Search

Purpose: In the context of ophthalmologic practice, there has been a rapid increase in the amount of data collected using electronic health records (EHR). Artificial intelligence (AI) offers a promising means of centralizing data collection and analysis, but to date, most AI algorithms have only been applied to analyzing image data in ophthalmologic practice. In this review we aimed to characterize the use of AI in the analysis of EHR, and to critically appraise the adherence of each included study to the CONSORT-AI reporting guideline. Methods: A comprehensive search of three relevant databases (MEDLINE, EMBASE, and Cochrane Library) from January 2010 to February 2023 was conducted. The included studies were evaluated for reporting quality based on the AI-specific items from the CONSORT-AI reporting guideline. Results: Of the 4,968 articles identified by our search, 89 studies met all inclusion criteria and were included in this review. Most of the studies utilized AI for ocular disease prediction (n = 41, 46.1%), and diabetic retinopathy was the most studied ocular pathology (n = 19, 21.3%). The overall mean CONSORT-AI score across the 14 measured items was 12.1 (range 8–14, median 12). Categories with the lowest adherence rates were: describing handling of poor quality data (48.3%), specifying participant inclusion and exclusion criteria (56.2%), and detailing access to the AI intervention or its code, including any restrictions (62.9%). Conclusions: In conclusion, we have identified that AI is prominently being used for disease prediction in ophthalmology clinics, however these algorithms are limited by their lack of generalizability and cross-center reproducibility. A standardized framework for AI reporting should be developed, to improve AI applications in the management of ocular disease and ophthalmology decision making. [ABSTRACT FROM AUTHOR]

The list of standard abbreviations for JDS is available at adsa.org/jds-abbreviations-24. Nonstandard abbreviations are available in the Notes. The deployment of diverse data-generating technologies in livestock farming holds the promise of early disease detection and improved animal well-being. In this paper, we combine routinely collected dairy farm and herd data with weather and high-frequency sensor data from 6 farms to predict new lameness events in various future periods, spanning from the following day to 3 wk. A Random Forest classifier, using input features selected by the Boruta algorithm, was used for the prediction task; effects of individual features were further assessed using partial dependence plots. We achieve precision scores of up to 93% when predicting lameness for the next 3 wk and when using information from the last 3 wk, combined with a balanced accuracy of 79%. Removing sensor data results has a tendency to reduce the precision for predictions, especially when using information from the last 1, 2, or 3 wk. Moving to a larger dataset (without sensor data) of 44 farms keeps the similar balanced accuracy but reduces precision by more than 30%, revealing a substantial a trade-off in model quality between false positives (false lameness alerts) and false negatives (missed lameness events). Sensor data holds promise to further improve the precision of these models, but can be partially compensated by high-resolution data from other systems, such as automated milking systems. [ABSTRACT FROM AUTHOR]

Effectively and equitably allocating medical resources, particularly for minority groups, is a critical issue that warrants further investigation in rural hospitals. Machine learning techniques have gained significant traction and demonstrated strong performance across various fields in recent years. The determination of hyperparameters significantly influences the performance of machine learning models. Thus, this study employs Optuna, a framework specifically designed for optimizing the hyperparameters of machine learning models. Building on prior research, machine learning models with Optuna (MLOPTA) are introduced to forecast diseases of indigenous patients. The numerical results reveal that the designed MLOPTA system can accurately capture the occurrences of specified diseases. Therefore, the MLOPTA system offers a promising approach for disease forecasting. The disease forecasting results can serve as crucial references for allocating hospital resources. [ABSTRACT FROM AUTHOR]

Patient representation learning based on electronic health records (EHR) is a critical task for disease prediction. This task aims to effectively extract useful information on dynamic features. Although various existing works have achieved remarkable progress, the model performance can be further improved by fully extracting the trends, variations, and the correlation between the trends and variations in dynamic features. In addition, sparse visit records limit the performance of deep learning models. To address these issues, we propose the multi-perspective patient representation Extractor (MPRE) for disease prediction. Specifically, we propose frequency transformation module (FTM) to extract the trend and variation information of dynamic features in the time–frequency domain, which can enhance the feature representation. In the 2D multi-extraction network (2D MEN), we form the 2D temporal tensor based on trend and variation. Then, the correlations between trend and variation are captured by the proposed dilated operation. Moreover, we propose the first-order difference attention mechanism (FODAM) to calculate the contributions of differences in adjacent variations to the disease diagnosis adaptively. To evaluate the performance of MPRE and baseline methods, we conduct extensive experiments on two real-world public datasets. The experiment results show that MPRE outperforms state-of-the-art baseline methods in terms of AUROC and AUPRC. [ABSTRACT FROM AUTHOR]

