Stroke Risk Prediction Models Research Articles

Plasma metabolites, systolic blood pressure, lifestyle, and stroke risk: A prospective cohort study.

To estimate the associations of stroke risk with plasma metabolites, metabolic risk score (MRS), the combinations of MRS with hypertension or lifestyle, and lifestyle-related metabolic signature. To assess the improvement of the stroke risk prediction model through the incorporation of MRS. A total of 77,315 participants from the UK Biobank were included in this study. Xgboost and LASSO-Cox regression were used to select metabolites and construct MRS. Elastic net regression was utilized to construct the lifestyle-related metabolic signature. Multivariate Cox regression was used to estimate the associations between metabolites, MRS, the combinations of MRS with hypertension or lifestyle, lifestyle-related metabolic signature, and stroke risk. We identified 48, 63, 39, and 4 metabolites associated with the risk of stroke, ischemic stroke (IS), subarachnoid hemorrhage (SAH), and intracerebral hemorrhage (ICH), respectively. High MRS significantly increased the risk of stroke (HR = 2.65 (95% confidence interval (CI): 2.09-3.35)), IS (HR = 2.45 (95% CI: 1.89-3.17)), ICH (HR = 2.74 (95% CI: 1.55-4.85)), and SAH (HR = 4.64 (95% CI: 2.25-9.56)). In the combination analyses, compared with normal systolic blood pressure (SBP) and low MRS, normal/high SBP, and high MRS significantly increased stroke risk (HR = 5.80 (95% CI: 2.75-12.27)/6.37 (95% CI: 3.22-12.62)). A favorable/unfavorable lifestyle and high MRS also significantly increased stroke risk (HR = 2.38 (95% CI: 1.73-3.28)/3.86 (95% CI: 2.63-5.67)) compared with a favorable lifestyle and low MRS. Incorporating MRS into the 15-year stroke and IS risk prediction model increased the areas under the curves (AUCs) from 0.746 to 0.766 and from 0.771 to 0.811, respectively. The metabolic signature was correlated with adherence to a healthy lifestyle (r = 0.414; P = 2.22e-16) and inversely associated with stroke risk (HR = 0.80 (95% CI: 0.73-0.86)). Various metabolites and MRS were significantly associated with the risk of stroke, IS, ICH, and SAH. Individuals with a high MRS may face an elevated stroke risk among populations with high SBP or unhealthy lifestyle, even those with normal SBP or healthy lifestyle. MRS provided modest improvement to the stroke risk prediction model. The lifestyle-related metabolic signature could reduce 20% stroke risk.

Design Analysis of Stroke Risk Prediction Model Employing the Hybrid Structure Implementation of Deep Transfer Learning System

Design Analysis of Stroke Risk Prediction Model Employing the Hybrid Structure Implementation of Deep Transfer Learning System

Abstract WP236: Harmonization of Stroke Risk Prediction Variables Using Natural Language Processing

Introduction: Early identification of stroke risk can profoundly influence an individual's chance of survival. While machine learning-based stroke risk prediction models perform better than traditional counterparts, they require integrating diverse open data sources. Variable discrepancies - where the same concept is described differently - pose challenges for data integration, requiring time-intensive manual harmonization. To address this, we developed an automated harmonization technique using natural language models. Methods: We utilized data from the Atherosclerosis Risk in Communities (ARIC), Multi-Ethnic Study of Atherosclerosis (MESA), Framingham Offspring Study, and Reasons for Geographic and Racial Differences in Stroke (REGARDS) cohorts. Our approach involved training a two-layer fully connected neural network (FCN) to predict a harmonized ‘concept variable’ based on variable descriptions. These predictions employed embeddings from Bidirectional Encoder Representations from Transformers for Biomedical Text Mining (BioBERT). Further refining was achieved through paired sentence classification tasks to determine if description pairs shared identical concepts. We experimented with both separate cohort training and combined cohort training to evaluate context dependency. We used cosine similarity between the embeddings of the pre-trained BioBERT encoder as a prediction score for our baseline method. Results: The harmonization methods using natural language models outperformed the baseline method. The FCN model improved the area under the receiver operating characteristic from 0.787 (baseline) to 0.985 for the paired classification task. For harmonization concept prediction, the top-2 accuracy - correct prediction in the top two answers - improved from 11.9% (baseline) to 54.6%, while the top-5 accuracy improved from 16.1% (baseline) to 67.6%. Conclusions: Utilizing natural language processing for data harmonization provides a scalable approach that improves accuracy and efficiency of prediction. This facilitates the inclusion of diverse cohorts, broadening the sample size and range of indicators available, which in turn advances the versatility and the generalizability of stroke risk prediction models.

