A comparison of learning algorithms for Bayesian networks: a case study based on data from an emergency medical service

Due to the uncertainty of many of the factors that influence the performance of an emergency medical service, we propose using Bayesian networks to model this kind of system. We use different algorithms for learning Bayesian networks in order to build several models, from the hospital manager’s point of view, and apply them to the specific case of the emergency service of a Spanish hospital. This first study of a real problem includes preliminary data processing, the experiments carried out, the comparison of the algorithms from different perspectives, and some potential uses of Bayesian networks for management problems in the health service.     

Noise reduction of renograms: a new algorithm applied to simulated renograms for evaluation of the renal retention function.

Physiological information about an organ may be assessed from the retention function as derived by deconvolution analysis. However, noise in data may cause distortion of the retention curve and potentially induce methodological errors. To take full advantage of all parts of the retention function, i.e. even the vascular part, we have developed a non-linear noise reduction algorithm. The algorithm is an adaptive polynomial fit (APF) in a sliding segment over the renogram to be deconvoluted. Each segment is modelled by the lowest polynomial resulting in a root of mean square error lower than a pre-set value. APF was tested in comparison with conventional repetitive 1:2:1 smoothing and rectangular window smoothing, using a set of simulated retention functions and corresponding renograms with superimposed artificial noise. The outcome was evaluated by comparing the generated retention functions with the simulated ones. The conventional smoothing algorithms, as well as APF, induced some distortion of the retention function, but the deviation from the true retention function is essentially lower in the case of APF. In addition, APF seems to be more robust in cases of essentially reduced renal function and thereby relative high noise levels. APF reduces noise in data, leading to retention functions with reliable information in terms of a tracer's first passage, uptake and outflow.     

Multi-label classification of chronically ill patients with bag of words and supervised dimensionality reduction algorithms

ObjectiveThis research is motivated by the issue of classifying illnesses of chronically ill patients for decision support in clinical settings. Our main objective is to propose multi-label classification of multivariate time series contained in medical records of chronically ill patients, by means of quantization methods, such as bag of words (BoW), and multi-label classification algorithms. Our second objective is to compare supervised dimensionality reduction techniques to state-of-the-art multi-label classification algorithms. The hypothesis is that kernel methods and locality preserving projections make such algorithms good candidates to study multi-label medical time series.MethodsWe combine BoW and supervised dimensionality reduction algorithms to perform multi-label classification on health records of chronically ill patients. The considered algorithms are compared with state-of-the-art multi-label classifiers in two real world datasets. Portavita dataset contains 525 diabetes type 2 (DT2) patients, with co-morbidities of DT2 such as hypertension, dyslipidemia, and microvascular or macrovascular issues. MIMIC II dataset contains 2635 patients affected by thyroid disease, diabetes mellitus, lipoid metabolism disease, fluid electrolyte disease, hypertensive disease, thrombosis, hypotension, chronic obstructive pulmonary disease (COPD), liver disease and kidney disease. The algorithms are evaluated using multi-label evaluation metrics such as hamming loss, one error, coverage, ranking loss, and average precision.ResultsNon-linear dimensionality reduction approaches behave well on medical time series quantized using the BoW algorithm, with results comparable to state-of-the-art multi-label classification algorithms. Chaining the projected features has a positive impact on the performance of the algorithm with respect to pure binary relevance approaches.ConclusionsThe evaluation highlights the feasibility of representing medical health records using the BoW for multi-label classification tasks. The study also highlights that dimensionality reduction algorithms based on kernel methods, locality preserving projections or both are good candidates to deal with multi-label classification tasks in medical time series with many missing values and high label density.     

微生物分类单元聚类算法比较研究

目的 随着高通量测序技术的发展,产生了大量的微生物16S rRNA基因序列数据.对该数据进行精确的微生物操作分类单元（operational taxonomic unit,OTU）划分,有助于了解环境中微生物的种群组成及分布.方法 本文在真实数据集与模拟数据集上,对现有的7种流行OTU单元聚类算法进行了对比研究,并分析了这些算法的优缺点及使用范围.结果 序列长度、测序深度对聚类结果均有影响.结论 相同的序列相似性阈值下,不同的聚类算法聚类结果差异较大,其中CROP算法的鲁棒性和抗噪性较好.     

