Deep learning for the identification of pre-and post-capillary pulmonary hypertension on cine MRI

Background: The aim of the present study was to develop a deep learning (DL) framework that can identify elevated left heart pressure in pulmonary hypertension (PH) based on cine MRI-derived left ventricular (LV) motion/deformation patterns. Methods: Fifty-four PH patients (23 males, 58.9±13.5 years old) were retrospectively included in the present study. Heart deformation analysis (HDA) was applied to acquire LV displacement, velocity strain and strain rate on cine MRI datasets. Peak values of motion/deformation indices at early and late diastole entered an artificial neural network (ANN), which was developed with Python, to discriminate cases of pre-capillary PH [defined as mean pulmonary arterial pressure (mPAP) ≥25 mmHg and pulmonary capillary wedge pressure (PCWP) ≤15 mmHg] from post-capillary PH (mPAP ≥25 mmHg and PCWP >15 mmHg). Results: Cine MRI datasets of 54 PH patients were eligible for HDA processing. Peak radial and circumferential displacement, velocity and strain rates in systole, early and late diastole were extracted. The ANN model was fit and trained with cine MRI-derived indices from 40 randomly chosen PH patients. Then, the model successfully identified the type of PH in the remaining 14 patients. The accuracy, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) to discriminate post-capillary PH from pre-capillary PH were 86%, 83%, 88%, 83% and 88%, respectively. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was 0.85 [95% confident interval (CI): 0.629–1]. Conclusions: DL can identify elevated left heart pressure underlying post-capillary PH based on LV motion/deformation patterns presented on regular cine MRI datasets.