Cardiovascular diseases (CVD) have become a major public health problem affecting the national economy and social development, and have become one of the major causes of death. Therefore, the prevention, control and risk assessment of CVD have been increasingly emphasized. However, current CVD prediction models face limitations in capturing complex relationships within physiological data, potentially hindering accurate risk assessment. This study addresses this gap by proposing a novel Framework for Multi-Input, One-dimensional Convolutional Neural Network (1D-CNN) with Attention Mechanism for CVD (FMI-CAECD). This framework leverages the feature extraction capabilities of Convolutional Neural Network (CNN) alongside an Attention Mechanism to adaptively identify critical features and non-linear relationships within the data. Additionally, Shapley Additive Explanations (SHAP) analysis is incorporated to provide deeper insights into individual feature importance for disease prediction. Performance evaluation on the BRFSS 2022 dataset demonstrates that FMI-CAECD achieves superior accuracy (97.45%), sensitivity (96.84%), specificity (95.07%), and F1-score (92.44%) compared to traditional machine learning baselines and other deep learning models. These findings suggest that FMI-CAECD offers a promising approach for CVD risk assessment. [ABSTRACT FROM AUTHOR]

The gut microbiota plays a vital role in human health, and significant effort has been made to predict human phenotypes, especially diseases, with the microbiota as a promising indicator or predictor with machine learning (ML) methods. However, the accuracy is impacted by a lot of factors when predicting host phenotypes with the metagenomic data, e.g. small sample size, class imbalance, high-dimensional features, etc. To address these challenges, we propose MicroHDF, an interpretable deep learning framework to predict host phenotypes, where a cascade layers of deep forest units is designed for handling sample class imbalance and high dimensional features. The experimental results show that the performance of MicroHDF is competitive with that of existing state-of-the-art methods on 13 publicly available datasets of six different diseases. In particular, it performs best with the area under the receiver operating characteristic curve of 0.9182 ± 0.0098 and 0.9469 ± 0.0076 for inflammatory bowel disease (IBD) and liver cirrhosis, respectively. Our MicroHDF also shows better performance and robustness in cross-study validation. Furthermore, MicroHDF is applied to two high-risk diseases, IBD and autism spectrum disorder, as case studies to identify potential biomarkers. In conclusion, our method provides an effective and reliable prediction of the host phenotype and discovers informative features with biological insights. [ABSTRACT FROM AUTHOR]

The widespread issue of data imbalance in healthcare classification tasks presents a significant challenge, as minority class instances, often representing critical yet less firequent medical conditions such as rare diseases, are overshadowed by the majority class. This imbalance can lead to poor predictive performance in medical diagnostics, where accurately identifying rare but severe conditions is crucial for patient outcomes. To address this, this study introduces a stacking ensemble classifier based on a Generative Adversarial Network (GAN), enhanced by the Shufed Frog Leaping Algorithm (SFLA). The core innovation of this approach lies in the utilization of SFLA to optimize the GAN's generative process, producing highly representative synthetic instances of the minority class. These synthetic instances enhance the authenticity of rare disease cases, which are often underrepresented in clinical datasets. By refining these instances through iterative interactions with the stacking ensemble, the GAN adapts them to closely resemble misclassified samples, thereby improving the system's diagnostic accuracy. This adaptive integration of SFLA and GANs results in a more robust ensemble classifier specifically tailored to handle the complexities of medical data. The technique was tested on imbalanced medical datasets, including those for thyroid and skin disorders, yielding an F1-score and AUC between 0.8 and 1. It outperformed the compared algorithms by 10-15%, significantly enhancing the accuracy of healthcare system evaluations, particularly for rare conditions. [ABSTRACT FROM AUTHOR]

