of Floating Macroalgae in the East China Sea and Its Vicinity Waters: A Comparison between 2017 and 2023

Ulva prolifera and Sargassum are two common floating macroalgae in China’s coastal algal bloom events. Ulva prolifera frequently emerges concomitantly with Sargassum outbreaks, thereby presenting challenges to the monitoring of algal blooms, thereby presenting challenges to the monitoring of algae. To tackle the challenge of differentiating between Ulva prolifera and Sargassum, this study employs Sentinel-2 MSI data for spectral analysis. Notably, significant disparities in the Remote Top of Atmosphere Reflectance (Rtoa) between Ulva prolifera and Sargassum are observed. This study proposes a random forest-based algorithm for discriminating between Ulva prolifera and Sargassum in the regions of the Yellow Sea and East China Sea. The algorithm introduced in this study attains remarkable accuracy in distinguishing Ulva prolifera and Sargassum within Sentinel-2 MSI data, achieving identical F1 scores of 99.1% for both. Moreover, when tested with GF-1 WFV data, the algorithm showcases outstanding performance; this demonstrates the algorithm’s robustness and its ability to mitigate the uncertainty linked to threshold selection. Simultaneously, a comparative analysis of algae distribution was conducted for both 2017 and the period from January to May 2023. Experimental results indicate that the algorithm exhibits high accuracy in distinguishing between Ulva prolifera and Sargassum. This capability will significantly enhance the monitoring of large algae in maritime regions; this holds crucial theoretical significance and offers substantial practical value in the realm of marine ecological conservation.

A multivariate blood metabolite algorithm stably predicts risk and resilience to major depressive disorder in the general population.

Socioeconomic pressures, sex, and physical health status strongly influence the development of major depressive disorder (MDD) and mask other contributing factors in small cohorts. Resilient individuals overcome adversity without the onset of psychological symptoms, but resilience, as for susceptibility, has a complex and multifaceted molecular basis. The scale and depth of the UK Biobank affords an opportunity to identify resilience biomarkers in rigorously matched, at-risk individuals. Here, we evaluated whether blood metabolites could prospectively classify and indicate a biological basis for susceptibility or resilience to MDD. Using the UK Biobank, we employed random forests, a supervised, interpretable machine learning statistical method to determine the relative importance of sociodemographic, psychosocial, anthropometric, and physiological factors that govern the risk of prospective MDD onset (total n=15,710). We then used propensity scores to rigorously match individuals with a history of MDD (n=491) against a resilient subset of individuals without an MDD diagnosis (retrospectively or during follow-up; n=491) using an array of key social, demographic, and disease-associated drivers of depression risk. 381 blood metabolites and clinical chemistry variables and 4 urine metabolites were integrated to generate a multivariate random forest-based algorithm using 10-fold cross-validation to predict prospective MDD risk and resilience. In unmatched individuals, a first case of MDD, with a median time-to-diagnosis of 72 years, can be predicted using random forest classification probabilities with an area under the receiver operator characteristic curve (ROC AUC) of 0.89. Prospective resilience/susceptibility to MDD was then predicted with a ROC AUC of 0.72 (x˜= 3.2 years follow-up) and 0.68 (x˜= 7.2 years follow-up). Increased pyruvate was identified as a key biomarker of resilience to MDD and was validated retrospectively in the TwinsUK cohort. Blood metabolites prospectively associate with substantially reduced MDD risk. Therapeutic targeting of these metabolites may provide a framework for MDD risk stratification and reduction. New York Academy of Sciences' Interstellar Programme Award; Novo Fonden; Lincoln Kingsgate award; Clarendon Fund; Newton-Abraham studentship (University of Oxford). The funders had no role in the development of the present study.

Optimization of biocementation responses by artificial neural network and random forest in comparison to response surface methodology.