Short- and long-term outcomes of patients with minor stroke and nonvalvular atrial fibrillation

Background and purposeNonvalvular atrial fibrillation (NVAF) is a risk factor for stroke. This study was undertaken to determine the influence of NVAF on the mortality and recurrent stroke after a minor stroke event.MethodsData were derived from the Third China National Stroke Registry (CNSR-III) which enrolled 15,166 subjects during August 2015 through March 2018 in China. Patients with minor stroke (NIHSS ≤ 5) within 24 h after onset were included. Clinical outcomes including all-cause mortality, cardiovascular death, recurrent ischemic stroke, and recurrent hemorrhagic stroke were collected. The Cox proportional hazards models were used to determine the association between NVAF and clinical outcomes.ResultsA total of 4,753 patients were included in our study. Of them, 222 patients had NVAF (4.7%) (mean age, 71.1 years) and 4,531 patients were without AF (95.3%) (mean age, 61.4 years). NVAF was associated with 12-month cardiovascular mortality in both univariate (hazards ratio [HR], 4.13; 95% confidence interval [CI], 1.84 to 9.31; P < 0.001) and multivariate analyses (HR, 4.66; 95% CI, 1.79 to 12.15; P = 0.001). There was no difference in the in-hospital ischemic stroke recurrence rate between the two groups (HR, 0.45 [95% CI, 0.19 to 1.05] P = 0.07 at discharge). However, patients with NVAF had a lower rate of recurrent ischemic stroke at medium- (3 months and 6 months) and long-term (12 months) follow-up (HR, 0.33 [95% CI, 0.16 to 0.68] P = 0.003 at 3 months; 0.49 [95% CI, 0.27 to 0.89] P = 0.02 at 6 months; 0.55 [95% CI, 0.32 to 0.94] P = 0.03 at 12 months, respectively) compared with those without. There was no difference in all-cause mortality and hemorrhagic stroke between the two groups during follow-up.ConclusionsMinor stroke patients with NVAF were at higher risk of cardiovascular death but had a lower rate of recurrent ischemic stroke compared to those without during the subsequent year after stroke event. A more accurate stroke risk prediction model for NVAF is warranted for optimal patient care strategies.

Abstract 18629: Discriminative Accuracy of Cha 2 Ds 2 Vasc Score, and Development of Predictive Accuracy Model Using Machine Learning for Ischemic Stroke in Cardiac Amyloidosis