Polycystic ovary syndrome (PCOS) requires early and precise diagnosis to manage and prevent long-term health consequences effectively. In this research, a large dataset of healthcare data gathered from various hospitals in Kerala, India, was evaluated using multiple machine learning (ML) and deep learning (DL) models to identify a highly reliable and accurate prediction of PCOS. The six algorithms used for comparison with the proposed DL model are support vector classification, random forest, logistic regression, k-nearest neighbors, and gaussian naive Bayes; they were selected due to their strengths in handling features in large datasets. The highly parameterized neural networks were tuned using efficient approaches like Optuna and genetic algorithms. The results indicated that the model implemented using our proposed combination of DL model and Optuna, outperformed the traditional models, achieving 93.55% reliability. This suggests the potential for using deep learning for decision-making in diagnosing PCOS. This method demonstrates the importance of integrating various data types with powerful analytic tools in medical diagnostics to support customized therapy. [ABSTRACT FROM AUTHOR]

This research offered an in-depth evaluation of "BambooShoots.AI," a platform aiding bamboo cultivators in harvest timing and pest detection. Using a quantitative method, it incorporated Lund A.M.'s USE Questionnaire (2001) and usage data analysis to gauge user satisfaction and effectiveness. The majority of participants were mid-aged cultivators, providing insights on system usefulness, ease of use, learning, and overall satisfaction. Demographics showed a primary user base of 35-44-year-olds with balanced gender representation and a high rate of Bachelor’s degree holders, highlighting the platform's broad appeal. BambooShoots.AI was noted for its significant usability, scoring well in all evaluated aspects. The study suggested enhancing the interface for older users, continuous feedback integration for improvement, and specialized training programs. It emphasized the need for accessible and inclusive design, aligning with evolving user needs. BambooShoots.AI emerged as a potent, user-focused tool in agricultural technology, pointing to its potential for wider adoption and development in the farming community, and affirming its role as a critical asset in modern agriculture.

ABSTRACT: The deployment of diverse data-generating technologies in livestock farming holds the promise of early disease detection and improved animal well-being. In this paper, we combine routinely collected dairy farm and herd data with weather and high-frequency sensor data from 6 farms to predict new lameness events in various future periods, spanning from the following day to 3 wk. A Random Forest classifier, using input features selected by the Boruta algorithm, was used for the prediction task; effects of individual features were further assessed using partial dependence plots. We achieve precision scores of up to 93% when predicting lameness for the next 3 wk and when using information from the last 3 wk, combined with a balanced accuracy of 79%. Removing sensor data results has a tendency to reduce the precision for predictions, especially when using information from the last 1, 2, or 3 wk. Moving to a larger dataset (without sensor data) of 44 farms keeps the similar balanced accuracy but reduces precision by more than 30%, revealing a substantial a trade-off in model quality between false positives (false lameness alerts) and false negatives (missed lameness events). Sensor data holds promise to further improve the precision of these models, but can be partially compensated by high-resolution data from other systems, such as automated milking systems.

Boosting algorithms are a set of machine learning techniques that are predicated on the notion that a weak learner's acquisition of multiple basic classifiers might yield results that are superior to those of any one simple classifier used alone. A comprehensive evaluation of regularly used boosting techniques against highly investigated diseases is lacking, despite the fact that boosting approaches have been used for disease prediction in many studies. Thus, the purpose of this work is to highlight the main algorithms and strategies in the boosting learning. The results of this work will help academics identify a more appropriate boosting approach to predict disease, as well as better understand current patterns and hotspots in diseases prediction models that use boosting learning. The results showed that adaboost algorithm outperformed other algorithms in terms of accuracy, achieving above 90%. This review also demonstrates how combining two boosting methods can increase the basic classifier's accuracy. By using AdaBoost and LightGBM, the accuracy reached 99.75%. XGBoost and Gradient Boosting techniques were employed more frequently in researches than other boosting algorithms.

