Machine learning assisted feature identification and prediction of hemodynamic endpoints using computed tomography in patients with CTEPH

• DOI: 10.1007/s10554-023-03026-2
• Publication Date: 2023-12-24T20:01:35Z
• Authors: Joshua Gawlitza; Sophie Endres; Peter Fries; Markus Graf; Heinrike Wilkens; Jonas Stroeder; Arno Buecker; Alexander Massmann; Sebastian Ziegelm...
• Keywords: Univariate analysis

Abstract Chronic thromboembolic pulmonary hypertension (CTEPH) is a rare but potentially curable cause of (PH). Currently PH diagnosed by right heart catheterisation. Computed tomography (CT) used for ruling out other causes and operative planning. This study aims to evaluate importance different quantitative/qualitative imaging features develop supervised machine learning (ML) model predict hemodynamic risk groups. 127 Patients with CTEPH who received preoperative catheterization thoracic CTA examinations (39 ECG-gated; 88 non-ECG gated) were included. 19 qualitative/quantitative 3 parameters [mean artery pressure, atrial pressure (RAP), oxygen saturation (PA SaO2)] gathered. Diameter-based CT measured in axial adjusted multiplane reconstructions (MPR). Univariate analysis was performed qualitative quantitative features. A random forest algorithm trained on Feature calculated all models. Qualitative showed no significant differences between ECG gated CTs. Depending reconstruction plane, five significantly different, mean absolute difference (MPR vs. axial) 0.3 mm correlation parameters. moderate strong multiple The achieved an AUC score 0.82 the mPAP based stratification 0.74 PA SaO2 stratification. Contrast agent retention hepatic vein, mosaic attenuation pattern ratio atrium/left ventricle most important among Quantitative correlate patients—regardless MPR adaption. Machine can be non-invasive seem more than previously anticipated.