AI-based diagnosis algorithm of pulmonary arterial hypertension using echocardiography

<bold>Background:</bold> Pulmonary arterial hypertension (PAH) is a progressive and potentially life-threatening condition. Echocardiography (EchoCG) is a widely used imaging modality for PAH diagnosis, but accurate interpretation can be challenging due to subjective assessment and interobserver variability. Artificial intelligence (AI) holds promise for improving diagnostic accuracy by leveraging machine learning algorithms to analyze echocardiography parameters. <bold>Methods:</bold> The study involving 156 patients with suspected PAH who underwent EchoCG between 2018 to 2023. EchoCG images were analyzed using a deep learning algorithm trained on a dataset of annotated echocardiograms. The algorithm was developed to detect specific parameters indicative of PAH, including right ventricular hypertrophy, dilatation, and systolic dysfunction. Diagnostic performance metrics, including sensitivity, specificity, and AUC ROC, were calculated to evaluate the algorithm's accuracy in detecting PAH. <bold>Results:</bold> The AI-based diagnostic algorithm demonstrated high sensitivity 78 %, specificity 74 %, and AUC 0.81 for the detection of PAH compared to conventional echocardiography interpretation. The algorithm effectively identified key echocardiography parameters associated with PAH, including right ventricular enlargement and dysfunction, with high precision and consistency. Furthermore, the AI algorithm exhibited robust performance across different patient subgroups, including those with coexisting cardiovascular comorbidities. <bold>Conclusion:</bold> Our study demonstrates the potential of AI-based analysis of echocardiography parameters for improving the diagnosis of PAH.