The integration of federated learning, attention-based models like BERT, and blockchain technology presents a transformative approach for managing medical records. This paper introduces a hybrid framework combining the latter technologies to solve critical challenges pertaining to the secure management of healthcare data. Federated learning provides a distributed learning of machine models, where sensitive patient data does not need to be transferred, while BERT models improve the precision in processing medical records using natural language understanding. Blockchain adds a layer of security by recording model updates transparently to ensure tamper-proofing and transactions. A concrete methodology for the implementation of the introduced framework including the design of the smart contract in Solidity is provided to secure recording the model updates. Various tests assessing the performance of the proposed system show a significant improvement in data privacy, model security and precision, compared to the other systems. This hybrid methodology offers advances in handling medical records and elaborates a new benchmark in integrating AI and blockchain for healthcare. This framework thus redefines secure and collaborative healthcare data management, setting the stage for further enhancements in privacy-focused AI applications in medical contexts. [ABSTRACT FROM AUTHOR]

The integration of the Internet of Things with the Cloud improves our lives by facilitating smooth connections between people and items. Predictive analytics, fueled by cutting-edge machine learning and artificial intelligence, turns reactive healthcare initiatives into proactive ones. A subset of machine learning called deep learning is essential for quickly analyzing large datasets, producing insightful conclusions, and efficiently addressing challenging problems. For early interventions and preventive care, especially for those who are at risk, accurate and timely illness prediction is crucial. Making accurate prediction models becomes crucial when utilizing electronic medical records. Accuracy is improved by using deep learning variations of recurrent neural networks that can handle sequential time-series data. Predictive analytics is applied to cloud-stored electronic medical records and data from Internet of Things devices in this suggested system. With a remarkable accuracy of 98.86%, the smart healthcare system is intended to monitor and anticipate the risk of heart disease utilizing Bi-LSTM (bidirectional long short-term memory). Furthermore, it reaches 98.9% accuracy, 98.8% sensitivity, 98.89% specificity, and 98.86% F-measure. These outcomes greatly surpass the performance of current smart heart disease prediction systems. [ABSTRACT FROM AUTHOR]

As medical imaging technologies advance, these tools are playing a more and more important role in assisting clinical disease diagnosis. The fusion of biomedical imaging and multi-modal information is profound, as it significantly enhances diagnostic precision and comprehensiveness. Integrating multi-organ imaging with genomic information can significantly enhance the accuracy of disease prediction because many diseases involve both environmental and genetic determinants. In the present study, we focused on the fusion of imaging-derived phenotypes (IDPs) and polygenic risk score (PRS) of diseases from different organs including the brain, heart, lung, liver, spleen, pancreas, and kidney for the prediction of the occurrence of nine common diseases, namely atrial fibrillation, heart failure (HF), hypertension, myocardial infarction, asthma, type 2 diabetes, chronic kidney disease, coronary artery disease (CAD), and chronic obstructive pulmonary disease, in the UK Biobank (UKBB) dataset. For each disease, three prediction models were developed utilizing imaging features, genomic data, and a fusion of both, respectively, and their performances were compared. The results indicated that for seven diseases, the model integrating both imaging and genomic data achieved superior predictive performance compared to models that used only imaging features or only genomic data. For instance, the Area Under Curve (AUC) of HF risk prediction was increased from 0.68 ± 0.15 to 0.79 ± 0.12, and the AUC of CAD diagnosis was increased from 0.76 ± 0.05 to 0.81 ± 0.06. [ABSTRACT FROM AUTHOR]

Graph neural networks (GNNs) have gained significant attention in disease prediction where the latent embeddings of patients are modeled as nodes and the similarities among patients are represented through edges. The graph structure, which determines how information is aggregated and propagated, plays a crucial role in graph learning. Recent approaches typically create graphs based on patients' latent embeddings, which may not accurately reflect their real-world closeness. Our analysis reveals that raw data, such as demographic attributes and laboratory results, offers a wealth of information for assessing patient similarities and can serve as a compensatory measure for graphs constructed exclusively from latent embeddings. In this study, we first construct adaptive graphs from both latent representations and raw data respectively, and then merge these graphs via weighted summation. Given that the graphs may contain extraneous and noisy connections, we apply degree-sensitive edge pruning and kNN sparsification techniques to selectively sparsify and prune these edges. We conducted intensive experiments on two diagnostic prediction datasets, and the results demonstrate that our proposed method surpasses current state-of-the-art techniques. • We propose a novel graph construction using multimodal embeddings and patient features via weighted sum. • We use edge pruning and kNN sparsification to reduce redundant and noisy edges in graph creation. • We design a novel loss function to optimize modality-shared and specific representations separately. • Our method outperforms others on two datasets; ablation studies and visualizations confirm its effectiveness. [ABSTRACT FROM AUTHOR]