Background: Cardiac amyloidosis (CA) in conjunction with atrial fibrillation (AF) presents unique management challenges. CHA2DS2VASc score in these patients is believed to underestimate the risk of ischemic stroke, necessitating a better predictive model in these patients. Methods: Data was obtained from the National Readmission Database (NRD). Outcomes between CA-AF and no-CA-AF were compared using multivariate regression analysis to calculate adjusted odds ratios (aOR). AutoScore; an interpretable machine learning framework, was used to develop a stroke risk prediction model, the predictive accuracy of which was evaluated with an area under the curve (AUC) using the receiver operating characteristic analysis. Results: A total of 11,860,804 (CA-AF 22,687 [0.19%] and no-CA-AF 11,838,117) patients were identified from 2015-2019. The adjusted odds of mortality (aOR 1.41 and 1.29), stroke (aOR 1.78 and 1.74), non-intracranial hemorrhage (aOR 2.10 and aOR 1.85), and intracranial hemorrhage (aOR 14.4 and aOR 4.26) were significantly higher in CA-AF compared with non-CA-AF at both index admission and 30-days, respectively. The CHA 2 DS 2 VASc score had a poor discriminative accuracy for stroke at 30-days in CA-AF (AUC 49%, 95%CI 47%-51%, p=0.54). The machine learning autoscore integrative model revealed that the predictive ability of our newly proposed E-CHADS score (end-stage renal disease (ESRD), congestive heart failure, hypertension, active cancer, dementia, and diabetes mellitus) for 30-day risk of ischemic stroke in CA-AF was excellent (for a cutoff of 52 points random forest score) with an AUC of 80% (95%CI 74%-86%) Conclusion: Cardiac amyloidosis carries a high risk of ischemic stroke that is not accurately predicted by the CHA 2 DS 2 VASc score. Our proposed model (E-CHADS) identifies 3 new variables (ESRD, dementia, and cancer) that have higher discriminative accuracy for ischemic stroke in these patients.

Ischemic stroke prediction using machine learning in elderly Chinese population: The Rugao Longitudinal Ageing Study.

Compared logistic regression (LR) with machine learning (ML) models, to predict the risk of ischemic stroke in an elderly population in China. We applied 2208 records from the Rugao Longitudinal Ageing Study (RLAS) for ischemic stroke risk prediction assessment. Input variables included 103 phenotypes. For 3-year ischemic stroke risk prediction, we compared the discrimination and calibration of LR model and ML methods, where ML methods include Random Forest (RF), Gaussian kernel Support Vector Machines (SVM), Multilayer perceptron (MLP), K-Nearest Neighbors Algorithm (KNN), and Gradient Boosting Decision Tree (GBDT) to develop an ischemic stroke risk prediction model. Age, pulse, waist circumference, education level, β2-microglobulin, homocysteine, cystatin C, folate, free triiodothyronine, platelet distribution width, QT interval, and QTc interval were significant induced predictors of ischemic stroke. For ischemic stroke prediction, the ML approach was able to tap more biochemical and ECG-related multidimensional phenotypic indicators compared to the LR model, which placed more importance on general demographic indicators. Compared to the LR model, SVM provided the best discrimination and calibration (C-index: 0.79vs. 0.71, 11.27% improvement in model utility), with the best performance in both validation and test data. In a comparison of LR with five ML models, the accuracy of ischemic stroke prediction was higher by combining ML with multiple phenotypes. Combined with other studies based on elderly populations in China, ML techniques, especially SVM, have shown good long-term predictive performance, inspiring the potential value of ML use in clinical practice.

CA-UNet Segmentation Makes a Good Ischemic Stroke Risk Prediction.

Stroke is still the World's second major factor of death, as well as the third major factor of death and disability. Ischemic stroke is a type of stroke, in which early detection and treatment are the keys to preventing ischemic strokes. However, due to the limitation of privacy protection and labeling difficulties, there are only a few studies on the intelligent automatic diagnosis of stroke or ischemic stroke, and the results are unsatisfactory. Therefore, we collect some data and propose a 3D carotid Computed Tomography Angiography (CTA) image segmentation model called CA-UNet for fully automated extraction of carotid arteries. We explore the number of down-sampling times applicable to carotid segmentation and design a multi-scale loss function to resolve the loss of detailed features during the process of down-sampling. Moreover, based on CA-Unet, we propose an ischemic stroke risk prediction model to predict the risk in patients using their 3D CTA images, electronic medical records, and medical history. We have validated the efficacy of our segmentation model and prediction model through comparison tests. Our method can provide reliable diagnoses and results that benefit patients and medical professionals.