Predicting the outcome of pregnancies of unknown location: Bayesian networks with expert prior information compared to logistic regression

BACKGROUND: As women present at earlier gestations to early pregnancy units (EPUs), the number of women diagnosed with a pregnancy of unknown location (PUL) increases. Some of these women will have an ectopic pregnancy (EP), and it is this group in the PUL population that poses the greatest concern. The aim of this study was to develop Bayesian networks to predict EPs in the PUL population. METHODS: Data were gathered in a single EPU from all women with a PUL. This data set was divided into a model-building (599 women with 44 EPs) and a validation (257 women with 22 EPs) data set and consisted of the following variables: vaginal bleeding, fluid in the pouch of Douglas, midline echo, lower abdominal pain, age, endometrial thickness, gestation days, the ratio of HCG at 48 and 0 h, progesterone levels (0 and 48 h) and the clinical outcome of the PUL. We developed Bayesian networks with expert information using this data set to predict EPs. RESULTS: The best Bayesian network used the gestational age, HCG ratio and the progesterone level at 48 h and had an area under the receiver operator characteristic curve (AUC) of 0.88 for predicting EPs when tested prospectively. CONCLUSIONS: Discrete-valued Bayesian networks are more complex to build than, for example, logistic regression. Nevertheless, we have demonstrated that such models can be used to predict EPs in a PUL population. Prospective interventional multicentre studies are needed to validate the use of such models in clinical practice.     

Effects of thrombin produced in blood prior to mixing with anticoagulant on the platelet function and blood coagulation tests

The amount of thrombin produced in the withdrawn blood of healthy donors prior to mixing with anticoagulants was estimated with thrombin-antithrombin III complex (TAT) methods. No samples showed production of thrombin in the blood mixed with anticoagulants within 2 min. After 3 min or more incubation, four donors out of 14 showed elevation of TAT, and a technician could suspect trouble only in a case during taking blood. Although two samples with low thrombin production (5.8, 12 ng/ml, normal: less than 5 ng/ml) showed no changes of APTT or prothrombin time, samples with high thrombin production (74, 98 ng/ml) showed 1.4-2 sec shortening of APTT and 0.8-1 sec shortening of prothrombin time. Platelet counts and platelet volume showed no remarkable changes even when samples with large amount of thrombin were suspended in tubes containing EDTA-2 K. Plasma beta-thromboglobulin concentration was elevated in the blood containing high amount of thrombin and close correlation was observed amount of thrombin in the blood and the amount of plasma beta-thromboglobulin (r = 0.90, P value less than 0.001). Platelet aggregation induced by collagen was not influenced by amount of thrombin produced. ADP-induced aggregation was suppressed in 3 out of 4 cases with thrombin production although no close correlation was observed between maximal aggregation rate and the concentration of thrombin. As improper handling of blood samples including withdrawal of blood leads to misunderstanding of test results. So, blood samples only for these sensitive tests should be taken and mix with anticoagulants within 2 min.     

TNF-α在CCD模型大鼠背根神经节中的表达及机制

目的:探讨TNF-α和NF-κB在背根神经节慢性压迫(CCD)模型大鼠背根神经节(DRG)中的表达变化及其对疼痛学行为的影响。方法:建立CCD大鼠模型,采用von Frey纤维丝监测机械痛阈的改变;通过Western blotting检测TNF-α和NF-κB在DRG中的表达变化趋势,分析其与疼痛行为之间的相关性;并采用免疫荧光双染技术研究TNF-α在DRG中的表达位置。结果:CCD组的50%机械缩足阈值在术后1 d即开始明显下降(P0.01),7~14 d达到高峰,其后逐渐上升,直至术后35 d仍明显低于术前及sham组(P0.01)。而DRG上的TNF-α及NF-κB于造模后各时点均显著增多(P0.01),且TNF-α的表达趋势与50%机械缩足阈值显著相关(P0.05)。结论:DRG慢性压迫可促进其上的TNF-α和NF-κB的合成和分泌,进而诱发机械痛觉过敏。因此,TNF-α/NF-κB信号通路可能是CCD模型疼痛形成的重要通路之一。     