In this article, the optimization of the specific urease activity (SUA) and the calcium carbonate (CaCO3) using microbially induced calcite precipitation (MICP) was compared to optimization using three algorithms based on machine learning: random forest regressor, artificial neural networks (ANNs), and multivariate linear regression. This study applied the techniques in two existing response surface method (RSM) experiments involving MICP technique. Random forest-based models and artificial neural network-based models were submitted through the optimization of hyperparameters via cross-validation technique and grid search, to select the best-optimized model. For this study, the random forest-based algorithm is aimed at having the best performance of 0.9381 and 0.9463 in comparison to the original r2 of 0.9021 and 0.8530, respectively. This study is aimed at exploring the capability of using machine learning-based models in small datasets for the purpose of optimization of experimental variables in MICP technique and the meaningfulness of the models by their specificities in the small experimental datasets applied to experimental designs. This study is aimed at exploring the capability of using machine learning-based models in small datasets for experimental variable optimization in MICP technique. The use of these techniques can create prerogatives to scale and mitigate costs in future experiments associated to the field.

Classification of Arctic Sea Ice Type in CFOSAT Scatterometer Measurements Using a Random Forest Classifier

The Ku-band scatterometer called CSCAT onboard the Chinese–French Oceanography Satellite (CFOSAT) is the first spaceborne rotating fan-beam scatterometer (RFSCAT). A new algorithm for classification of Arctic sea ice types on CSCAT measurement data using a random forest classifier is presented. The random forest classifier is trained on the National Snow and Ice Data Center (NSIDC) weekly sea ice age and sea ice concentration product. Five feature parameters, including the mean value of horizontal and vertical polarization backscatter coefficient, the standard deviation of horizontal and vertical polarization backscatter coefficient and the copol ratio, are innovatively extracted from orbital measurement for the first time to distinguish water, first-year ice (FYI) and multi-year ice (MYI). The overall accuracy and kappa coefficient of sea ice type model are 93.35% and 88.53%, respectively, and the precisions of water, FYI, and MYI are 99.67%, 86.60%, and 79.74%, respectively. Multi-source datasets, including daily sea ice type from the EUMETSAT Ocean and Sea Ice Satellite Application Facility (OSI SAF), NSIDC weekly sea ice age, multi-year ice concentration (MYIC) provided by the University of Bremen, and SAR-based sea ice type released by Copernicus Marine Environment Monitoring Service (CMEMS) have been used for comparison and validation. It is shown that the most obvious difference in the distribution of sea ice types between the CSCAT results and OSI SAF sea ice type are mainly concentrated in the marginal zones of FYI and MYI. Furthermore, compared with OSI SAF sea ice type, the area of MYI derived from CSCAT is more homogeneous with less noise, especially in the case of younger multiyear ice. In the East Greenland region, CSCAT identifies more pixels as MYI with lower MYIC values, showing better accuracy in the identification of areas with obvious mobility of MYI. In conclusion, this research verifies the capability of CSCAT in monitoring Arctic sea ice classification, especially in the spatial homogeneity and detectable duration of sea ice classification. Given the high accuracy and processing speed, the random forest-based algorithm can offer good guidance for sea ice classification with FY-3E/RFSCAT, i.e., a dual-frequency (Ku and C band) scatterometer called WindRAD.

Using a Machine Learning Algorithm to Predict Online Patient Portal Utilization: A Patient Engagement Study.

There is a low rate of online patient portal utilization in the U.S. This study aimed to utilize a machine learning approach to predict access to online medical records through a patient portal. This is a cross-sectional predictive machine learning algorithm-based study of Health Information National Trends datasets (Cycles 1 and 2; 2017-2018 samples). Survey respondents were U.S. adults (≥18 years old). The primary outcome was a binary variable indicating that the patient had or had not accessed online medical records in the previous 12 months. We analyzed a subset of independent variables using k-means clustering with replicate samples. A cross-validated random forest-based algorithm was utilized to select features for a Cycle 1 split training sample. A logistic regression and an evolved decision tree were trained on the rest of the Cycle 1 training sample. The Cycle 1 test sample and Cycle 2 data were used to benchmark algorithm performance. Lack of access to online systems was less of a barrier to online medical records in 2018 (14%) compared to 2017 (26%). Patients accessed medical records to refill medicines and message primary care providers more frequently in 2018 (45%) than in 2017 (25%). Privacy concerns, portal knowledge, and conversations between primary care providers and patients predict portal access. Methods described here may be employed to personalize methods of patient engagement during new patient registration.

