不同海拔高度健康藏汉族青年脉象图研究

本文通过对不同海拔高度藏汉族健康青青年的４００幅脉图分析后认为，健康藏汉族青年的脉象以滑、平、弦、缓为多。结古和西宁的藏汉族青年脉力图的主要生理参数中不同海拔高度上有一定差异，表现为结古青年弦脉较多，而西宁教地区滑脉较多，究其形成原因，主要与高海拔缺氧、寒冷环境有密切关系；在民族间也存在一定差异，表现为汉族弦脉多，藏族滑脉多，它们的形成与移居高原后心血管的反庆和代偿机制有关，在性别间有明星的差异。     

PCA在过程故障检测与诊断中的应用

讨论了基于主无分析（PCA）的过程故障检测与诊断的原理,运用T^2统计、Q统计方法,结合贡献图对一典型过程进行了仿真分析,结果表明PCA方法可对简单传感器故障进行检测与诊断,并指出了该方法中的不足,提出了将PCA方法同基于过程动态模型的故障诊断方法相结合的研究思路。     

Quality of diabetes care in a university health center in Lebanon.

OBJECTIVE: To assess the quality of care provided to diabetic patients by family physicians in a university health clinic, using measures of glycemic and cardiovascular risk control as well as documentation of and adherence to World Health Organization (WHO) guidelines for diabetes primary care. DESIGN: Chart review of the previous year's medical notes for all identified diabetics in the practice over 2.5 years. RESULTS: Two-hundred and four diabetic patients were identified, with an estimated prevalence of 4.1%. The majority was type II diabetics, on oral hypoglycemic agents. Glycosylated hemoglobin was documented in 39.7% of patients, fasting plasma glucose in 99%, cholesterol in 93.1%, triglycerides in 91.2% and blood pressure in 85.8%; optimal control of these indicators was noted in 28.4%, 17.8%, 34%, 29.6% and 55.4% respectively. Fifty percent of the diabetics were referred for retinal checks. Physicians documented the presence of nephropathy in 46.8% and neuropathy in 59.6%; however, they documented patient instruction on foot care, diet, exercise and diabetes self-care poorly. CONCLUSION: There is a need for interventions to improve management and documentation in diabetes care in order to achieve early detection and prevention of complications. Developing a protocol for the clinic based on standard guidelines, and the use of flow sheets may be helpful in improving these intermediate indicators of quality of care.     

高海拔地区健康藏汉族青年脉象图的研究

通过对高海拔地区（海拔３７０３ｍ）健康藏汉族青年的２００幅脉图分析后认为，健康藏汉族青年的脉象以滑、平、弦、缓为多。该地区藏汉族青年脉图的主要生理参数在不同民族间没有明显的差异；而在不同性别上存在着差异，表现为男性滑脉较多，女性弦脉较多，究其形成差异的原因，与男女血液动力学的差异有一定的关系。     

质量控制图在农村互助医疗保险项目监测中的应用

以中国农村互助医疗保险项目试点乡铁厂镇的门诊服务利用人次和门诊补偿费用的监测为例，介绍了质量控制图的制作方法以及质量控制图在项目运行监测中的应用，旨在探索我国的医疗保险制度的监测管理方法。     

直线加速器日质控数据分析及过程评估

目的 应用统计过程控制方法对加速器日质控(QC)数据进行分析,并对使用晨检仪的QC过程进行评估。方法 加速器、晨检仪校准后分别收集由技师、物理师摆位的100组、30组QC数据,设备第2次校准后再次收集技师摆位的QC数据100组,分析两次校准后技师摆位数据(各100组)的归一化信噪比的变化规律。使用由技师和物理师摆位的QC数据(各30组)绘制控制图,比较中心线位置和上下控制线范围的不同。计算由技师、物理师摆位的3个组日QC的过程能力指数。结果 两次校准的技师摆位数据归一化信噪比均为前6周变化较大,6~8周后趋于稳定,8周后逐渐变小。物理师摆位的QC数据在输出量一致性方面,上下控制线范围更窄;在平坦度、对称性方面,中心线更接近目标值0。对输出量一致性、平坦度方面,3个组日QC的过程能力指数均满足≥1要求;对称性Transverse方向均不满足。结论 应采用30~40个数据点绘制加速器日QC过程的控制图。QC过程应由相对固定且较少的QC人员完成,检测项目也应设置更适合的容差。     

