To address perception distortion, difficulties in heterogeneous data fusion, and missed detection of small faults in defect-elimination robots for ultra-high-voltage transmission lines under strong electromagnetic interference, a physics-aware spatiotemporal graph fusion network PA-STGFN is proposed. The Middleton Class-A model is employed to estimate the parameters of non-Gaussian impulsive noise, combined with robust statistics for adaptive suppression. EfficientNet enhanced with coordinate attention and MS-TCN are used to extract visual and vibration–acoustic features, respectively. Modality uncertainty is quantified based on Dirichlet evidence, enabling dynamic reconstruction of graph connections and multi-source confidence-weighted fusion. The model is jointly trained using focal loss, reconstruction loss, and uncertainty regularization. Validation on transmission-line data shows that the feature-matching accuracy and anomaly-detection rate reach 95.8% and 96.5%, respectively, with a performance degradation of less than 3% at 10 dB and a single-frame inference latency of 78 ms. The proposed method can improve the perception reliability of defect-elimination robots in strong electromagnetic environments.
Objective To systematically evaluate the effects of different nursing intervention models on the quality of life of elderly patients with cardiovascular diseases, integrate current clinical research evidence, and provide evidence-based references for optimizing standardized nursing protocols and improving long-term prognoses for elderly patients with cardiovascular diseases. Methods Databases including CNKI, Wanfang, VIP, PubMed, Embase and Cochrane Library were electronically searched from database inception to June 2026. Randomized controlled trials and prospective cohort studies involving elderly patients with cardiovascular diseases receiving specialized nursing interventions were included. Basic experiments, reviews, case reports, studies with missing data and non-Chinese-English publications were excluded. Two researchers independently completed literature screening, data extraction and quality assessment. A descriptive systematic review was adopted for qualitative integration and analysis of included studies, so as to summarize the application effects and existing research limitations of different nursing interventions. Results A total of 28 publications were finally enrolled, including 21 Chinese core articles and 7 English articles, with an overall sample size of 4 326 participants. The intervention models of included studies were categorized into five types: transitional care, comprehensive individualized nursing, mindfulness-based psychological nursing, cardiac exercise rehabilitation nursing, and special-topic health education. Multiple studies indicated that multi-mode integrated nursing interventions could significantly improve the multi-dimensional quality of life (physiological, psychological and social functioning) in elderly patients with cardiovascular diseases, enhance medication adherence, and reduce the incidence of adverse cardiovascular events and unplanned readmissions. In contrast, single short-term health education and basic nursing interventions exerted limited effects on improving long-term quality of life. Existing studies had problems such as lack of unified standards for intervention protocols, large variations in follow-up duration, high research heterogeneity, and insufficient evidence for subgroups of frail advanced-age patients. Conclusion Diversified integrated nursing interventions are effective in improving the quality of life of elderly patients with cardiovascular diseases. Among them, transitional comprehensive nursing and exercise rehabilitation combined with psychological nursing yield the most prominent benefits. Standardization of relevant clinical research needs to be further improved. High-quality studies with large-sample, multi-center design and long-term follow-up should be carried out in the future to construct standardized and individualized nursing systems suitable for elderly Chinese patients with cardiovascular diseases.
Objective To investigate the association between thoracic irradiation (TIR) and the severity of coronary artery calcification (CAC) in patients with lung cancer, and to provide evidence-based basis for cardiovascular risk stratification and early intervention in lung cancer patients undergoing radiotherapy. Methods A retrospective cohort study design was adopted. A total of 126 patients with primary lung cancer who received thoracic radiotherapy in the oncology department of a tertiary hospital from January 2018 to December 2023 were consecutively enrolled as the radiotherapy group, and 126 lung cancer patients diagnosed during the same period but without radiotherapy were matched at a 1:1 ratio as the non-radiotherapy group. Propensity score matching was performed for baseline characteristics including age, sex, smoking history, hypertension, diabetes, dyslipidemia and tumor stage. The CAC score was semiquantitatively calculated by the Agatston method based on follow-up chest CT images obtained at least 12 months after radiotherapy, and the severity was classified into three grades according to the number of involved vessels: no calcification, single-vessel calcification and multi-vessel calcification. Dosimetric parameters of the radiotherapy group were collected, including mean heart dose (MHD), heart V30 and left anterior descending artery (LAD) dose. Inter-group differences were compared using t-test, χ2 test and rank-sum test. Multivariate ordinal Logistic regression was used to analyze the independent influencing factors of CAC severity. Results The median CAC score in the radiotherapy group was significantly higher than that in the non-radiotherapy group [426.3 (IQR: 118.5–982.7) AU vs. 187.6 (IQR: 42.3–563.2) AU, Z=-4.215, P<0.001]. The proportion of multi-vessel calcification in the radiotherapy group was significantly higher than that in the non-radiotherapy group (46.0% vs. 27.0%, χ2=10.286, P=0.006). Multivariate ordinal Logistic regression analysis showed that after adjusting for age, hypertension, diabetes, dyslipidemia and smoking history, thoracic radiotherapy was an independent risk factor for increased CAC severity (OR=2.87, 95%CI: 1.64–5.02, P<0.001). In the subgroup analysis of the radiotherapy group, MHD≥10 Gy (OR=2.31, 95%CI: 1.12–4.76, P=0.023) and LAD maximum dose≥35 Gy (OR=2.05, 95%CI: 1.03–4.08, P=0.041) were independently associated with more severe CAC. Conclusion Thoracic radiotherapy in lung cancer patients is independently associated with increased severity of coronary artery calcification, and patients with higher cardiac radiation dose tend to have greater CAC severity. Routine baseline CAC assessment before radiotherapy, optimization of cardiac dose constraints during treatment, and strengthened cardiovascular event monitoring and early intervention during follow-up are recommended.
This paper investigates the robust stability of a class of linear time-delay systems subject to unknown parameters and delays. Since the exact values of these quantities are often unavailable in practice, an analytical framework based on the distribution of characteristic roots is developed. Through theoretical derivation of the characteristic equation for the perturbed system, an upper bound on its unstable eigenvalues is established. Subsequently, by employing the argument principle, sufficient conditions for robust stability are derived for both first-order and high-order time-delay systems. These conditions are easy to verify with low computational cost and are convenient for direct engineering application.
Smart-grid operational analytics rely on heterogeneous data, analytical services, and distributed computing resources, but existing discovery and unified-access mechanisms do not ensure executable orchestration under operational constraints. This paper proposes BS-DCSO, a business-semantic-aware method for jointly selecting data services, computing services, and resource nodes in a smart-grid data-fabric environment. BS-DCSO integrates type- and permission-aware filtering, QoS-based top-k pruning, business-context scoring, complete-plan evaluation, feasibility-first lexicographic selection, and failure-triggered re-orchestration. Experiments on a synthetic benchmark covering five representative smart-grid scenarios and ten random seeds show a provisioning success rate of 0.4897 ± 0.1139, a feasible-instance success rate of 0.8320 ± 0.0562, and a constraint violation rate of 0.0809 ± 0.0213. Compared with ResourceFirstScheduling, BS-DCSO improves paired provisioning success by 0.1423 ± 0.0552, while outperforming five controlled baselines across the three main metrics. Re-orchestration recovers feasible replacements in 0.8282 ± 0.0398 of eligible single-service failures. Under the sequential fixed-catalogue timing protocol, the method requires 151.39–175.58 ms per request for 100–1000 requests, indicating a practical feasibility–computation trade-off under the tested synthetic setting rather than high-concurrency scalability.