Change detection in a rural landscape: A case study of processes and main driving factors along with its response to thermal environment in Farim, Iran.

This study aims to investigate the alteration of Land Use/Land Cover (LULC) change and its response to changes in land surface temperature (LST) and heat island phenomena of a rural district known as Farim in the north of Iran from 1990 to 2020 using multi-date Landsat data. The random forest-based algorithm, supported by Google Earth Engine, is used to execute the LULC classification with an overall accuracy of more than 92%. Based on the LULC results, in terms of area changes, the classes of bare land, rice fields, and water bodies encountered an increase, but woods and dry farms decreased. The present study also incorporates the trends of land cover change that are analyzed using regression based on the temporal datasets of the three leading driving factors: temperature, precipitation, and population. The result demonstrates that the main changing factors of the mostly changed class (bare land) are population/precipitation and temperature/population. Additionally, the effect of LULC change on seasonal LST and urban heat island (UHI) is also analyzed in this study. The result witnessed a significant LST rise in the summer and winter seasons of about 12.87°C and 14.2°C, respectively over the study period. The Urban Thermal Field Variance Index (UTFVI), characterizing the heat island phenomenon, shows that the strongest UTFVI zone is in the central area and thenone UTFVI zone is in the surrounding region. Moreover, both seasons have seen a significant rise in none UTFVI zones compared to decreasing strongest UTFVI zone. The result of the present study will be helpful for urban planners and climate researchers who study future land cover change and its associated driving factors.

Deriving alpha angle from anterior-posterior dual-energy x-ray absorptiometry scans: an automated and validated approach.

Introduction: Alpha angle (AA) is a widely used imaging measure of hip shape that is commonly used to define cam morphology, a bulging of the lateral aspect of the femoral head. Cam morphology has shown strong associations with hip osteoarthritis (OA) making the AA a clinically relevant measure. In both clinical practice and research studies, AA tends to be measured manually which can be inconsistent and time-consuming. Objective: We aimed to (i) develop an automated method of deriving AA from anterior-posterior dual-energy x-ray absorptiometry (DXA) scans; and (ii) validate this method against manual measures of AA. Methods: 6,807 individuals with left hip DXAs were selected from UK Biobank. Outline points were manually placed around the femoral head on 1,930 images before training a Random Forest-based algorithm to place the points on a further 4,877 images. An automatic method for calculating AA was written in Python 3 utilising these outline points. An iterative approach was taken to developing and validating the method, testing the automated measures against independent batches of manually measured images in sequential experiments. Results: Over the course of six experimental stages the concordance correlation coefficient, when comparing the automatic AA to manual measures of AA, improved from 0.28 [95% confidence interval 0.13-0.43] for the initial version to 0.88 [0.84-0.92] for the final version. The inter-rater kappa statistic comparing automatic versus manual measures of cam morphology, defined as AA ³≥60°, improved from 0.43 [80% agreement] for the initial version to 0.86 [94% agreement] for the final version. Conclusions: We have developed and validated an automated measure of AA from DXA scans, showing high agreement with manually measuring AA. The proposed method is available to the wider research community from Zenodo.

Comparison of Stress Detection through ECG and PPG signals using a Random Forest-based Algorithm.