An Interpretable Prediction Model for Stroke Based on XGBoost and SHAP

Objective: To establish a stroke prediction and feature analysis model integrating XGBoost and SHAP to aid the clinical diagnosis and prevention of stroke. Methods: Based on the open data set on Kaggle, with the help of data preprocessing and grid parameter optimization, an interpretable stroke risk prediction model was established by integrating XGBoost and SHAP and an explanatory analysis of risk factors was performed. Results: The XGBoost model’s accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC) were 96.71%, 93.83%, 99.59%, and 99.19%, respectively. Our explanatory analysis showed that age, type of residence, and history of hypertension were key factors affecting the incidence of stroke. Conclusion: Based on the data set, our analysis showed that the established model can be used to identify stroke, and our explanatory analysis based on SHAP increases the transparency of the model and facilitates medical practitioners to analyze the reliability of the model.

Simple CaroTID-VasC score to predict one-year risk of stroke in symptomatic carotid stenosis patients

BackgroundA subset of extracranial symptomatic carotid stenosis (ESCS) patients may fare well on current optimal medical therapy (OMT), and surgery may be avoided in these patients. Therefore, we aimed to develop and validate a stroke risk prediction model to stratify the risk among ESCS patients. MethodsAdult ESCS patients who denied revascularization procedures were enrolled prospectively and prescribed OMT. Patients were followed-up for twelve months after assessing the clinical, imaging, and hemodynamics-based risk predictors at baseline. Cox regression analysis was performed on predictors which were significant in univariate analysis. Beta coefficients of significant predictors in Cox regression were used to generate a numeric score. The model was internally validated using bootstrapping. ResultsA total of 20 (20.2%) out of 99 patients had event recurrence during the follow-up. Transient ischemic attack index event (P = 0.014), diabetes mellitus (P = 0.018), contralateral significant stenosis (P = 0.007), echolucent plaque (P = 0.011), and impaired vasomotor reactivity (P = 0.006) were significant predictors in Cox regression analysis. A points score (0–6) was derived from regression coefficients of the significant predictors. The area under ROC was 0.884 for the developed model and 0.832 for the bootstrapped model. Youden's index divided the score into low-risk (2.2%) and high-risk (35.8%) groups, and the difference in risk was significant (P < 0.001). ConclusionsMost ESCS patients benefited from OMT, and the CaroTID-VasC score was effective in stratifying patients for risk of endpoint occurrence. The developed model may help identify high-risk subgroups of ESCS patients and assist the decision-making of carotid interventions.

Predictive Accuracy of Stroke Risk Prediction Models Across Black and White Race, Sex, and Age Groups

Stroke is the fifth-highest cause of death in the US and a leading cause of serious long-term disability with particularly high risk in Black individuals. Quality risk prediction algorithms, free of bias, are key for comprehensive prevention strategies. To compare the performance of stroke-specific algorithms with pooled cohort equations developed for atherosclerotic cardiovascular disease for the prediction of new-onset stroke across different subgroups (race, sex, and age) and to determine the added value of novel machine learning techniques. Retrospective cohort study on combined and harmonized data from Black and White participants of the Framingham Offspring, Atherosclerosis Risk in Communities (ARIC), Multi-Ethnic Study for Atherosclerosis (MESA), and Reasons for Geographical and Racial Differences in Stroke (REGARDS) studies (1983-2019) conducted in the US. The 62 482 participants included at baseline were at least 45 years of age and free of stroke or transient ischemic attack. Published stroke-specific algorithms from Framingham and REGARDS (based on self-reported risk factors) as well as pooled cohort equations for atherosclerotic cardiovascular disease plus 2 newly developed machine learning algorithms. Models were designed to estimate the 10-year risk of new-onset stroke (ischemic or hemorrhagic). Discrimination concordance index (C index) and calibration ratios of expected vs observed event rates were assessed at 10 years. Analyses were conducted by race, sex, and age groups. The combined study sample included 62 482 participants (median age, 61 years, 54% women, and 29% Black individuals). Discrimination C indexes were not significantly different for the 2 stroke-specific models (Framingham stroke, 0.72; 95% CI, 0.72-073; REGARDS self-report, 0.73; 95% CI, 0.72-0.74) vs the pooled cohort equations (0.72; 95% CI, 0.71-0.73): differences 0.01 or less (P values >.05) in the combined sample. Significant differences in discrimination were observed by race: the C indexes were 0.76 for all 3 models in White vs 0.69 in Black women (all P values <.001) and between 0.71 and 0.72 in White men and between 0.64 and 0.66 in Black men (all P values ≤.001). When stratified by age, model discrimination was better for younger (<60 years) vs older (≥60 years) adults for both Black and White individuals. The ratios of observed to expected 10-year stroke rates were closest to 1 for the REGARDS self-report model (1.05; 95% CI, 1.00-1.09) and indicated risk overestimation for Framingham stroke (0.86; 95% CI, 0.82-0.89) and pooled cohort equations (0.74; 95% CI, 0.71-0.77). Performance did not significantly improve when novel machine learning algorithms were applied. In this analysis of Black and White individuals without stroke or transient ischemic attack among 4 US cohorts, existing stroke-specific risk prediction models and novel machine learning techniques did not significantly improve discriminative accuracy for new-onset stroke compared with the pooled cohort equations, and the REGARDS self-report model had the best calibration. All algorithms exhibited worse discrimination in Black individuals than in White individuals, indicating the need to expand the pool of risk factors and improve modeling techniques to address observed racial disparities and improve model performance.

