Machine Learning–Based Predictive Model for Post-Stroke Dementia

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: March 28, 2024

Backgound: Post-stroke dementia (PSD), a common complication, diminishes rehabilitation efficacy and affects disease prognosis in stroke patients. Many factors may be related to PSD, including demographic, comorbidities, examination characteristics. However, most existing methods are qualitative evaluations of independent factors, which ignore the interaction amongst various factors. Therefore, purpose this study is explore applicability machine learning for predicting PSD. Methods: 9 acceptable features were screened out by Spearman correlation analysis Boruta algorithm. We developed evaluated 8 (ML) models: logistic regression, elastic net, k-nearest neighbors, decision tree, extreme gradient boosting, support vector machine, random forest, multilayer perceptron. Results: A total 539 patients included study. Among models used predict boosting forest showed highest area under curve (AUC), with values 0.7287 0.7285, respectively. The important PSD age, high sensitivity C-reactive protein, side location, occurrence cerebral hemorrhage. Conclusion: Our findings suggest that ML models, especially can best risk

Language: Английский

---

Functional connectivity and graph theory of impaired central visual pathways in acute ischemic stroke based on fMRI

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: April 19, 2024

Abstract In the study of this paper, we first performed analysis whole brain static functional connectivity, divided into 90 regions interest (ROIs) by applying AAL mapping, compared connectivity 14 patients and 26 healthy volunteers (HC) who completed 3-months experiment (3months), 7-days (7days), 12 ( HC), 7-day 3-month (3months) were analysed for whole-brain in all three groups, ROIs mapped to Yeo7 network analysis. sFC analyses revealed significant alterations patients' VAN, DMN networks. Secondly, dynamic based on mapping with sliding window method separately, identified two pattern characteristics, i.e., state 1 a dominated high-frequency weak 2 low-frequency strong connectivity.Stroke spent significantly more time 1, number switches stroke 7days higher likely switch mode 2. Significant changes observed DMN, VIS, FPN, LIM. Finally, built five machine learning models SFC features that differ between namely linear support vector (SVM), radial basis function (SVM-RBF), k nearest neighbours (KNN), random forest (RF), decision tree (TREE). Based maximum AUC optimal feature subset found within LIM networks contributed classification AIS HCs alike.The variation FC may provide new insights neural mechanisms patients.

Language: Английский

---

Forecasting the Survival and Mortality of Patients by Machine Learning Trained on Heart Failure Clinical Imbalanced Data

Published: March 14, 2024

Language: Английский

---

A survey on common biostatistics tools in neuroscience: Machine learning and Bayesian modeling

Advances in Engineering Technology Research, Journal Year: 2024, Volume and Issue: 9(1), P. 650 - 650

Published: Jan. 25, 2024

Machine learning was characterized by building models and finding correlations between data features, while logistic regression, decision trees, support vector machines (SVM), random forest (RF) neural networks were recognized as common ML approaches. Bayesian modeling model uncertainty, which can estimate the features from dataset directly instead of sampling distribution. Their roles extremely useful for detection progression diseases in neuroscience. This review summarize different approaches various diseases, hoping to introduce potential biostatistics tools

Language: Английский

---

Predictive models for secondary epilepsy within 1 year in patients with acute ischemic stroke: a multicenter retrospective study

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: June 28, 2024

Abstract Objective Post-stroke epilepsy (PSE) is a major complication that worsens both prognosis and quality of life in patients with ischemic stroke. This study aims to develop an interpretable machine learning model predict PSE using medical records from four hospitals Chongqing. Methods We collected analyzed records, imaging reports, laboratory test results 21,459 diagnosed Traditional univariable multivariable statistical analyses were performed identify key predictive factors. The dataset was divided into 70% training set 30% testing set. To address class imbalance, the Synthetic Minority Oversampling Technique combined Edited Nearest Neighbors used. Nine widely applied algorithms evaluated compared relevant prediction metrics. SHAP (SHapley Additive exPlanations) used interpret model, assessing contributions different features. Results Regression showed complications such as hydrocephalus, cerebral hernia, deep vein thrombosis, well brain regions (frontal, parietal, temporal lobes), significantly contributed PSE. Factors like age, gender, NIH Stroke Scale (NIHSS) scores, WBC count D-dimer levels associated higher risk Among models, tree-based methods Random Forest, XGBoost, LightGBM demonstrated strong performance, achieving AUC 0.99. Conclusion Our successfully predicts risk, models showing superior performance. NIHSS score, count, identified most important predictors.

Language: Английский

---

Specific Mode Electroacupuncture Stimulation Mediates the Delivery of NGF Across the Hippocampus Blood–Brain Barrier Through p65-VEGFA-TJs to Improve the Cognitive Function of MCAO/R Convalescent Rats

Molecular Neurobiology, Journal Year: 2024, Volume and Issue: unknown

Published: July 12, 2024

Abstract Cognitive impairment frequently presents as a prevalent consequence following stroke, imposing significant burdens on patients, families, and society. The objective of this study was to assess the effectiveness underlying mechanism nerve growth factor (NGF) in treating post-stroke cognitive dysfunction rats with cerebral ischemia–reperfusion injury (MCAO/R) through delivery into brain using specific mode electroacupuncture stimulation (SMES). From 28th day after modeling, were treated NGF mediated by SMES, function observed treatment. Learning memory ability evaluated behavioral tests. impact SMES blood–brain barrier (BBB) permeability, enhancement MCAO/R, including transmission electron microscopy, enzyme-linked immunosorbent assay, immunohistochemistry, immunofluorescence, TUNEL staining. We reported that demonstrates safe efficient open BBB during ischemia repair phase, facilitating p65-VEGFA-TJs pathway. Graphical By Figdraw