Stress has been classified as the health epidemic of the 21st century with an increasingly active research interest within the fields of psychology, neuroscience, medicine, and more recently affective computing. At present, stress is identified through cortisol levels in saliva but there is no unanimously accepted standard for continuous stress evaluation. With recent development in wearable sensors, many scientists are interested in stress identification through physiological signals such as the Heart rate variability (HRV). In this paper, we present a supervised machine learning-based algorithm to detect stress from HRV derived from electrocardiograms (ECG) as well as photoplethysmograms (PPG), as a low cost alternative to ECG. HRV features from ECG and PPG signals of 46 healthy subjects were analysed and used to separately train and test a subject-independent Random Forest algorithm. In both datasets, stress was accurately identified with more than 80% F1-score and 90% AUC. Results show that PPG is a good surrogate to ECG for HRV analysis and stress detection. The proposed algorithm has the potential to assist researchers and clinicians in the automated continuous analysis of stress.

TRForest: a novel random forest-based algorithm for tRNA-derived fragment target prediction.

tRNA fragments (tRFs) are small RNAs comparable to the size and function of miRNAs. tRFs are generally Dicer independent, are found associated with Ago, and can repress expression of genes post-transcriptionally. Given that this expands the repertoire of small RNAs capable of post-transcriptional gene expression, it is important to predict tRF targets with confidence. Some attempts have been made to predict tRF targets, but are limited in the scope of tRF classes used in prediction or limited in feature selection. We hypothesized that established miRNA target prediction features applied to tRFs through a random forest machine learning algorithm will immensely improve tRF target prediction. Using this approach, we show significant improvements in tRF target prediction for all classes of tRFs and validate our predictions in two independent cell lines. Finally, Gene Ontology analysis suggests that among the tRFs conserved between mice and humans, the predicted targets are enriched significantly in neuronal function, and we show this specifically for tRF-3009a. These improvements to tRF target prediction further our understanding of tRF function broadly across species and provide avenues for testing novel roles for tRFs in biology. We have created a publicly available website for the targets of tRFs predicted by tRForest.

Machine learning-based diagnosis support system for differentiating between clinical anxiety and depression disorders

In light of the need for objective mechanism-based diagnostic tools, the current research describes a novel diagnostic support system aimed to differentiate between anxiety and depression disorders in a clinical sample.Eighty-six psychiatric patients with clinical anxiety and/or depression were recruited from a public hospital and assigned to one of the experimental groups: Depression, Anxiety, or Mixed. The control group included 25 participants with no psychiatric diagnosis.Participants performed a battery of six cognitive-behavioral tasks assessing biases of attention, expectancies, memory, interpretation and executive functions. Data were analyzed with a machine-learning (ML) random forest-based algorithm and cross-validation techniques. The model assigned participants to clinical groups based solely on their aggregated cognitive performance.By detecting each group's unique performance pattern and the specific measures contributing to the prediction, the ML algorithm predicted diagnosis classification in two models: (I) anxiety/depression/mixed vs. control (76.81% specificity, 69.66% sensitivity), and (II) anxiety group vs. depression group (80.50% and 66.46% success rates in classifying anxiety and depression, respectively).The findings demonstrate that the cognitive battery can be utilized as a support system for psychiatric diagnosis alongside the clinical interview. This implicit tool, which is not based on self-report, is expected to enable the clinician to achieve increased diagnostic specificity and precision. Further, this tool may increase the confidence of both clinician and patient in the diagnosis by equipping them with an objective assessment tool. Finally, the battery provides a profile of biased cognitions that characterizes the patient, which in turn enables more fine-tuned, individually-tailored therapy.

Implementation of a Machine-Learning Algorithm in the Electronic Health Record for Targeted Screening for Familial Hypercholesterolemia: A Quality Improvement Study.

Implementation of a Machine-Learning Algorithm in the Electronic Health Record for Targeted Screening for Familial Hypercholesterolemia: A Quality Improvement Study.