Theileriosis, a parasitic disease, caused by Theileria spp. and transmitted through ticks, poses a significant threat to livestock, leading to elevated morbidity and mortality rates. This study investigated the incidence trend of theileriosis in Kerala, India, over three years (2019–21). Notably, the research unveiled a substantial upsurge in bovine theileriosis cases within Kerala during this period, partly attributed to the state’s severe flooding and landslides in 2018, triggered by incessant monsoon rains. The present study envisaged pinpointing the risk factors underlying the prevalence of theileriosis in Kerala. Employing linear discriminant analysis, key environmental and remote sensing variables influencing the disease’s incidence were identified. Subsequently, these risk factors underwent climate disease modelling, leading to the creation of risk maps. To predict areas sensitive to theileriosis outbreaks in Kerala, two regression models and nine machine learning models were employed. The gradient boost and random forest models demonstrated the most accurate fit among these. The study also estimated the basic reproduction number (R0), which ranged from 0.89 to 1.8. This value indicates a high potential for Theileria spp. transmission within the study area. Consequently, the research outcomes offer valuable insights into pinpointing high risk theileriosis locations in livestock in Kerala. [ABSTRACT FROM AUTHOR]

Internet of Things (IoT) frameworks generates data for large and remote agricultural areas through sensors and use this data for crop predictions by several machine learning algorithms. Farming is the practice of producing crops, rearing of livestock, cultivation of soil and it is important for economic development of the country. Farmers have been following traditional farming practices till now. These techniques were imprecise and reduced productivity and time consumption. Determining the steps that essential for practicing at its appropriate season helps to increase the productivity of precision farming. In this research, the primary objectives are to enhance precision farming practices by introducing a comprehensive IoT-based framework and employing advanced machine learning algorithms. Therefore, this research work introduces crop recommendation and disease prediction that helps farmers to increase productivity and reduce manual labor. The proposed Multi-level Kronecker Guided Pelican Convolutional Neural Network (MKGPCNN) focuses on crop recommendation, providing forecasts for suitable crops in the agricultural sector. Simultaneously, the Combined Graph Sample and Aggregate Attention Network (CGSAAN) are introduced for crop disease identification and recommending appropriate fertilizers to manage diseases and enhance harvests. The evaluation of both systems on publicly accessible datasets, namely the crop recommendation dataset and the new plant diseases dataset, demonstrates higher accuracy rates of 99% and 98%, precision of 99.5% and 99%, recall of 99.6% and 98.9%. The results suggest that the introduced system have the potential to significantly assist farmers in smarter crop management and harvesting, contributing to increased productivity and reduced manual labor. [ABSTRACT FROM AUTHOR]

Objective Develop a novel technique to identify an optimal number of regression units corresponding to a single risk point, while creating risk scoring systems from logistic regression-based disease predictive models. The optimal value of this hyperparameter balances simplicity and accuracy, yielding risk scores of small scale and high accuracy for patient risk stratification. Materials and Methods The proposed technique applies an adapted line search across all potential hyperparameter values. Additionally, DeLong test is integrated to ensure the selected value produces an accuracy insignificantly different from the best achievable risk score accuracy. We assessed the approach through two case studies predicting diabetic retinopathy (DR) within six months and hip fracture readmissions (HFR) within 30 days, involving cohorts of 90 400 diabetic patients and 18 065 hip fracture patients. Results Our scores achieve accuracies insignificantly different from those obtained by existing approaches, reaching AUROCs of 0.803 and 0.645 for DR and HFR predictions, respectively. Regarding the scale, our scores ranged 0-53 for DR and 0-15 for HFR, while scores produced by existing methods frequently spanned hundreds or thousands. Discussion According to the assessment, our risk scores offer simple and accurate predictions for diseases. Furthermore, our new DR score provides a competitive alternative to state-of-the-art risk scores for DR, while our HFR case study presents the first risk score for this condition. Conclusion Our technique offers a generalizable framework for crafting precise risk scores of compact scales, addressing the demand for user-friendly and effective risk stratification tool in healthcare. [ABSTRACT FROM AUTHOR]