Development of stroke predictive model in community-dwelling population: A longitudinal cohort study in Southeast China.

Stroke has been the leading cause of death and disability in the world. Early recognition and treatment of stroke could effectively limit brain damage and vastly improve outcomes. This study aims to develop a highly accurate prediction model of stroke with a list of lifestyle behaviors and clinical characteristics to distinguish high-risk groups in the community-dwelling population. Participants in this longitudinal cohort study came from the community-dwelling population in Suzhou between November 2018 and June 2019. A total of 4,503 residents participated in the study, while stroke happened to 22 participants in the 2-year follow-up period. Baseline information of each participant was acquired and enrolled in this study. T-test, Chi-square test, and Fisher's exact test were used to examine the relationship of these indexes with stroke, and a prediction scale was constructed by multivariate logistic regression afterward. Receiver operating characteristic analysis was applied to testify to the prediction accuracy. A highly accurate prediction model of stroke was constructed by age, gender, exercise, meat and vegetarian diet, BMI, waist circumference, systolic blood pressure, Chinese visceral adiposity index, and waist-height ratio. Two additional prediction models for overweight and non-overweight individuals were formulated based on crucial risk factors, respectively. The stroke risk prediction models for community-dwelling and overweight populations had accuracies of 0.79 and 0.82, severally. Gender and exercise were significant predictors (χ2 > 4.57, p < 0.05) in the community-dwelling population model, while homocysteine (χ2 = 4.95, p < 0.05) was significant in the overweight population model. The predictive models could predict 2-year stroke with high accuracy. The models provided an effective tool for identifying high-risk groups and supplied guidance for improving prevention and treatment strategies in community-dwelling population.

Prediction Model of Ischemic Stroke Based on Machine Learning

Machine learning (ML) can be used for deep mining and analysis of multidimensional medical data. At present, it has been widely used in medical diagnosis and prognosis prediction. This paper aims to make the existing research no longer focus on identifying key risk factors of stroke, and predict stroke risk more accurately. We collected the data of 3,962 cerebral apoplexy patients from 2019 to 2020, according to gender (male: 2,613; female: 1,349) and age (16–40 years old; 41–54 years old; 55–69 years old; 70 years old and above) layered. After data preprocessing, a stroke risk prediction model was built using principal component analysis (PCA) and extreme learning institutions (ELM). The prediction accuracy of PCA-ELM was as high as 97%. In this model, total cholesterol and high density lipoprotein are taken as 10 important factors that affect the incidence of stroke. This method can timely and efficiently mine the factors influencing the incidence of cerebral apoplexy from the data, and can predict the incidence of cerebral apoplexy. It has high value in practical application. This paper has great reference value in the research of brain death.

