Electrocardiogram Detection of Pulmonary Hypertension Using Deep Learning.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• Pulmonary hypertension (PH) is a serious condition often diagnosed late.
• Early detection of PH is crucial for effective management and improved patient outcomes.
• Current diagnostic methods can be time-consuming, leading to delays in treatment initiation.

Purpose of the Study:

• To evaluate the efficacy of a deep learning approach in detecting PH using only electrocardiogram (ECG) data.
• To determine if this AI-driven method can identify clinically significant PH subtypes.
• To assess the potential of ECG-based deep learning to expedite PH diagnosis.

Main Methods:

• A deep convolutional neural network was developed and trained on 12-lead ECG voltage data from adult patients.
• The study retrospectively analyzed data from patients with and without PH, confirmed by right heart catheterization or echocardiogram.
• The model was validated on a separate test dataset, including ECGs obtained up to two years prior to diagnosis.

Main Results:

• The deep learning model demonstrated high accuracy in detecting PH (AUC 0.89), precapillary PH (AUC 0.91), and pulmonary arterial hypertension (AUC 0.88).
• The algorithm successfully identified PH subtypes, including Group 3 PH (AUC 0.80).
• ECGs acquired up to two years before diagnosis were also accurately analyzed by the model (AUC ≥ 0.79).

Conclusions:

• A deep learning algorithm utilizing ECG data can effectively detect pulmonary hypertension and its subtypes.
• This AI approach can identify PH even from ECGs taken up to two years before formal diagnosis.
• The findings suggest a significant potential for reducing diagnostic delays in pulmonary hypertension management.