Translational application of a self-organized deep feature engineering pipeline for non-invasive pulmonary hypertension classification from routine chest radiographs
T Kivrak, M Gelen, O Salkin, O Karaca, P D Barua, S Dogan, T Tuncer, R S Tan, M Salvi, V S Subbhuraam, U R AcharyaAbstract
Background
Pulmonary hypertension (PH) causes high mortality and poses diagnostic challenges. Current guidelines require invasive right heart catheterization (RHC) to confirm mean pulmonary artery pressure ≥ 25 mmHg. Delayed diagnosis impairs timely treatment. It is unknown whether standard chest X-rays can stratify PH severity. The study aimed to develop and validate Exemplar MobileNet (ExMobileNet), an explainable AI model that classifies PH into hemodynamic categories from routine chest X-ray images and thus supports non-invasive severity assessment.
Methods
We collected 1,293 de-identified chest X-rays obtained from 2018 to 2023. The cohort
Translational study 135 PH patients confirmed by RHC and 551 healthy controls. We defined seven multi-class tasks for key hemodynamic parameters (mean pulmonary artery pressure, pulmonary vascular resistance, cardiac index, etc.). The ExMobileNet workflow consists of: (i) Feature extraction via MobileNetV2, (ii) Feature selection by neighborhood component analysis and Chi2 feature selectors, (iii) Classification with k-nearest neighbors (kNN) and support vector machines (SVM) and (iv) Decision fusion by majority vote and greedy optimization.
Results
Task-level accuracy ranged from 90.3% to 93.2%. Geometric mean scores ranged from 78.9 % to 85.1 %. Overall sensitivity and specificity were 88.5% and 91.3%, respectively. Mean accuracy across all tasks was 92.0 % (±1.2%). Average inference time was 2.3 ± 0.4 second per image on CPU-only hardware.
Conclusions
ExMobileNet achieves high agreement with RHC-based assessments using routine chest X-rays. This AI tool may enable earlier, non-invasive PH screening in clinical practice.Confusion matrixes of ExMobileNetFor image description, please refer to the figure legend and surrounding text.Computed ROC curvesFor image description, please refer to the figure legend and surrounding text.