A review of risk concepts and models for predicting the risk of primary stroke.

Predicting an individual's risk of primary stroke is an important tool that can help to lower the burden of stroke for both the individual and society. There are a number of risk models and risk scores in existence but no review or classification designed to help the reader better understand how models differ and the reasoning behind these differences. In this paper we review the existing literature on primary stroke risk prediction models. From our literature review we identify key similarities and differences in the existing models. We find that models can differ in a number of ways, including the event type, the type of analysis, the model type and the time horizon. Based on these similarities and differences we have created a set of questions and a system to help answer those questions that modelers and readers alike can use to help classify and better understand the existing models as well as help to make necessary decisions when creating a new model.

STROKE RISK PREDICTION MODEL FOR AFRICAN POPULATIONS: AN APPROACH FOR EARLY IDENTIFICATION OF POPULATIONS AT RISK

Background & Aim Early identification and treatment of populations at risk of stroke in Africa have been delimited by the data to design a risk prediction model tailor-made for indigenous Africans. This study developed and tested an Afrocentric risk scoring model to predict stroke occurrence among West Africans. Method 7066 case-control pairs (identified in the SIREN study) were randomly partitioned into testing and validation datasets (85:15), and conditional logistic regression models were developed from seventeen traditional factors of stroke in the training dataset. Significant factors were assigned constant, statistically weighted (based on regression coefficient), and receiver operating characteristics curves were constructed (from the testing dataset) to estimate cut-off points for discriminate stroke cases from healthy controls (to build an additive risk scoring model for stroke). Sensitivity, specificity, positive and negative predicted values were estimated to assess the strength of the stroke risk prediction model at P < 0.05. Results Cohen's kappa for validity was maximal at a total risk score of 56% using statistical weighting approaches for risk quantification in both datasets after identifying 15 traditional vascular factors for stroke occurrence. The stroke risk prediction model had a strong predictive accuracy (76%, 95%CI: 74%–79%), sensitivity (80.3%), specificity (63.0%), positive predictive value (68.5%) and negative predictive value (76.2%) in the validation dataset. Conclusion The Afrocentric stroke risk models had good accuracy and might be helpful for stroke prediction as a viable approach in de-escalating the rising burden of stroke across multi-ethnic indigenous African populations. Early identification and treatment of populations at risk of stroke in Africa have been delimited by the data to design a risk prediction model tailor-made for indigenous Africans. This study developed and tested an Afrocentric risk scoring model to predict stroke occurrence among West Africans. 7066 case-control pairs (identified in the SIREN study) were randomly partitioned into testing and validation datasets (85:15), and conditional logistic regression models were developed from seventeen traditional factors of stroke in the training dataset. Significant factors were assigned constant, statistically weighted (based on regression coefficient), and receiver operating characteristics curves were constructed (from the testing dataset) to estimate cut-off points for discriminate stroke cases from healthy controls (to build an additive risk scoring model for stroke). Sensitivity, specificity, positive and negative predicted values were estimated to assess the strength of the stroke risk prediction model at P < 0.05. Cohen's kappa for validity was maximal at a total risk score of 56% using statistical weighting approaches for risk quantification in both datasets after identifying 15 traditional vascular factors for stroke occurrence. The stroke risk prediction model had a strong predictive accuracy (76%, 95%CI: 74%–79%), sensitivity (80.3%), specificity (63.0%), positive predictive value (68.5%) and negative predictive value (76.2%) in the validation dataset. The Afrocentric stroke risk models had good accuracy and might be helpful for stroke prediction as a viable approach in de-escalating the rising burden of stroke across multi-ethnic indigenous African populations.

Age Specific Models to Capture the Change in Risk Factor Contribution by Age to Short Term Primary Ischemic Stroke Risk.

