您选择的条件: 2023-03-22
  • 阻塞性睡眠呼吸暂停低通气综合征对高血压患者血压变异性和心率变异性影响研究

    分类: 医学、药学 >> 临床医学 提交时间: 2023-03-22 合作期刊: 《中国全科医学》

    摘要: 背景 阻塞性睡眠呼吸暂停低通气综合征(OSAHS)在高血压患者中患病率高,但诊断率低,其中心率变异性(HRV)和血压变异性(BPV)都是心血管事件相关预测因子,但目前关于OSAHS与高血压患者BPV和HRV内在联系的相关研究较少。目的 本研究旨在探讨OSAHS对高血压患者HRV、BPV的影响,并开发和内、外部验证一种通过HRV和BPV相关指标预测高血压患者OSAHS患病风险的列线图。 方法 选取2018年1月至2020年12月在中南大学湘雅二医院收治的228例高血压患者作为研究对象,根据OSAHS诊断标准分为单纯高血压组(n=114)和高血压合并OSAHS组(n=114);另外收集2021年1月2月住院的34例高血压伴或不伴OSAHS患者作为独立的外部验证组。收集研究对象的一般资料〔年龄、性别、体质指数(BMI)等〕、平均血压水平〔夜间收缩压(nSBP)等〕、BPV相关指标〔夜间收缩压标准差(nSSD)、夜间舒张压标准差(nDSD)、24h舒压标准差(24hDSD)等〕、血压昼夜节律、HRV相关指标〔RR间期平均值标准差(SDANN)、低频带(LF)等〕、多导睡眠监测(PSG)参数〔氧减指数(ODI)、睡眠呼吸暂停低通气指数(AHI)、最低血氧饱和度(MinSpO2)等〕。比较单纯高血压组和高血压合并OSAHS组一般资料、平均血压水平、BPV相关指标、血压昼夜节律、HRV相关指标的差异;并依据AHI标准和MinSpO2标准将高血压合并OSAHS组患者分别分为轻度组、中度组和高度组,比较不同严重程度组高血压合并OSAHS组患者一般资料、平均血压水平、BPV和HRV相关指标的差异;采用Pearson相关性分析和多元线性回归分析探究HRV和BPV相关影响因素;并绘制限制性立方样条图检验高血压患者平血压水平、BPV和HRV相关指标与OSAHS患病风险的相关性;通过多因素Logistic回归模型分析高血压患者患OSAHS的影响因素,构建列线图预测模型,采用Bootstrap方法在检验内、外部组验证组在列线图模型中的性能;采用受试者工作特征(ROC)曲线评估内、外部验证组列线图在高血压患者OSAHS患病风险的预测价值,计算ROC曲线面积(AUC)等指标。 结果 228例高血压患者中男性168例(73.7%),女性60例(26.3%),平均年龄为(62.012.9)岁,BMI为(24.23.5)kg/m2。高血压合并OSAHS组患者的BMI高于单纯高血压组(P<0.05),高血压合并OSAHS组夜间BPV相关指标高于单纯高血压组,HRV相关指标低于单纯高血压组(P<0.05)。AHI标准和MinSpO2标准下不同严重程度高血压合并OSAHS组患者相关资料比较结果显示,重度组与轻、中度组的BMI、BPV、HRV相关指标比较,差异有统计学意义(P<0.05)。Pearson相关性分析和多元线性回归分析结果显示:高血压合并OSAHS组患者的nSSD、nDSD水平、HRV相关指标与BMI、AHI、ODI水平存在相关关系(P<0.01),BMI、ODI、MinSpO2是高血压合并OSAHS组患者nSSD、nDSD水平和HRV相关指标的独立影响因素(P<0.05);限制性立方样条模型结果显示BPV、HRV相关指标与发生OSAHS存在非线性相关(P<0.05),纳入多元Logistic回归分析模型后发现nSBP、nSSD、24hDSD、SDANN、LF、年龄、BMI是高血压患者发生OSAHS的影响因素(P<0.05),以年龄、BMI、nSBP、nSSD、24hDSD、SDANN、LF为预测因子构建列线图预测模型,Bootstrap方法验证结果显示,内、外部验证组的绝对误差分别为0.013、0.021,表明列线图模型的校准度良好。内、外部验证组列线图预测高血压患者OSAHS患病风险的AUC分别为0.861、0.744〔95%CI(0.818,0.919),P<0.001;95%CI(0.691,0.839),P<0.001〕。 结论 OSAHS可增加高血压患者夜间BPV,降低HRV,HRV和BPV均与OSAHS病情严重程度密切相关,夜间缺氧或许更能引起血压和心率变化。本研究构建的列线图也许可用于高血压患者发生OSAHS风险的个体化预测,HRV和BPV参数或许是筛选OSAHS的有力工具。

  • Curvature-Balanced Feature Manifold Learning for Long-Tailed Classification

    分类: 计算机科学 >> 计算机科学技术其他学科 提交时间: 2023-03-22

    摘要: To address the challenges of long-tailed classification, researchers have proposed several approaches to reduce model bias, most of which assume that classes with few samples are weak classes. However, recent studies have shown that tail classes are not always hard to learn, and model bias has been observed on sample-balanced datasets, suggesting the existence of other factors that affect model bias. In this work, we systematically propose a series of geometric measurements for perceptual manifolds in deep neural networks, and then explore the effect of the geometric characteristics of perceptual manifolds on classification difficulty and how learning shapes the geometric characteristics of perceptual manifolds. An unanticipated finding is that the correlation between the class accuracy and the separation degree of perceptual manifolds gradually decreases during training, while the negative correlation with the curvature gradually increases, implying that curvature imbalance leads to model bias. Therefore, we propose curvature regularization to facilitate the model to learn curvature-balanced and flatter perceptual manifolds. Evaluations on multiple long-tailed and non-longtailed datasets show the excellent performance and exciting generality of our approach, especially in achieving significant performance improvements based on current state-ofthe-art techniques. Our work opens up a geometric analysis perspective on model bias and reminds researchers to pay attention to model bias on non-long-tailed and even samplebalanced datasets. The code and model will be made public.