Multimodal data integration for enhanced longitudinal prediction for cardiac and cerebrovascular events following initial diagnosis of obstructive sleep apnea syndrome

Multimodal Visualization and Explainable Machine Learning-Driven Markers Enable Early Identification and Prognosis Prediction for Symptomatic Aortic Stenosis and Heart Failure With Preserved Ejection Fraction After Transcatheter Aortic Valve Replacement: Multicenter Cohort Study

BACKGROUND

Currently, there is a paucity of literature addressing personalized risk stratification using multimodal data in patients with symptomatic aortic stenosis and heart failure with preserved ejection fraction (HFpEF) following transcatheter aortic valve replacement (TAVR).

OBJECTIVE

This study aimed to enhance the performance of risk assessment models in this patient population by developing a predictive model for adverse outcomes using various machine learning (ML) techniques.

METHODS

This multicenter cohort study included 326 patients diagnosed with severe AS and HFpEF who underwent TAVR between January 2017 and December 2023. Patients were allocated to training (n=195) and independent validation (n=131) sets based on hospital affiliation. A dual-phase feature selection process, combining least absolute shrinkage and selection operator logistic regression and the Boruta algorithm, was used to identify relevant variables from the multimodal dataset. A total of 5 ML model-decision trees, K-nearest neighbors, random forest, support vector machine, and extreme gradient boosting were used to construct a visualization and explainable predictive framework to elucidate model decision-making processes.

RESULTS

The primary features identified included age, N-terminal pro-brain natriuretic peptide, fasting blood glucose, triglyceride/high-density lipoprotein cholesterol ratio, triglyceride glucose index, triglyceride glucose-BMI index, atherogenic index of plasma index, and Apolipoprotein B. Among the 5 models, the support vector machine demonstrated the best predictive performance for major adverse cardiovascular and cerebrovascular events in patients with severe AS and HFpEF following TAVR, achieving an area under the curve of 0.756 (95% CI 0.631-0.881) in the independent validation set. The model exhibited good calibration and robust predictive power in both training and validation sets and demonstrated the highest net benefit in decision curve analysis compared to other models. To extract significant variables influencing the algorithm and ensure model appropriateness, we interpreted cohort and personalized model predictions using Shapley Additive Explanations values.

CONCLUSIONS

Our ML-based multimodal model, incorporating 8 readily accessible predictors, demonstrated robust predictive capability for 12 months of major adverse cardiovascular and cerebrovascular events risk. This model can be used to identify high-risk individuals with AS and HFpEF following TAVR, potentially aiding in risk stratification and personalized treatment strategies.

Hyper-DREAM, a Multimodal Digital Transformation Hypertension Management Platform Integrating Large Language Model and Digital Phenotyping: Multicenter Development and Initial Validation Study

Within the mHealth framework, systematic research that collects and analyzes patient data to establish comprehensive digital health archives for hypertensive patients, and leverages large language models (LLMs) to assist clinicians in health management and Blood Pressure (BP) control remains limited. In this study, our aims to describe the design, development and usability evaluation process of a management platform (Hyper-DREAM) for hypertension. Our multidisciplinary team employed an iterative design approach over the course of a year to develop the Hyper-DREAM platform. This platform's primary functionalities encompass multimodal data collection (personal hypertensive digital phenotype archive), multimodal interventions (BP measurement, medication assistance, behavior modification, and hypertension education) and multimodal interactions (clinician-patient engagement and BP Coach component). In August 2024, the mHealth App Usability Questionnaire (MAUQ) was conducted involving 51 hypertensive patients recruited from three distinct centers. In parallel, six clinicians engaged in management activities and contributed feedback via the Doctor's Software Satisfaction Questionnaire (DSSQ). Concurrently, a real-world comparative experiment was conducted to evaluate the usability of the BP Coach, ChatGPT-4o Mini, ChatGPT-4o and clinicians. The comparative experiment demonstrated that the BP Coach achieved significantly higher scores in utility (mean scores 4.05, SD 0.87) and completeness (mean scores 4.12, SD 0.78) when compared to ChatGPT-4o Mini, ChatGPT-4o, and clinicians. In terms of clarity, the BP Coach was slightly lower than clinicians (mean scores 4.03, SD 0.88). In addition, the BP Coach exhibited lower performance in conciseness (mean scores 3.00, SD 0.96). Clinicians reported a marked improvement in work efficiency (2.67 vs. 4.17, P < .001) and experienced faster and more effective patient interactions (3.0 vs. 4.17, P = .004). Furthermore, the Hyper-DREAM platform significantly decreased work intensity (2.5 vs. 3.5, P = .01) and minimized disruptions to daily routines (2.33 vs. 3.55, P = .004). The Hyper-DREAM platform demonstrated significantly greater overall satisfaction compared to the WeChat-based standard management (3.33 vs. 4.17, P = .01). Additionally, clinicians exhibited a markedly higher willingness to integrate the Hyper-DREAM platform into clinical practice (2.67 vs. 4.17, P < .001). Furthermore, patient management time decreased from 11.5 min (SD 1.87) with Wechat-based standard management to 7.5 min (SD 1.84, P = .01) with Hyper-DREAM. Hypertensive patients reported high satisfaction with the Hyper-DREAM platform, including ease of use (mean scores 1.60, SD 0.69), system information arrangement (mean scores 1.69, SD 0.71), and usefulness (mean scores 1.57, SD 0.58). In conclusion, our study presents Hyper-DREAM, a novel artificial intelligence-driven platform for hypertension management, designed to alleviate clinician workload and exhibiting significant promise for clinical application. The Hyper-DREAM platform is distinguished by its user-friendliness, high satisfaction rates, utility, and effective organization of information. Furthermore, the BP Coach component underscores the potential of LLMs in advancing mHealth approaches to hypertension management.