Age is one of the most important risk factors when it comes to stroke risk prediction. However, including age as a risk factor in a stroke prediction model can give rise to a number of difficulties. Age often dominates the risk score, and also not all risk factors contribute proportionally to stroke risk by age. In this study we investigate a number of common stroke risk factors, using Framingham heart study data from the NHLBI Biologic Specimen and Data Repository Information Coordinating Center to determine if they appear to contribute proportionally by age to a stroke risk score. As we find evidence that there is some non-proportionality by age, we then create a set of logistic regression risk models that each predict the 5 year stroke risk for a different age group. The age group models are shown to be better calibrated when compared to a model for all ages that includes age as a risk factor. This suggests that to get better predictions for stroke risk it may be necessary to consider alternative methods for including age in stroke risk prediction models that account for the non-proportionality of the other risk factors as age changes.

Stroke Risk Prediction With Hybrid Deep Transfer Learning Framework

Stroke has become a leading cause of death and long-term disability in the world with no effective treatment. Deep learning-based approaches have the potential to outperform existing stroke risk prediction models, but they rely on large well-labeled data. Due to the strict privacy protection policy in health-care systems, stroke data is usually distributed among different hospitals in small pieces. In addition, the positive and negative instances of such data are extremely imbalanced. Transfer learning can solve small data issue by exploiting the knowledge of a correlated domain, especially when multiple source of data are available. In this work, we propose a novel Hybrid Deep Transfer Learning-based Stroke Risk Prediction (HDTL-SRP) scheme to exploit the knowledge structure from multiple correlated sources (i.e., external stroke data, chronic diseases data, such as hypertension and diabetes). The proposed framework has been extensively tested in synthetic and real-world scenarios, and it outperforms the state-of-the-art stroke risk prediction models. It also shows the potential of real-world deployment among multiple hospitals aided with 5G/B5G infrastructures.

The Construction of a Risk Prediction Model Based on Neural Network for Pre-operative Acute Ischemic Stroke in Acute Type A Aortic Dissection Patients.

Objective: To establish a pre-operative acute ischemic stroke risk (AIS) prediction model using the deep neural network in patients with acute type A aortic dissection (ATAAD).Methods: Between January 2015 and February 2019, 300 ATAAD patients diagnosed by aorta CTA were analyzed retrospectively. Patients were divided into two groups according to the presence or absence of pre-operative AIS. Pre-operative AIS risk prediction models based on different machine learning algorithm was established with clinical, transthoracic echocardiography (TTE) and CTA imaging characteristics as input. The performance of the difference models was evaluated using the receiver operating characteristic (ROC), precision-recall curve (PRC) and decision curve analysis (DCA).Results: Pre-operative AIS was detected in 86 of 300 patients with ATAAD (28.7%). The cohort was split into a training (211, 70% patients) and validation cohort (89, 30% patients) according to stratified sampling strategy. The constructed deep neural network model had the best performance on the discrimination of AIS group compare with other machine learning model, with an accuracy of 0.934 (95% CI: 0.891–0.963), 0.921 (95% CI: 0.845–0.968), sensitivity of 0.934, 0.960, specificity of 0.933, 0.906, and AUC of 0.982 (95% CI: 0.967–0.997), 0.964 (95% CI: 0.932–0.997) in the training and validation cohort, respectively.Conclusion: The established risk prediction model based on the deep neural network method may have the big potential to evaluate the risk of pre-operative AIS in patients with ATAAD.

Developing a Stroke Risk Prediction Model Using Cardiovascular Risk Factors: The Suita Study