儿科应用氨基糖苷类抗菌药物研究进展

目的：了解我院儿童药品不良反应（ADR）的规律及特点。方法：采用回顾性分析方法,统计我院2017-2018年收集的348例儿童ADR报告,分别从患儿性别与年龄、药品种类、发生时间、ADR累及系统/器官分布、临床转归等方面进行Pareto最优分析。结果：348例儿童ADR中,1~12岁患儿累计构成比为77.59%（A类因素）,抗感染药、心血管系统药、呼吸系统及抗肿瘤药引起的ADR累计构成比为77.87%（A类因素）;涉及的抗感染药主要为头孢菌素和大环内酯类（65.22%）,药品剂型主要为注射剂及粉针剂（77.59%）,给药途径主要为静脉滴注（67.24%）,发生ADR时间主要为用药后2 d内（79.60%）,累及系统/器官主要为皮肤及其附件和胃肠系统（77.78%）;临床转归主要为好转（62.07%）。结论：我院儿童ADR涉及的药品种类主要为抗感染药、心血管系统药、呼吸系统及抗肿瘤药,主要表现为皮肤及其附件损害和胃肠系统紊乱。在ADR监测时应重点关注1~12岁儿童,同时应掌握合理的用药方式、强化监护环节,尽可能减少ADR的发生,以确保儿童临床用药安全、有效。     

Divisive normalization is an efficient code for multivariate Pareto-distributed environments

Divisive normalization is a canonical computation in the brain, observed across neural systems, that is often considered to be an implementation of the efficient coding principle. We provide a theoretical result that makes the conditions under which divisive normalization is an efficient code analytically precise: We show that, in a low-noise regime, encoding an n-dimensional stimulus via divisive normalization is efficient if and only if its prevalence in the environment is described by a multivariate Pareto distribution. We generalize this multivariate analog of histogram equalization to allow for arbitrary metabolic costs of the representation, and show how different assumptions on costs are associated with different shapes of the distributions that divisive normalization efficiently encodes. Our result suggests that divisive normalization may have evolved to efficiently represent stimuli with Pareto distributions. We demonstrate that this efficiently encoded distribution is consistent with stylized features of naturalistic stimulus distributions such as their characteristic conditional variance dependence, and we provide empirical evidence suggesting that it may capture the statistics of filter responses to naturalistic images. Our theoretical finding also yields empirically testable predictions across sensory domains on how the divisive normalization parameters should be tuned to features of the input distribution.