A random forest-based algorithm for data-intensive spatial interpolation in crop yield mapping

High-resolution yield maps are an essential tool in modern agriculture. Using spatial interpolation, spatially discrete sampled yield data from yield monitors can be transformed into continuous yield maps. However, spatial interpolation is usually performed using methods that can be computationally demanding or that lack credibility measurements. The objectives of this work were to improve and evaluate a spatial machine learning algorithm for yield mapping at a fine scale. The core method used for mapping is Quantile Regression Forest Spatial Interpolation (QRFI), in which covariates from the spatial neighborhood of the sampled yields are used to predict yields at unsampled sites. To assess the algorithm performance, more than one thousand yield monitor datasets from several plant species were processed with QRFI, and other geostatistical (ordinary kriging, KG) and non-geostatistical (spatial inverse distance interpolation, IDW) methods. We illustrated the application of QRFI for yield mapping using yield monitor datasets of different grain crops from the Argentine Pampas. Evaluation of the methods showed that all statistical metrics suggested better results for yield maps obtained by QRFI than by KG or IDW. Globally, prediction error of QRFI was 11.5%, which was on average at least 16% better than the corresponding results of the other spatial interpolation methods. The machine learning algorithm QRFI can be successfully applied to perform spatial interpolation of yields at the field scale and to assess the associated prediction uncertainty.

Deriving alpha angle from anterior-posterior dual-energy x-ray absorptiometry scans: an automated and validated approach

Introduction: Alpha angle (AA) is a widely used measure of hip shape that is commonly used to define cam morphology, a bulging of the lateral aspect of the femoral head. Cam morphology has shown strong associations with hip osteoarthritis (OA) making the AA a clinically relevant measure. In both clinical practice and research studies, AA tends to be measured manually which can be inconsistent and time-consuming. Objective: We aimed to (i) develop an automated method of deriving AA from anterior-posterior dual-energy x-ray absorptiometry (DXA) scans; and (ii) validate this method against manual measures of AA. Methods: 6,807 individuals with left hip DXAs were selected from UK Biobank. Outline points were manually placed around the femoral head on 1,930 images before training a Random Forest-based algorithm to place the points on a further 4,877 images. An automatic method for calculating AA was written in Python 3 utilising these outline points. An iterative approach was taken to developing and validating the method, testing the automated measures against independent batches of manually measured images in sequential experiments. Results: Over the course of six experimental stages the concordance correlation coefficient, when comparing the automatic AA to manual measures of AA, improved from 0.28 [95% confidence interval 0.13-0.43] for the initial version to 0.88 [0.84-0.92] for the final version. The inter-rater kappa statistic comparing automatic versus manual measures of cam morphology, defined as AA ³≥60°, improved from 0.43 [80% agreement] for the initial version to 0.86 [94% agreement] for the final version. Conclusions: We have developed and validated an automated measure of AA from DXA scans, showing high agreement with manually measuring AA. The proposed method is available to the wider research community from Zenodo.

Adaptive LiDAR Sampling and Depth Completion Using Ensemble Variance.

This work considers the problem of depth completion, with or without image data, where an algorithm may measure the depth of a prescribed limited number of pixels. The algorithmic challenge is to choose pixel positions strategically and dynamically to maximally reduce overall depth estimation error. This setting is realized in daytime or nighttime depth completion for autonomous vehicles with a programmable LiDAR. Our method uses an ensemble of predictors to define a sampling probability over pixels. This probability is proportional to the variance of the predictions of ensemble members, thus highlighting pixels that are difficult to predict. By additionally proceeding in several prediction phases, we effectively reduce redundant sampling of similar pixels. Our ensemble-based method may be implemented using any depth-completion learning algorithm, such as a state-of-the-art neural network, treated as a black box. In particular, we also present a simple and effective Random Forest-based algorithm, and similarly use its internal ensemble in our design. We conduct experiments on the KITTI dataset, using the neural network algorithm of Ma et al. and our Random Forest-based learner for implementing our method. The accuracy of both implementations exceeds the state of the art. Compared with a random or grid sampling pattern, our method allows a reduction by a factor of 4-10 in the number of measurements required to attain the same accuracy.