Diabetes is a terrible health situation characterized by high-rise blood glucose levels. If it is not predicted at an early stage, then it generates the problems in the human body like kidney failure or premature death, and stroke. Controlling blood glucose levels provides patients with helpful dietary recommendations, which are critical components of diabetes management. In the past decades, diverse conventional approaches have been executed to predict the beginning stages of diabetes mellitus depending on physical and substance tests. Still, developing a new framework that can effectively diagnose diabetes mellitus-affected patients is required. To this end, the major target of this task is to predict diabetes mellitus with an advanced accuracy rate with the help of the Heuristic-based Ensemble Model Selection Strategy (H-EMSS). In the data collection phase, the Pima Indian Diabetes dataset (PID) is taken from the storage area of UCI. The data cleaning is performed in the pre-processing stage, which is the technique of removing or fixing, corrupted, incorrect, duplicate, incomplete data, or incorrectly formatted, inside a dataset. Then, the diabetes prediction is accomplished by the H-EMSS. Here, 10 base learners like Naive Bayes (NB), Convolutional Neural Network (CNN), Linear Regression (LR), Deep Neural Network (DNN), Support Vector Machine (SVM), Artificial Neural Network (ANN), Decision Tree (DT), Random Forest (RF), Auto Encoder (AE) and Recurrent Neural Network (RNN) are considered. From these, three classifiers are optimally selected by the Modified Scalar Factor-based Elephant Herding Optimization (MSF-EHO), so that the prediction rate will be high. The suggested methodology's efficacy is also compared and analyzed, with the findings demonstrating the recommended model's superiority. The overall evaluation is that the Root Mean Square Error (RMSE) of the designed Modified Scalar Factor-based Elephant Herding Optimization-Heuristic-based Ensemble Model Selection Strategy (MSF-EHO-H-EMSS) attains 4.601% and also the Mean Absolute Error (MAE) on the designed method achieves 0.99%. Thus, the given outcomes of the designed method revealed that it achieves elevated performance than the other existing techniques regarding diverse error metrics. [ABSTRACT FROM AUTHOR]

A growing number of studies are incorporating microbiome variation into machine learning-based models for the prediction of disease. The prediction of disease is generally improved by considering microbiome signatures. The most promising improvements in microbiome-augmented disease prediction are seen for liver disease, respiratory diseases, diabetes mellitus, inflammatory bowel disease, and neuropsychiatric conditions. As improved prediction relies on accurate microbial definition, detection, and diversity metrics, important considerations during bioinformatics and experimental design must be taken into consideration to limit the effect of unwanted technical bias. The human microbiome has been increasingly recognized as having potential use for disease prediction. Predicting the risk, progression, and severity of diseases holds promise to transform clinical practice, empower patient decisions, and reduce the burden of various common diseases, as has been demonstrated for cardiovascular disease or breast cancer. Combining multiple modifiable and non-modifiable risk factors, including high-dimensional genomic data, has been traditionally favored, but few studies have incorporated the human microbiome into models for predicting the prospective risk of disease. Here, we review research into the use of the human microbiome for disease prediction with a particular focus on prospective studies as well as the modulation and engineering of the microbiome as a therapeutic strategy. [ABSTRACT FROM AUTHOR]

