Huyut Mehmet Tahir, Velichko Andrei, Belyaev Maksim, Izotov Yuriy, Karaoğlanoğlu Şebnem, Sertoğullarından Bünyamin, Keskin Sıddık, Korzun Dmitry
Department of Biostatistics and Medical Informatics, Faculty of Medicine, Erzincan Binali Yıldırım University, Erzincan, 24000, Turkey.
Petrozavodsk State University, 33 Lenin Ave., Petrozavodsk, 185910, Russia.
Sci Rep. 2025 Jul 12;15(1):25262. doi: 10.1038/s41598-025-00274-1.
Right ventricular dysfunction (RVD) is strongly associated with increased mortality in patients with acute pulmonary embolism (PE), making its early detection crucial. Identifying RVD risk factors rapidly, accurately, and economically within the acute PE population could significantly improve diagnosis and treatment, potentially reducing mortality rates. This study evaluates the performance of LogNNet and supervised machine learning (ML) models for diagnosing RVD using a repeated stratified hold-out validation procedure. An ensemble-based LogNNet model is proposed for practical application. The LogNNet model identified gender, coronary artery disease, Comorbid Disease (especially hypertension), age (above 74-years), Thrombus segment and un/bilateral Thrombus as the most significant predictors for RVD diagnosis. Additionally, combinations of these features demonstrated high predictive power. LogNNet achieved robust results with only a few selected features, making it suitable for applications in resource-limited environments. LogNNet provides a practical and accessible tool for early RVD detection using PE patient data and has been shown to support applications in healthcare innovations aimed at improving patient outcomes and resilience in edge devices, clinical decision support systems, and challenging environments. Furthermore, these findings could be used as promising applications by integrating with advances in digital health and human health monitoring systems, such as bionic clothing and smart sensor networks.
右心室功能障碍（RVD）与急性肺栓塞（PE）患者死亡率增加密切相关，因此早期检测至关重要。在急性PE人群中快速、准确且经济地识别RVD风险因素可显著改善诊断和治疗，有可能降低死亡率。本研究使用重复分层留出验证程序评估LogNNet和监督机器学习（ML）模型诊断RVD的性能。提出了一种基于集成的LogNNet模型以供实际应用。LogNNet模型确定性别、冠状动脉疾病、合并症（尤其是高血压）、年龄（74岁以上）、血栓节段以及单侧/双侧血栓是RVD诊断的最显著预测因素。此外，这些特征的组合显示出高预测能力。LogNNet仅使用少数选定特征就取得了稳健的结果，使其适用于资源有限的环境。LogNNet提供了一种实用且易于使用的工具，可利用PE患者数据进行RVD早期检测，并且已被证明可支持在旨在改善患者预后和边缘设备、临床决策支持系统及具有挑战性环境中的恢复力的医疗创新中的应用。此外，通过与数字健康和人类健康监测系统（如仿生服装和智能传感器网络）的进展相结合，这些发现可作为有前景的应用。