Multimodal data-based longitudinal prognostic model for predicting atrial fibrillation recurrence after catheter ablation in patients with patent foramen ovale and paroxysmal atrial fibrillation

BACKGROUND

Clinical studies on atrial fibrillation (AF) recurrence after catheter ablation in patients diagnosed with patent foramen ovale (PFO) and paroxysmal AF (PAF) are scarce. Here, we aimed to develop a nomogram model utilizing multimodal data for the risk stratification of AF recurrence following catheter ablation in individuals diagnosed with PFO and new-onset PAF.

METHODS

Patients with PFO and PAF who underwent catheter ablation at the Renmin Hospital of Wuhan University from January 2018 to June 2020 were consecutively enrolled. The identification of potential risk factors was conducted using the regression method known as least absolute shrinkage and selection operator. Subsequently, multivariate COX regression analysis was conducted to determine the independent risk factors, after which a nomogram scoring system was developed. The nomogram's performance was assessed via various statistical measures, including receiver operating characteristic curve analysis, calibration curve, and decision curve analysis (DCA).

RESULTS

The dataset was partitioned into the development cohort (n = 102) and the validation cohort (n = 43) using a 7:3 ratio. The constructed nomogram included four clinical variables: age, diabetes mellitus, lipoprotein (a), and right ventricular diameter. The area under the curve values of the development and validation cohorts at 1, 2, and 3 years post-catheter ablation were 0.911, 0.812, and 0.786 and 0.842, 0.761, and 0.785, respectively. Additionally, the nomogram demonstrated a significant correlation between the predicted and actual outcomes in the development and validation cohorts, indicating its excellent calibration. Lastly, the DCA findings suggested that the model had notable clinical applicability in predicting the likelihood of AF recurrence within 1, 2, and 3 years after catheter ablation.

CONCLUSION

The incorporation of multimodal data in a nomogram visualization tool facilitates the concise representation of multimodal data, thereby enhancing the comprehension of the clinical status of patients with PFO and PAF following catheter ablation and providing accurate risk stratification at 1, 2, and 3 years post-treatment.

TRIAL REGISTRATION

This trial was registered in the Chinese Clinical Trial Registry. (ChiCTR2300072320).

Inflammation Biomarker-Driven Vertical Visualization Model for Predicting Long-Term Prognosis in Unstable Angina Pectoris Patients with Angiographically Intermediate Coronary Lesions

Objective

Angina, a prevalent manifestation of coronary artery disease, is primarily associated with inflammation, an established contributor to the pathogenesis of atherosclerosis and acute coronary syndromes (ACS). Various inflammatory markers are employed in clinical practice to predict patient prognosis and optimize clinical decision-making in the management of ACS. This study investigated the prognostic significance of integrating commonly used, easily repeatable inflammatory biomarkers within a multimodal preoperative prediction model in patients presenting with unstable Angina Pectoris (UAP) and intermediate coronary lesions.

Methods

This retrospective analysis included patients diagnosed with UAP and intermediate coronary lesions (50%-70% stenosis) who underwent coronary angiography at our hospital between January 2019 and June 2021. The assessed outcome was the occurrence of major adverse cardiac and cerebrovascular events (MACCEs). The Boruta algorithm was applied to identify potential risk factors and develop a prognostic multimodal model.