Language: Английский

---

Investigate the therapeutic differences between Temporal Interference Stimulation and Transcranial Alternating Current Stimulation on Post-stroke cognitive dysfunction: A Protocol for Clinical Trial

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: July 18, 2024

Introduction Transcranial alternating current stimulation (tACS) and temporal interference (TIS) as electrical neuromodulation therapy, have shown promising applications in cognitive impairments. Meanwhile TIS technique is more novel with deep non-invasive brain . At present, the therapeutic or differences between tACS on Post-stroke dysfunction(PSCI) still unclear. Here, we aim to compare analysis model clinical performances of tACS. Methods analysis The prospective, single-blind randomized controlled trial will be conducted over a two-week period. Through precise statistical sample size calculation,thirty-six eligible participants mild PSCI recruited randomly allocated either group. Participants group receive at frequencies 2005Hz 2010Hz hippocampus target(in hippocampal region). Those undergo 5Hz dorsolateral prefrontal cortex (DLPFC). intervention last for two weeks, each receiving 25-minute sessions once day, five times per week. primary outcome measure Montreal assessment (MoCA), while secondary outcomes include performance N-back task, digital span test (DST), shape trails (STT) functional near-infrared spectroscopy (fNIRS). All assessments collected time points: pre-intervention (T1) post-intervention (T2). Trial registration protocol registered www.chictr.org.cn under registration number ChiCTR2400081207.Registered February 26, 2024.

Language: Английский

---

Prediction of cognitive impairment among Medicare beneficiaries using a machine learning approach

Archives of Gerontology and Geriatrics, Journal Year: 2024, Volume and Issue: 128, P. 105623 - 105623

Published: Sept. 6, 2024

Language: Английский

---

Development and validation of domain-specific clinical prediction models of post-stroke cognitive impairment

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 6, 2024

Background Post-stroke cognitive impairment (PSCI) is highly prevalent across multiple domains. Individualised PSCI prognosis has mainly been researched using global outcomes. Here, we develop and externally validate clinical prediction models for overall domain-specific PSCI, including language, memory, attention, executive function, numeracy, praxis. Methods N =430 stroke survivors completed the Oxford Cognitive Screen (OCS) in acute care at 6-month follow-up (binarized outcome; impaired vs unimpaired). Logistic regression were fitted comprising both mandatory clinically-relevant (age, sex, severity, education, hemisphere, PSCI) data-driven (acute mood difficulties, length of stay care, multimorbidity) predictors backward elimination ( p < 0.10) on multiply imputed data. Internal validation used bootstrapping to obtain optimism-adjusted performance estimates. External C-Slope as a uniform shrinkage factor. Results Compared model (C-Statistic=0.76 [95% CI=0.71–0.80]), comparable or improved was observed language (C-Statistic=0.77 CI=0.72–0.81]) memory (C-Statistic=0.72 CI=0.65–0.75]), attention (C-Statistic=0.74 [0.69–0.78]). Numeracy (C-Statistic=0.69 CI=0.63–0.74]), function (C-Statistic=0.71 CI=0.65–0.76]), praxis (C-Statistic=0.60 CI=0.53–0.65]) showed weaker performance. In external validation, development data CI=0.67–0.79]). Conclusions Domain-specific have potential offer more meaningful prognoses compared models. show promise different severity cohorts. Future recalibration would be beneficial.

Language: Английский

---

Predictive models for secondary epilepsy within 1 year in patients with acute ischemic stroke: a multicenter retrospective study

Published: Oct. 15, 2024

Post-stroke epilepsy (PSE) is a significant complication that has negative impact on the prognosis and quality of life ischemic stroke patients. We collected medical records from 4 hospitals in Chongqing created an interpretable machine learning model for prediction.We records, imaging reports, laboratory tests 21459 patients with diagnosis stroke. conducted traditional univariable multivariable statistics analyses to compare identify important features. Then data was divided into 70% training set 30% testing set. employed Synthetic Minority Oversampling Technique combined Edited Nearest Neighbors method resample imbalanced dataset Nine commonly used methods were build models, relevant prediction metrics compared select best-performing model. Finally, we SHAP(SHapley Additive exPlanations) interpretability analysis, assessing contribution clinical significance different features prediction.In regression complications such as hydrocephalus, cerebral hernia, uremia, deep vein thrombosis; brain regions included involvement cortical including frontal lobe, parietal occipital temporal subcortical region basal ganglia, thalamus so contributed PSE. General age, gender, National Institutes Health Stroke Scale score, well indicators WBC count, D-dimer, lactate, HbA1c associated higher likelihood Patients conditions fatty liver, coronary heart disease, hyperlipidemia, low HDL had developing The particularly tree models Random Forest, XGBoost, LightGBM, demonstrated good predictive performance AUC 0.99.The built large can effectively predict PSE, tree-based performing best. NIHSS count D-dimer found have greatest impact.

Language: Английский

---