Pulmonary Hypertension Detection

This AI algorithm was developed to predict pulmonary hypertension (PH) using electrocardiography (ECG) and validated in a historical cohort study across multiple hospitals. The algorithm, which uses an ensemble neural network, demonstrated high performance with an area under the receiver operating characteristic curve of 0.859 in internal validation and 0.902 in external validation. The AI tool accurately identified patients at higher risk for developing PH, and its sensitivity map indicated key ECG wave components (S-wave, P-wave, and T-wave) were critical for its predictions.