Diagnosis and prognosis are especially difficult areas of medical research related to cancer due to the high incidence of breast cancer, which has surpassed all other cancers in terms of female mortality. Another factor that has a substantial influence on the quality of life of cancer patients is the fear that they may experience a relapse of their disease. The objective of the study is to give medical practitioners a more effective strategy for using ensemble learning techniques to forecast when breast cancer may recur. This research aimed to investigate the usage of deep neural networks (DNNs) and artificial neural networks (ANNs) in addition to machine learning (ML) based approaches, including bagging, averaging, and voting, to enhance the efficacy of breast cancer relapse diagnosis on two breast cancer relapse datasets. Results from the empirical study demonstrate that the proposed ensemble learning-enabled approach improves accuracies by 96.31% and 95.81%, precisions by 96.70% and 96.15%, sensitivities by 98.88% and 98.68%, specificities by 84.62% in both, F1-scores by 97.78% and 97.40%, and area under the curve (AUCs) of 0.987 and 0.978, with University Medical Centre, Institute of Oncology (UMCIO) and Wisconsin prognostic breast cancer (WPBC) datasets respectively. Consequently, these improved disease outcomes may encourage physicians to use this model to make better treatment choices. [ABSTRACT FROM AUTHOR]

Background Atrial fibrillation (AF) coexisting with coronary artery disease (CAD) remains a prevailing issue that often results in poor short- and long-term patient outcomes. Screening has been proposed as a method to increase AF detection rates and reduce the incidence of poor prognosis through early intervention. Nevertheless, due to the cost implications and uncertainty over the benefits of a systematic screening programme, the International Task Force currently recommends against screening. This study is to employ Bayesian networks (BN) for assessing the pre-test probability (PTP) of AF in patients with CAD.Methods A total of 12,552 patients with CAD were divided into the CAD patients with AF group (CHD-AF group) and the CAD patients without AF group (non-AF group). Univariate analysis and LASSO regression method were used to screen for potential risk factors. The maximum-minimum climb (MMHC) algorithm was used to construct the directed acyclic graph (DAG) of BN. Predictive power was tested using internal validation, external validation and 10-fold internal cross-validation. Finally, the generated BN model was compared with four machine learning algorithms.Results Fourteen indicators were included in the BN, including age, gender, systolic blood pressure (SBP), low-density lipoprotein cholesterol (LDL-C), serum uric acid (UA), gamma-glutamyltransferase (GGT), direct bilirubin (DBIL), lipoproteins [LP(a)], NYHA cardiac function grading, diabetes mellitus and hypertension, palpitation, dyspnoea and the left atrial diameter. The BN model performs well on both the test set (AUC = 0.90) and internal 10-fold cross-validation (AUC = 0.89 ± 0.01).Conclusion The prediction model of AF with CAD constructed based on BN has high prediction performance and may provide a new tool for large-scale AF screening.

Background: Various disease prediction models have been developed, capitalizing on the wide use of electronic health records, but environmental factors that are important in the development of noncommunicable diseases are rarely included in the prediction models. Hypertensive disorders of pregnancy are leading causes of maternal morbidity and mortality and are known to cause several serious complications later in life. Objective: This study aims to develop early hypertensive disorders of pregnancy prediction models using comprehensive environmental factors based on self-report questionnaires in early pregnancy. Study Design: We developed machine learning and artificial intelligence models for the early prediction of hypertensive disorders of pregnancy using early pregnancy data from approximately 23,000 pregnancies in the Tohoku Medical Megabank Birth and Three Generation Cohort Study. We clarified the important features for prediction based on regression coefficients or Gini coefficients of the interpretable artificial intelligence models (i.e., logistic regression, random forest and XGBoost models) among our developed models. Results: The performance of the early hypertensive disorders of pregnancy prediction models reached an area under the receiver operating characteristic curve of 0.93, demonstrating that the early hypertensive disorders of pregnancy prediction models developed in this study retain sufficient performance in hypertensive disorders of pregnancy prediction. Among the early prediction models, the best performing model was based on self-reported questionnaire data in early pregnancy (mean of 20.2 gestational weeks at filling) which consist of comprehensive lifestyles. The interpretation of the models reveals that both eating habits were dominantly important for prediction. Conclusion: We have developed high-performance models for early hypertensive disorders of pregnancy prediction using large-scale cohort data from the Tohoku Medical Megabank project. Our study clearly revealed that the use of comprehensive lifestyles from self-report questionnaires led us to predict hypertensive disorders of pregnancy risk at the early stages of pregnancy, which will aid early intervention to reduce the risk of hypertensive disorders of pregnancy.