Blood Dialysis Room, Tonglu Hospital of Traditional Chinese Medicine, Tonglu, Hangzhou 311500, Zhejiang, China.
Contrast Media Mol Imaging. 2022 Aug 27;2022:9356108. doi: 10.1155/2022/9356108. eCollection 2022.
This study is aimed at exploring the diagnostic value of digital subtraction angiography (DSA) based on faster region-based convolutional networks (Faster-RCNN) deep learning for maintenance hemodialysis (MHD) diseases and to provide a theoretical basis for clinical nursing. A total of 50 MHD patients who were clinically diagnosed in the Blood Purification Center were randomly divided into the control group and the experimental group (25 cases for each group). The control group was given routine nursing intervention, and the experimental group was given overall nursing intervention under the supervision of DSA. A faster RCNN multitarget detection network was constructed to analyze the average accuracy of various vascular structures in the test set. The self-rating anxiety scale (SAS) and self-rating depression scale (SDS) were used to evaluate the degree of anxiety and depression. The urine volume before and after the operation, local hematoma after a puncture, the incidence of complications, and nursing satisfaction were recorded. The results showed that the average accuracy of the vein, internal carotid artery, circle of Willis, venous sinus, and venous vessels was 0.876, 0.916, 0.994, 0.925, and 0.732, respectively. The success rate of surgery in the experiment group was higher than that in the control group, and the difference had statistical significance ( < 0.05). The SAS score and SDS score in the experimental group were significantly lower than those in the control group ( < 0.05). The total incidence rate of complications in the experimental group (16.00%) was significantly lower than that in the control group (44.00%) ( < 0.05). The satisfaction rate of the experimental group was significantly higher than that of the control group ( < 0.05). The Faster-RCNN model had the best effect in differentiating the circle of Willis and a poor effect in differentiating venous vessels. DSA based on Faster-RCNN can significantly improve the success rate of puncture in MHD patients. The implementation of holistic nursing intervention under its supervision can significantly reduce postoperative complications and improve patient satisfaction with nursing compared with routine nursing.
本研究旨在探讨基于快速区域卷积网络（Faster-RCNN）深度学习的数字减影血管造影（DSA）对维持性血液透析（MHD）疾病的诊断价值，并为临床护理提供理论依据。将 50 例临床诊断为 MHD 的患者在血液净化中心随机分为对照组和实验组（每组 25 例）。对照组给予常规护理干预，实验组在 DSA 监护下给予整体护理干预。构建 Faster-RCNN 多目标检测网络，分析测试集中各种血管结构的平均准确率。采用焦虑自评量表（SAS）和抑郁自评量表（SDS）评估焦虑和抑郁程度。记录手术前后尿量、穿刺局部血肿、并发症发生率和护理满意度。结果显示，静脉、颈内动脉、Willis 环、静脉窦和静脉血管的平均准确率分别为 0.876、0.916、0.994、0.925 和 0.732。实验组手术成功率高于对照组，差异有统计学意义（<0.05）。实验组 SAS 评分和 SDS 评分明显低于对照组（<0.05）。实验组总并发症发生率（16.00%）明显低于对照组（44.00%）（<0.05）。实验组的满意度明显高于对照组（<0.05）。Faster-RCNN 模型在区分 Willis 环方面效果最佳，在区分静脉血管方面效果较差。基于 Faster-RCNN 的 DSA 可显著提高 MHD 患者穿刺成功率。在其监护下实施整体护理干预可显著降低术后并发症，提高患者对护理的满意度。