Results

A total of 773 patients were enrolled and divided into a training cohort (n=463) and validation cohort (n=310). A nomogram was constructed to predict the probability of MACCE-free survival based on five clinical features: diabetes mellitus, current smoking, history of myocardial infarction, neutrophil-to-lymphocyte ratio, and fasting blood glucose. In the training cohort, the area under the curve values for the nomogram at 24, 32, and 40 months were 0.669, 0.707, and 0.718, respectively, while those in the validation cohort were 0.613, 0.612 and 0.630, respectively. The model demonstrated good calibration in both cohorts with predicted outcomes aligning well with actual results at all time points up to 40 months. Furthermore, decision curve analysis showed significant clinical utility of the model across the specified time intervals.

Conclusion

The developed preoperative prognostic model visually illustrates the association among inflammation, blood glucose level, established risk factors, and long-term MACCEs in UAP patients with intermediate coronary lesions.

Multimodal Data-Driven Prognostic Model for Predicting Long-Term Prognosis in Patients With Ischemic Cardiomyopathy and Heart Failure With Preserved Ejection Fraction After Coronary Artery Bypass Grafting: A Multicenter Cohort Study

BACKGROUND

Limited data from the literature are available to assess the efficacy of coronary artery bypass grafting in patients with ischemic cardiomyopathy and heart failure with preserved ejection fraction. Therefore, our objective was to use machine learning techniques integrating clinical features, biomarker data, and echocardiography data to enhance comprehension and risk stratification in patients diagnosed with ischemic cardiomyopathy and heart failure with preserved ejection fraction who have undergone coronary artery bypass grafting surgery.

METHODS AND RESULTS

For this study, 294 patients with ischemic cardiomyopathy and heart failure with preserved ejection fraction who underwent coronary artery bypass grafting surgery were assigned to the development cohort (n=176) and the independent validation cohort (n=118). A total of 52 clinical variables were extracted for each patient. The principal clinical end point was the incidence of major adverse cardiovascular events, encompassing cardiac mortality, acute myocardial infarction, acute heart failure, and graft failure. From least absolute shrinkage and selection operator regression, 4 predictors were selected for the final prediction nomogram: diabetes, hypertension, the systemic immune-inflammation index, and NT-proBNP (N-terminal pro-B-type natriuretic peptide). The prediction nomogram achieved satisfactory prediction performance in both the development cohort (C index, 0.768 [95% CI, 0.701-0.835]) and independent validation cohort (C index, 0.633 [95% CI, 0.521-0.745]). Adequate calibration was noted for the likelihood of major adverse cardiovascular events in both the development and independent validation cohorts. Decision curve analysis confirmed the clinical usefulness of the established prediction nomogram.

CONCLUSIONS

A clinically feasible prognostic model, based on preoperative multimodal data, was developed for risk stratification of patients with ischemic heart and heart failure with preserved ejection fraction who receive coronary artery bypass grafting surgery.

REGISTRATION

https://www.chictr.org.cn; Unique identifier: ChiCTR2300074439.

Incorporating Inflammation Biomarker-Driven Multivariate Predictive Model for Coronary Microcirculatory Dysfunction in Acute Myocardial Infarction Following Emergency Percutaneous Coronary Intervention

BACKGROUND

Despite patients with successful revascularization as evidenced by angiographic findings, inadequate clinical management of coronary microcirculatory dysfunction (CMD) may result in preventable adverse outcomes. Therefore, it is imperative to use a multimodal data‑driven predictive model for the occurrence of CMD in patients with acute myocardial infarction (AMI) following emergency percutaneous coronary intervention (PCI).

METHODS

A prospective case-control analysis was conducted on a cohort of 77 patients with AMI who underwent PCI. The most informative predictors were selected for the predictive model through the application of LASSO analysis and multi-factor logistic regression. The diagnosis of CMD is based on findings from cardiac magnetic resonance (CMR).

RESULTS

Based on the findings from LASSO analysis and multi-factor logistic regression, variables including sex, neutrophil-to-lymphocyte ratio (NLR), Gensini score, and diabetes mellitus were identified as independent predictors for the development of CMD in AMI patients who underwent emergency PCI. The predictive model was evaluated using bootstrap self-sampling 500 times. The resulting predictive model demonstrated an AUC value of 0.897 (95% CI: 0.827-0.958). The calibration curves exhibited good concordance between the predictions generated by the model and the CMR analysis. Furthermore, decision curve analysis revealed that the predictive model provided valuable clinical benefit in predicting CMD.

CONCLUSIONS

The multivariate predictive model, constructed using readily available clinical variables in patients with AMI who underwent PCI, demonstrates satisfactory predictability for identifying comorbid CMD, thereby facilitating the identification of high-risk patients.