Prognostic Model for Long-Term Cardiac Function After Pulmonary Embolism Based on Dynamic Electrocardial Signal and Circulating Biomarkers

Overview

Pulmonary embolism (PE) is a highly morbid and fatal cardiovascular disease. Right ventricular dysfunction (RVD) secondary to PE indicates a poor prognosis and serves as a critical basis for risk stratification. Recent studies have shown that over one-third of patients continue to experience RVD one year after PE, with the mechanisms and regression remaining unclear. Although electrocardiography (ECG) is the most commonly used test for cardiac disease, its diagnostic specificity for PE is limited. In recent years, artificial intelligence (AI) has successfully extracted hundreds of features from data that are difficult for the human eye to recognize. The correlation between daily vital signs monitored by wearable devices and functional signs of chronic cardiovascular disease suggests the potential of AI in detecting disease progression. There is a lack of specific markers for right ventricular function post-PE, and the significance and changes of these markers in disease progression have not yet been explored. This study aims to develop a predictive model for the progression of RVD after PE using AI, combining electromyography, wearable devices, and vitality markers. In this prospective cohort study, 500 patients with acute PE at intermediate or higher risk were enrolled. Approximately 200 patients with RVD at discharge were followed for one year, with daily electromyographic data collected using portable electromyographs. Biospecimens were collected at the following time points: admission, discharge, and follow-up at 3, 6, and 12 months and a variety of inflammatory markers were measured using a multifactorial assay on liquid suspension cores. These data were integrated into a continuous disease diagnostic model based on a deep learning restrictive updating strategy. Ultimately, a continuous disease diagnosis and prognosis algorithm was developed, yielding a model for predicting the progression of RVD after PE using multifactorial assays on liquid suspension cores to measure various inflammatory markers.

Long-term functional impairment after acute PE has been increasingly concerned in recent years. Our previous meta-analysis indicated that 34% PE patients had RVD at 1 year after an acute episode. However, the mechanism and prognosis of long-term RVD are unknown but largely influence patients' life expectancy and quality. In recent years, hundreds of ECG features have been successfully identified by the development of artificial intelligence (AI) and electrocardio signal monitored by wearable devices have also been used to identify cardiac disease and may be promising in detecting potential manifestations for long-term cardiac function in PE patients. Inflammation is known to have an important role in RVD after PE, but the prognostic predicting value have not yet been explored, especially time-variant changes. Therefore, this study is to obtain a prognostic model to predict the occurrence and outcome of long-term RVD after PE, based on artificial intelligence and wearable devices, combining dynamic changes of ECG and biomarkers. 50 patients with acute intermediate and higher risk PE will be prospectively recruited. ECG signal will be collected by a wearable single-lead long-range ECG acquisition system during hospitalization. And those with RVD at discharge (approximately 20 patients) are followed up for 1 year after discharge. Daily ECG data will be collected using a portable ECG monitor device. Blood and urine samples will be obtained at the following time points: admission, discharge, and follow-up at 3, 6, and 12 months, to measure time-variant inflammatory markers using a multiplex immunoassay for inflammatory cytokines quantitation. According to baseline ECG, biomarkers and clinical features, a model based on deep learning algorithm predicting RVD at discharge in study population will be obtained.