Random Forest-based Algorithm for Sleep Spindle Detection in Infant EEG.

Sleep spindles are associated with normal brain development, memory consolidation and infant sleep-dependent brain plasticity and can be used by clinicians in the assessment of brain development in infants. Sleep spindles can be detected in EEG, however, identifying sleep spindles in EEG recordings manually is very time-consuming and typically requires highly trained experts. Research on the automatic detection of sleep spindles in infant EEGs has been limited to-date. In this study, we present a novel supervised machine learning-based algorithm to detect sleep spindles in infant EEG recordings. EEGs collected from 141 ex-term born infants and 6 ex-preterm born infants, recorded at 4 months of age (adjusted), were used to train and test the algorithm. Sleep spindles were annotated by experienced clinical physiologists as the gold standard. The dataset was split into training (81 ex-term), validation (30 ex-term), and testing (30 ex-term + 6 ex-preterm) set. 15 features were selected for input into a random forest algorithm. Sleep spindles were detected in the ex-term infant EEG test set with 92.1% sensitivity and 95.2% specificity. For ex-preterm born infants, the sensitivity and specificity were 80.3% and 91.8% respectively. The proposed algorithm has the potential to assist researchers and clinicians in the automated analysis of sleep spindles in infant EEG.

Improving Multi-Tumor Biomarker Health Check-up Tests with Machine Learning Algorithms.

Background: Tumor markers are used to screen tens of millions of individuals worldwide at annual health check-ups, especially in East Asia. Machine learning (ML)-based algorithms that improve the diagnostic accuracy and clinical utility of these tests can have substantial impact leading to the early diagnosis of cancer. Methods: ML-based algorithms, including a cancer screening algorithm and a secondary organ of origin algorithm, were developed and validated using a large real world dataset (RWD) from asymptomatic individuals undergoing routine cancer screening at a Taiwanese medical center between May 2001 and April 2015. External validation was performed using data from the same period from a separate medical center. The data set included tumor marker values, age, and gender from 27,938 individuals, including 342 subsequently confirmed cancer cases. Results: Separate gender-specific cancer screening algorithms were developed. For men, a logistic regression-based algorithm outperformed single-marker and other ML-based algorithms, with a mean area under the receiver operating characteristic curve (AUROC) of 0.7654 in internal and 0.8736 in external cross validation. For women, a random forest-based algorithm attained a mean AUROC of 0.6665 in internal and 0.6938 in external cross validation. The median time to cancer diagnosis (TTD) in men was 451.5, 204.5, and 28 days for the mild, moderate, and high-risk groups, respectively; for women, the median TTD was 229, 132, and 125 days for the mild, moderate, and high-risk groups. A second algorithm was developed to predict the most likely affected organ systems for at-risk individuals. The algorithm yielded 0.8120 sensitivity and 0.6490 specificity for men, and 0.8170 sensitivity and 0.6750 specificity for women. Conclusions: ML-derived algorithms, trained and validated by using a RWD, can significantly improve tumor marker-based screening for multiple types of early stage cancers, suggest the tissue of origin, and provide guidance for patient follow-up.

Multicentre validation of a computer-based tool for differentiation of acute Kawasaki disease from clinically similar febrile illnesses

BackgroundThe clinical features of Kawasaki disease (KD) overlap with those of other paediatric febrile illnesses. A missed or delayed diagnosis increases the risk of coronary artery damage. Our computer algorithm...