基于激光扫描与CT建立带牙根的三维数字化牙颌模型

目的　基于激光三维扫描牙颌模型,结合临床头部CT资料,建立包含牙根的数字化牙颌模型,记录牙冠和牙根解剖形态结构以及空间位置变化。 方法　1例牙列正常成年女性,头部CT数据利用Mimics11.0软件建立口腔内单独牙齿和整个上下牙列三维数字化模型;三维激光扫描仪分别扫描上下颌石膏模型及其咬合关系,在GeoMagic10.0软件中,结合CT重建的单独牙齿的三维数字化模型,重构带有牙根的牙列三维模型,同时实现上下颌三维模型符合口腔实际咬合关系的位置恢复。 结果　重建了带有牙根的三维数字化上下颌牙列模型,不仅清晰显示牙冠的解剖特征、牙列的排列、牙根的形态和位置关系,并且准确恢复了上下颌牙列的咬合关系。 结论　在正畸治疗过程中,利用激光三维扫描及CT数据建立带有牙根的三维数字化牙颌模型,不仅可以显示牙冠牙根的解剖形态结构,还能够监测牙冠和牙根的移动情况,便于指导治疗进程,评价治疗效果。     

基于故障树与贝叶斯网络的呼吸机故障智能诊断

为了快速准确地找出呼吸机故障原因,迅速排除故障,恢复设备的正常运行,本文采用基于故障树和贝叶斯网络的 方法对呼吸机常见故障进行分析。首先通过对呼吸机结构原理的综合分析,结合文献案例搭建呼吸机故障树,进行定性 分析;利用贝叶斯网络对呼吸机故障进行定量分析;最后用实际维修案例进行验证。结果表明,该方法得到的推理结果与 实际结果相符性达到84.54%,为建立呼吸机故障静态数据库并进行故障智能诊断提供了理论依据,具有一定的推广 价值。     

Electroconvulsive shock effects on food and water intake as a function of prior deprivational state

Two experiments, each using approximately 30 male and 30 female hooded rats, examined the effects of electroconvulsive shock (ECS) on water intake as a function of the water deprivation state of the animal and food intake and body weight as a function of food deprivation state. In Experiment 1, half the animals were subjected to water deprivation prior to ECS or sham ECS, and the other half were watered ad lib; in Experiment 2, food deprivation was the pre-ECS manipulation. As predicted, ECS produced a decrease in water intake in the ad lib watered animals but not in the deprived ones. A similar treatment by deprivation interaction was not found for food intake or body weight. In the males, food intake was decreased for one day by the treatment but recovered to a normal level two days following ECS; however, such an effect of ECS on food intake did not occur in the females. Since food and water intake was not differentially affected by ECS and sham ECS in deprived animals, previous cautions regarding ECS effects on motivational states may not be germane to many studies using appetitive conditioning paradigms.     

Rate-responsive pacing based on the atrio-ventricular conduction time: comparison of different algorithms

Rate-responsive pacing based on the atrio-ventricular conduction time is a promising therapy for restoring physiological heart rate control in chronotropic incompetent patients. This paper compares four different algorithms. Three of them had been formulated as patents, but no real test or application has been reported up to now. The fourth was recently published by the authors of this paper. There the steps involved in the development process were stationary and dynamic system identification, control system design and a pilot study with patients. The data obtained were used to formulate a simulation model of the cardiac system by means of which the other algorithms were tested. Test criteria were stability, the attenuation of disturbances and the response time to changes of the exercise rate. None of the three patents worked when being strictly implemented as described. The problems encountered were instability, unusable parameterisations and some questionable adaptation mechanisms. In a redesign we tried to improve the patents, but only in one case would the results obtained justify real use. In the other cases the variability of the pacing frequency was intolerably high.     