The brain has to make efficient use of its limited resources to represent and respond to the wide range of stimuli in its environment. An important mechanism by which this can be achieved is divisive normalization (1, 2), which is thought to be a canonical computation in the brain (3). This gain control mechanism (according to which the response of a neuron to its preferred stimulus is suppressed by the intensity of nonpreferred stimuli) permits the representation of potentially unbounded stimuli by biophysically feasible bounded firing rates. Originally proposed for individual neurons in the primary visual cortex (1, 4, 5), this computation has since also been observed at the population level in the primary visual cortex (6–8) and throughout the visual hierarchy (9, 10), as well as in several other neural systems including olfactory pathways (11), the middle temporal area (12, 13), the inferotemporal cortex (14), the hippocampus (15), and in multisensory integration (16). In addition, divisive normalization has been shown to play an important role in value representations (17, 18) and for choice behavior, where it has been proposed to account for violations of the independence of irrelevant alternatives (IIA) axiom of rational choice (19–23; but see refs. 24 and 25). The nonlinear computation has also been suggested to play a role in attentional modulation (12, 26, 27), the modulation of response variability (28), the representation of visual uncertainty (29), and probabilistic inference (30, 31). It is further used in neural network models of the visual system (32, 33) as well as in computer vision and image compression (34).This ubiquitous array of functions begs the question of what overarching objective the divisive normalization computation achieves. In this paper, we consider this computation’s information-theoretic properties and provide testable conditions for its efficiency that are both simple and general, making them applicable across many of the aforementioned settings.Since Schwartz and Simoncelli (35) showed empirically that divisive normalization reduces the statistical redundancy present in natural images, a common answer (36) has been that divisive normalization is an implementation of the efficient coding principle (37–41). This principle has been central to our understanding of the visual and other sensory systems (42–44), and it has also provided an account of biases in perception (45) and choice (46–49). Of course, divisive normalization has benefits beyond coding efficiency and redundancy reduction, such as permitting tuning curves that are invariant with respect to “nuisance” dimensions (e.g., maintaining discriminability of orientations regardless of contrast) or ensuring that population responses are easily decodable (e.g., by a linear classifier or winner-take-all competition), among other features (3). Its widespread implementation in the nervous system may thus simultaneously achieve a number of purposes. Here, we focus on the question of whether divisive normalization is indeed an efficient computation, which arises naturally in both the sensory and choice domains.Despite significant progress (50), an answer to this question in terms of testable conditions for efficiency has remained elusive, since formally relating neural computations to stimulus distributions has proved difficult: “The establishment of a precise quantitative relationship between environmental statistics and neural processing is important … [but] it has been surprisingly difficult to make the link quantitatively precise … [and] specification of a probability distribution over the space of input signals … is a difficult problem in its own right” (ref. 51, p. 1194). We close this gap with a theoretical result that makes precise the conditions on the input distribution under which divisive normalization encodes a stimulus efficiently.Existing analytical work in the domain of vision has demonstrated that divisive normalization approximately (but not entirely) removes the statistical dependence in models of filter responses to natural images (52–54) such as the conditional normal (35) or lognormal (55) distributions. Moreover, divisive normalization can be viewed as an approximation of the nonlinear radial Gaussianization transformation that removes the statistical dependence of non-Gaussian elliptically symmetric distributions (56), yet divisive normalization itself can do so only imperfectly owing to its bounded range (57, 58). Lyu (50, 59) has quantified the extent to which divisive normalization reduces the statistical dependence of one such elliptical distribution: the multivariate Student’s t distribution, which is in the class of Gaussian scale mixture models of natural images (60). He showed that even though divisive normalization approximates the transformation that eliminates this model’s statistical dependencies, it can also increase them in low-dimensional settings.This literature typically assumes a model of empirical stimulus statistics and derives the predictions of an (approximately) optimal code. It thus represents the first of two common approaches for testing the efficient coding hypothesis (51). The second approach is to examine the statistics of actual neural responses to naturalistic stimuli, in the spirit of Laughlin (40). Here, we pursue instead a third approach that consists of deriving analytically what stimulus distribution a given computation efficiently represents. This is in contrast to Malo and Laparra (54) or Lyu (50), for example, who use similar techniques but who start by assuming a given model of stimulus statistics. Instead, our approach is similar in spirit to that of Ballé et al. (61), who obtain a density model on images by inverting a generalized divisive normalization transform, except that we obtain the input density in analytical closed form. Without making any a priori assumptions about the stimulus statistics, the input distribution we find to be efficiently encoded captures many important features of naturalistic stimulus statistics, as we demonstrate in an analysis of image statistics. Our approach thus provides an additional perspective on the efficiency properties of divisive normalization.We consider a setting in which an n-dimensional input is to be encoded by the divisively normalized firing rates of n neurons. The input can be either a stimulus or a representation of a stimulus coming from another neural system upstream. In the context of visual stimuli, the multivariate input could arise, for instance, as the responses of a population of linear filters convolved with the stimulus (35). At least two conditions have to be satisfied for the resulting multivariate representation to be efficient in a low-noise regime. First, it ought to adhere to histogram equalization (40) along each input dimension, which ensures that each output is used equally often. Second, maximizing the Shannon entropy of the output distribution requires—in the absence of constraints—that any statistical dependence across dimensions be removed. We use a formulation of the efficient coding principle that, in a low-noise regime, implies both of these desiderata and thus gives rise to a multivariate analog of the classic criterion of histogram equalization. Specifically, we consider a neural code to be efficient if and only if it maximizes the Shannon mutual information (62, 63) between the n-dimensional input and its representation. Since, for sufficiently small noise, this criterion can be approximated arbitrarily well by the requirement that the output distribution is entropy-maximizing, divisive normalization is then efficient whenever it transforms the input distribution into an output distribution that is uniform over the range of values divisive normalization can attain (39, 64).^* This allows us to characterize the class of input distributions that are efficiently encoded in a low-noise regime.We prove that divisive normalization maximizes the entropy of the output distribution if and only if the distribution of inputs in the environment is multivariate Pareto. This suggests that divisive normalization may have evolved as an efficient encoding strategy for heavy-tailed, scale-invariant power-law distributions of the kind that occur in many ecological contexts (66); see also Discussion.The statistical dependence in the multivariate Pareto distribution is also consistent with the conditional variance dependence observed in natural image statistics (35), and it has a representation as a Gamma mixture of independent exponential random variables (providing a link to ref. 60). In an empirical analysis of naturalistic images, we demonstrate that the efficiently encoded Pareto distribution indeed captures the statistics of filter responses to natural images just as well as a common model of natural image statistics does. Divisive normalization may thus be an adaptation, in evolution or development, to various natural contexts with physical quantities whose distributions are characterized by heavy-tailed marginals and an empirically important form of statistical dependence.We generalize our result by allowing for a representation to come at an arbitrary metabolic cost, which affects the shape of an efficient code (67, 68), and we show how this impacts the optimally encoded input distribution. For example, if costs are linear in the total number of spikes (which constrains the average firing rate), then the entropy-maximizing output distribution is exponential, and the associated input distribution changes accordingly. We provide necessary and sufficient conditions on the stimulus distribution for divisive normalization to be efficient under any member of a large family of cost functions.Beyond providing a testable prediction on the shape of stimulus distributions that divisive normalization efficiently encodes, our theoretical result also yields empirically testable predictions across sensory domains on how the parameters of the divisive normalization transformation should be tuned to the parameters of the stimulus distribution (35). Specifically, the power index in the normalization function matches the shape parameter of the stimulus distribution, while the normalization weights are the inverses of the scale parameters of the stimulus distribution. Our theoretical predictions thus open the door to systematic experimental tests of the efficiency properties of empirically observed divisive normalization.