Pulmonary hypertension (PH) is a complex disease classified into five groups (I-V) by the European Society of Cardiology/European Respiratory Society (ESC/ERS) guidelines. Chest contrast-enhanced computed tomography (CECT) is crucial in the non-invasive PH assessment. This study aimed to develop a machine learning (ML)-based educational resource for classifying PH cases via CECT according to ESC/ERS groups. We retrospectively included 172 PH patients who underwent CECT at two University Hospitals (Udine and Brescia). Three chest-devoted radiologists independently reviewed the CECTs, reporting on 13 features, including lung conditions, heart abnormalities, chronic thromboembolism, and mediastinal findings. Readers assigned the features as absent/present except for the left atrium (LA) anteroposterior diameter (measured in millimeters) and classified PH cases I-V with likelihood scores (1-100 %) for each group. The majority decisions for features and average LA diameter were used as ML inputs. The highest average likelihood scores determined group assignments, serving as ground truth. Various ML algorithms were tested using the Weka software and evaluated by accuracy, area under the ROC curve (AUROC), and F1-score. After excluding three group V patients to avoid imbalance, the Naïve-Bayes algorithm showed 0.72 accuracy, 0.84 AUROC, and 0.72 F1-score. Accuracy values for group I-IV were 0.75, 0.78, 0.51, 0.79; AUROC values were 0.78, 0.84, 0.86, 0.87; F1-scores were 0.63, 0.79, 0.61, 0.84, respectively. This study is the first to develop an ML-driven tool for classifying PH via chest CECT. While performance metrics require improvement, including the need for a larger sample size, the resource can potentially train non-dedicated radiologists in PH classification, supporting multidisciplinary reasoning.

Background: Negative CCTA can effectively exclude significant CAD, eliminating the need for further noninvasive or invasive testing. However, in the presence of severe CAD, the accuracy declines, thus necessitating additional testing. The aim of our study was to evaluate the diagnostic performance of noninvasive cFFR derived from CCTA, compared to ICA in detecting hemodynamically significant stenoses in participants with high CAC scores (>400). Methods: This study included 37 participants suspected of having CAD who underwent CCTA and ICA. CAC was calculated and cFFR analyses were performed using an on-site machine learning-based algorithm. Diagnostic accuracy parameters of CCTA and cFFR were calculated on a per-vessel level. Results: The median total CAC score was 870, with an IQR of 642–1370. Regarding CCTA, sensitivity and specificity for RCA were 60% and 67% with an AUC of 0.639; a LAD of 87% and 50% with an AUC of 0.688; an LCX of 33% and 90% with an AUC of 0.617, respectively. Regarding cFFR, sensitivity and specificity for RCA were 60% and 61% with an AUC of 0.606; a LAD of 75% and 54% with an AUC of 0.647; an LCX of 50% and 77% with an AUC of 0.647. No significant differences between AUCs of coronary CTA and cFFR for each vessel were found. Conclusions: Our results showed poor diagnostic accuracy of CCTA and cFFR in determining significant ischemia-related lesions in participants with high CAC scores when compared to ICA. Based on our results and study limitations we cannot exclude cFFR as a method for determining significant stenoses in people with high CAC. A key issue is accurate and detailed lumen segmentation based on good-quality CCTA images.