3 September 2023 in The European Research Journal
Objective: The aim of this study was to investigate whether there is subtle cardiac dysfunction in patients with chronic venous insufficiency. Methods: Age and sex matched 56 patients with a score of C3 and above in the Clinical, Etiological, Anatomical, Pathophysiological classification and 56 healthy volunteers were included in the study. All subjects were evaluated by detailed echocardiographic examination, including two-dimensional strain echocardiographic analysis by speckle tracking method. Results: Mitral E wave deceleration time (EDT), E and A wave velocity, E/e' ratio for left ventricle, tricuspid EDT, E/e’ ratio for right ventricle and systolic pulmonary artery pressure were found high as significant statistically in patients groups (p < 0.05). But no any statistically significant difference was observed in other parameters between two groups. Conclusions: There may be an increase in diastolic filling pressures in patients with chronic venous insufficiency due to the increased preload in the supine position. This condition seems to be clinically important in patients at high risk for heart failure due to the presumption of the early treatment of chronic venous insufficiency may reduce the risk of heart failure evolvement.
7 October 2026 in Pattern Recognition Letters
24 October 2026 in Journal of Electrocardiology
17 February 2026 in SSRN Electronic Journal
Background and Objectives: Quantitative coronary angiography (QCA) requires accurate identification of coronary artery segments for lesion characterization and clinical scoring systems such as SYNTAX. Manual segment identification is time-consuming and subject to inter-observer variability, particularly when vessel appearance is ambiguous due to projection geometry. This study aims to develop a deep learning framework that integrates X-ray viewing angle information for improved automated coronary segment classification. Methods: We propose AngleAwareNet, a deep learning framework that integrates X-ray viewing angle information (primary LAO/RAO and secondary CRA/CAU angles) for automated coronary segment classification. The architecture employs a ResNet-18 encoder with a multi-task learning strategy combining vessel segmentation and segment classification. Viewing angles extracted from DICOM metadata are encoded through a learned embedding and fused with image features for classification. The model was trained on a combined dataset comprising lesion-centric ROI images from a single-center hospital dataset merged with the ARCADE training subset, and externally validated on the held-out ARCADE test set. Results: On internal validation, AngleAwareNet achieved 94.35% classification accuracy and 0.8028 Dice coefficient for segmentation. Ablation studies demonstrated that viewing angle integration improved overall accuracy by 1.01% compared to the baseline without angles (93.34%). External validation on the ARCADE benchmark showed a more pronounced benefit: the angle-aware model achieved 89.29% overall accuracy compared to 86.48% for the baseline (+2.81%), with substantial improvements for RCA (+6.9%), LAD (+4.7%), and LCX (+2.8%) classes. Conclusions: Integration of X-ray viewing geometry significantly improves automated coronary segment identification, particularly for anatomically adjacent vessels. The proposed lesion-centric approach is well-suited for integration into QCA workflows, enabling automated segment labeling for clinical reporting and SYNTAX score calculation.