Introduction: Stroke remains a major cause of death and disability in Japan and worldwide. Detecting individuals at high risk for stroke to apply preventive approaches is recommended. This study aimed to develop a stroke risk prediction model among urban Japanese using cardiovascular risk factors. Methods: We followed 6,641 participants aged 30–79 years with neither a history of stroke nor coronary heart disease. The Cox proportional hazard model estimated the risk of stroke incidence adjusted for potential confounders at the baseline survey. The model’s performance was assessed using the receiver operating characteristic curve and the Hosmer-Lemeshow statistics. The internal validity of the risk model was tested using derivation and validation samples. Regression coefficients were used for score calculation. Results: During a median follow-up duration of 17.1 years, 372 participants developed stroke. A risk model including older age, current smoking, increased blood pressure, impaired fasting blood glucose and diabetes, chronic kidney disease, and atrial fibrillation predicted stroke incidence with an area under the curve = 0.76 and p value of the goodness of fit = 0.21. This risk model was shown to be internally valid (p value of the goodness of fit in the validation sample = 0.64). On a risk score from 0 to 26, the incidence of stroke for the categories 0–5, 6–7, 8–9, 10–11, 12–13, 14–15, and 16–26 was 1.1%, 2.1%, 5.4%, 8.2%, 9.0%, 13.5%, and 18.6%, respectively. Conclusion: We developed a new stroke risk model for the urban general population in Japan. Further research to determine the clinical practicality of this model is required.

Accurate Prediction of Stroke for Hypertensive Patients Based on Medical Big Data and Machine Learning Algorithms: Retrospective Study.

BackgroundStroke risk assessment is an important means of primary prevention, but the applicability of existing stroke risk assessment scales in the Chinese population has always been controversial. A prospective study is a common method of medical research, but it is time-consuming and labor-intensive. Medical big data has been demonstrated to promote disease risk factor discovery and prognosis, attracting broad research interest.ObjectiveWe aimed to establish a high-precision stroke risk prediction model for hypertensive patients based on historical electronic medical record data and machine learning algorithms.MethodsBased on the Shenzhen Health Information Big Data Platform, a total of 57,671 patients were screened from 250,788 registered patients with hypertension, of whom 9421 had stroke onset during the 3-year follow-up. In addition to baseline characteristics and historical symptoms, we constructed some trend characteristics from multitemporal medical records. Stratified sampling according to gender ratio and age stratification was implemented to balance the positive and negative cases, and the final 19,953 samples were randomly divided into a training set and test set according to a ratio of 7:3. We used 4 machine learning algorithms for modeling, and the risk prediction performance was compared with the traditional risk scales. We also analyzed the nonlinear effect of continuous characteristics on stroke onset.ResultsThe tree-based integration algorithm extreme gradient boosting achieved the optimal performance with an area under the receiver operating characteristic curve of 0.9220, surpassing the other 3 traditional machine learning algorithms. Compared with 2 traditional risk scales, the Framingham stroke risk profiles and the Chinese Multiprovincial Cohort Study, our proposed model achieved better performance on the independent validation set, and the area under the receiver operating characteristic value increased by 0.17. Further nonlinear effect analysis revealed the importance of multitemporal trend characteristics in stroke risk prediction, which will benefit the standardized management of hypertensive patients.ConclusionsA high-precision 3-year stroke risk prediction model for hypertensive patients was established, and the model's performance was verified by comparing it with the traditional risk scales. Multitemporal trend characteristics played an important role in stroke onset, and thus the model could be deployed to electronic health record systems to assist in more pervasive, preemptive stroke risk screening, enabling higher efficiency of early disease prevention and intervention.

Prevalence, Mechanisms, and Management of Ischemic Stroke in Heart Failure Patients.

Heart failure (HF) and stroke, two of the major causes of death worldwide, are closely associated. Although atrial fibrillation (AF), which occurs in more than half of patients with HF, is a major risk factor for stroke, there is a great deal of evidence that HF itself increases the risk of stroke independent of AF. The main mechanism of stroke appears to be thromboembolism. However, previous studies have failed to demonstrate the benefit of warfarin in patients with HF without AF, as the benefit of stroke prevention was counteracted by the increased incidence of major bleeding. Recently, researchers have identified patients with HF at a particularly high risk for stroke who may benefit from anticoagulation therapy. Based on stroke-risk prediction models, it may be possible to make better stroke prevention decisions for patients with HF. Moreover, non-vitamin K oral anticoagulants have emerged as anticoagulants with a more favorable risk-benefit profile than warfarin. Future studies on selecting high-risk patients and using more appropriate antithrombotics will lead to improved management of patients with HF.