Deep Learning to Detect Pulmonary Hypertension from the Chest X-Ray Images of Patients with Systemic Sclerosis

Abstract

Pulmonary hypertension (PH) is a serious prognostic complication in patients with systemic sclerosis (SSc). Deep learning models can be applied to detect PH in the chest X-ray images of these patients. The aim of the study was to investigate the performance and prognostic implications of a deep learning algorithm for the diagnosis of PH in SSc patients using chest X-ray images.

Chest X-ray images were acquired from 230 SSc patients with suspected PH who underwent chest X-ray and right heart catheterization (RHC). A convolutional neural network was trained to identify the data of patients with PH (mean pulmonary arterial pressure > 20 mmHg). Kaplan-Meier analysis was used to evaluate survival. The area under the receiver operating characteristic curve (AUC) obtained with the deep learning algorithm was 0.826 while the AUC obtained with cardiologist assessments of the same images was 0.804. The 5-year prognosis was 83.4% in patients with PH detected by RHC, and 85% in those with PH detected by the model.

The deep learning model developed in this study can detect PH from the chest X-ray data of SSc patients. The prognostic accuracy of the model was demonstrated as well.