PH-DyPred: A Multimodal Dynamic Risk Prediction Study in Pulmonary Hypertension

Pulmonary hypertension (PH) is a progressive cardiopulmonary disease characterized by elevated pulmonary artery pressure and vascular remodeling, which leads to right heart failure and increased mortality. Despite advances in diagnostics, risk stratification remains limited due to the disease's heterogeneity. This study aims to develop and validate a dynamic risk prediction model for PH by integrating multimodal data-including echocardiography, Cardiac MRI, PET-MR, ECG, biomarkers, and clinical features-using advanced machine learning algorithms. The study will establish a prospective cohort of PH patients to explore predictive markers, stratify prognosis, and provide a scientific basis for early warning and individualized management.

This is a prospective, observational cohort study designed to investigate dynamic risk prediction in patients diagnosed with pulmonary hypertension (PH). The study will collect multimodal clinical data-comprising imaging (echocardiography, cardiac MRI, PET-MR), electrocardiographic parameters, blood-based biomarkers, and demographic and clinical information-at baseline and follow-up intervals. The core objective is to develop a data fusion-based prognostic model capable of predicting adverse outcomes such as hospitalization, functional deterioration, or mortality. Machine learning methods will be employed to identify key predictive features. The model will be validated internally and externally across different subgroups. The study seeks to inform individualized risk-based decision-making and advance precision screening in PH care. In addition, biospecimens will be collected to support comprehensive multi-omics profiling. Whole blood, serum, plasma, urine, and stool samples will be obtained and processed using standardized protocols. Blood-derived samples will be used for genomic, proteomic, metabolomic, and microRNA analyses; urine specimens will support metabolomic and renal biomarker assays; and stool samples will be used for gut microbiome sequencing. All biospecimens will be stored in a secure biobank and linked with clinical, imaging, and longitudinal follow-up data using de-identified subject codes to enable integrated multimodal analyses and facilitate future exploratory investigations of disease mechanisms and biomarker discovery. Health economic evaluation, including cost-effectiveness and budget impact analyses, will be conducted using collected data on healthcare resource utilization, direct medical costs, and clinical outcomes to inform future policy and reimbursement decision-making.