Effects of Prenatal Air Pollution on Family-Specific, Genome-Wide, Repetitive Element Methylation in Cord Blood

OPS 48: Air pollution and epigenetics, Room 417, Floor 4, August 26, 2019, 4:30 PM - 5:30 PM Background/Objective: Repetitive elements (REs) comprise more than 50% of the genome and methylation suppresses their transpositions (insertions into other genomic regions) that can damage DNA. Global average RE methylation is inversely associated with particulate matter ≤2.5µm (PM2.5) exposure, however, averaging all RE methylation may overlook their biological activity, as RE subfamilies and individual REs vary in methylation. We employed a method predicting genome-wide, RE subfamily-specific, methylation in cord blood and investigated their associations with prenatal PM2.5 exposure. Methods: Trimester average PM2.5 was estimated using a satellite-based spatio-temporally resolved prediction model and residence location of 443 women from the PROGRESS birth-cohort in Mexico City. We measured cord blood DNA methylation with the Illumina EPICArray using a random forest-based algorithm to predict genome-wide, locus-specific CpG methylation of long interspersed elements-1 (L1) and a short interspersed element (Alu) using nearby CpG sites. We modeled associations between PM2.5 with: 1) methylation of subfamilies within the most recent L1PA, L1PB, and Alu lineages averaged across the genome; and 2) genomic locus-specific aggregates of 2 or more adjacent RE CpGs from these subfamilies. Results: We imputed methylation of 1,183 CpG aggregates and 25 subfamilies within the L1PA and L1PB lineages and 69,596 aggregates and 36 Alu subfamilies with mean methylation of 92%. Average methylation of 20 L1 and 35 Alu subfamilies was inversely associated with 2nd trimester PM2.5. Restricting analysis to full-length L1, methylation of L1PA2, L1PA3, L1PA5, and the active element, L1HS, were inversely associated with PM2.5 in the 2nd and 3rd trimester. Investigations of individual CpGs yielded two aggregates associated with PM2.5 after multiple comparison corrections. Conclusions: Reduced RE subfamily methylation may promote transpositions and cell toxicity that could be related to health endpoints such as cancer. Future work will investigate RE expression and whether RE methylation/expression mediate relationships between PM2.5 exposure and health.

Patient-Level Prediction of Cardio-Cerebrovascular Events in Hypertension Using Nationwide Claims Data.

BackgroundPrevention and management of chronic diseases are the main goals of national health maintenance programs. Previously widely used screening tools, such as Health Risk Appraisal, are restricted in their achievement this goal due to their limitations, such as static characteristics, accessibility, and generalizability. Hypertension is one of the most important chronic diseases requiring management via the nationwide health maintenance program, and health care providers should inform patients about their risks of a complication caused by hypertension.ObjectiveOur goal was to develop and compare machine learning models predicting high-risk vascular diseases for hypertensive patients so that they can manage their blood pressure based on their risk level.MethodsWe used a 12-year longitudinal dataset of the nationwide sample cohort, which contains the data of 514,866 patients and allows tracking of patients’ medical history across all health care providers in Korea (N=51,920). To ensure the generalizability of our models, we conducted an external validation using another national sample cohort dataset, comprising one million different patients, published by the National Health Insurance Service. From each dataset, we obtained the data of 74,535 and 59,738 patients with essential hypertension and developed machine learning models for predicting cardiovascular and cerebrovascular events. Six machine learning models were developed and compared for evaluating performances based on validation metrics.ResultsMachine learning algorithms enabled us to detect high-risk patients based on their medical history. The long short-term memory-based algorithm outperformed in the within test (F1-score=.772, external test F1-score=.613), and the random forest-based algorithm of risk prediction showed better performance over other machine learning algorithms concerning generalization (within test F1-score=.757, external test F1-score=.705). Concerning the number of features, in the within test, the long short-term memory-based algorithms outperformed regardless of the number of features. However, in the external test, the random forest-based algorithm was the best, irrespective of the number of features it encountered.ConclusionsWe developed and compared machine learning models predicting high-risk vascular diseases in hypertensive patients so that they may manage their blood pressure based on their risk level. By relying on the prediction model, a government can predict high-risk patients at the nationwide level and establish health care policies in advance.