基于PCA和LDA数据降维的脑磁图脑机接口研究

脑磁图(MEG)信号作为一种新的脑-机接口(BCI)输入信号,含有手运动方向的模式信息。通常对MEG信号采用信号处理的特征提取和线性分类,识别率一直难于提高。本文提出用主成分分析(PCA)方法对其进行特征提取,并用线性判别分析(LDA)进行了优化,最后用最近邻非线性分类器进行分类,在分类结果的基础上分析了混淆矩阵。实验结果表明PCA+LDA方法能有效的分析多通道的MEG信号,平均识别率达到了53.0%,优于BCI竞赛Ⅳ的识别率46.9%。     

Predictors of urinary tract infection based on artificial neural networks and genetic algorithms

BACKGROUND: Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. METHODS: We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of > or =10(5) colony-forming units (CFU) per milliliter, and counts of > or =10(2) CFU per milliliter. RESULTS: Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of > or =10(5) CFU per milliliter) to 0.792 (for uropathogen counts of > or =10(2) CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. CONCLUSIONS: ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.     

Melanomas non-invasive diagnosis application based on the ABCD rule and pattern recognition image processing algorithms

In this paper an automated dermatological tool for the parameterization of melanomas is presented. The system is based on the standard ABCD Rule and dermatological Pattern Recognition protocols. On the one hand, a complete stack of algorithms for the asymmetry, border, color, and diameter parameterization were developed. On the other hand, three automatic algorithms for digital image processing have been developed in order to detect the appropriate patterns. These allow one to calculate certain quantitative features based on the aspect and inner patterns of the melanoma using simple-operation algorithms, in order to minimize response time. The database used consists of 160 500×500-pixel RGB images (20 images per pattern) cataloged by dermatologists, and the results have turned out to be successful according to assessment by medical experts. While the ABCD algorithms are mathematically reliable, the proposed algorithms for pattern recognition produced a remarkable rate of globular, reticular, and blue veil Pattern recognition, with an average above 85% of accuracy. It thus proves to be a reliable system when performing a diagnosis.     

Noise of surface bio-potential electrodes based on NASICON ceramic and Ag−AgCl

The electrochemical noise from dry NASICON-based surface electrodes and pregelled Ag−AgCl electrodes is evaluated in saline solutions and on the skin. The electrochemical noise from the electrode/electrolyte interface is found to be negligible (less than 1 μV peak to peak). On the skin, the noise level is highly dependent on the patient. At high frequencies, the skin/electrode interface noise is equal to ‘thermal noise’ and can be related to the real part of the skin/electrode impedance. At low frequencies (f<100 Hz), excess noise is observed that varies as f⁻². It is tentatively ascribed to a non-stationary process or noise of electrochemical origin due to the ionic nature of the skin. The contribution of residual EMG signal of low amplitude (5 μV peak to peak) is suggested for electrodes with large surface area. Reprint requests and correspondence should be addressed to Frist     

Effects of prior testing with the WAIS-R NI on subsequent retest with the WAIS-R.

Seven subtests from the Wechsler Adult Intelligence Scale, Revised as a Neuropsychological Instrument (WAIS-R NI) were administered to 20 nonreferred university students. The same participants were administered the corresponding subtests from the WAIS-R 3 to 4 weeks later. Data concerning amount and consistency of change in 'scaled scores' were compared to those reported by Wechsler (1981) for test-retest with the WAIS-R. Performance on standard items was also compared to performance on multiple choice items from the WAIS-R NI. Gains observed in subtest 'scaled scores' at retest were comparable to those reported by Wechsler (1981). A substantial minority of participants obtained lower scores on some multiple choice items than they did on corresponding standard items. Implications and limitations of the current data, and the pressing need for comprehensive normative data for the WAIS-R NI are discussed.