Patients with pulmonary hypertension (PH) represent a challenging and heterogeneous cohort with high morbidity and mortality if left untreated. To make a definitive diagnosis of PH, one needs to conduct an invasive right heart catheterization (RHC) in order to assess the mean pulmonary artery pressure (mPAP). As PH occurs sporadically in various medical conditions, including connective tissue disease, and congenital heart disease, and presenting symptoms are non-specific, there is a need to raise the suspicion of PH early in the community. For this reason, noninvasive tools that are widely available for upfront screening would be ideal to enable timely diagnosis of PH. Transthoracic echocardiography has emerged as the mainstay for screening of PH, yet the sensitivity and specificity of this approach remain limited even in experienced hands. As high-throughput technologies advance and access to PH big data improve, it will be critical to prudently select artificial intelligence approaches for data analysis, visualization, and interpretation. By combining the multimodal clinical data (such as indicators from chest X-ray, electrocardiography, and echocardiography), this study aims to develop artificial intelligence-assisted programs to assist the screening and diagnosis of PH, and to evaluate its diagnostic accuracy for PH as compared with RHC, and to estimate whether this approach outperforms the conventional echocardiographic method.
