The reliability and accuracy of an electromagnetic motion analysis system when used conjointly with an accelerometer

The effect of an accelerometer driven electronic postural monitor (Spineangel®) placed within the electromagnetic measurement field of the Polhemus Fastrak? is unknown. This study assessed the reliability and accuracy of Fastrak? linear and angular measurements, when the Spineangel® was placed close to the sensor(s) and transmitter. Bland Altman plots and intraclass correlation coefficient (2,1) were used to determine protocol reproducibility and measurement consistency. Excellent reliability was found for linear and angular measurements (0.96, 95% CI: 0.90–0.99; and 1.00, 95% CI: 1.00–1.00, respectively) with the inclusion of Spineangel®; similar results were found, without the inclusion of Spineangel®, for linear and angular measurements, (0.96, 95% CI: 0.89–0.99; and 1.00, 95% CI: 1.00–1.00, respectively). The greatest linear discrepancies between the two test conditions were found to be less than 3.5 mm, while the greatest angular discrepancies were below 3.5°. As the effect on accuracy was minimal, these findings support the conjoint use of the Fastrak? during validation studies of the Spineangel® device.

Statement of Relevance: Although previous studies have used the Fastrak? as the gold standard measurement system, the influence of an accelerometer driven postural monitor on accuracy has not been reported. The strength of the present study has been to determine the effect of accelerometer placement within the electromagnetic field on the reliability and accuracy of the Fastrak?.     

基于门控图卷积神经网络的有机化学反应预测

随着现代医药技术和计算机技术的发展,采用人工智能技术来加速药物的研发进度成为了研究热点,而对有机化学反应产物的高效预测是药物逆合成路线设计中的关键问题。针对样本数据集中化学反应类型分布不均匀的问题,提出了一种主动采样训练下的门控图卷积神经网络（ASGGCN）模型。首先,输入化学反应物的简化分子线性输入规范（SMILES）编码,通过门控图卷积神经网络（GGCN）以及注意力机制预测反应中心所在位置;然后,根据化学约束条件和候选反应中心枚举出可能的化学键组合来生成候选产物,再通过门控图卷积差分网络对候选产物进行筛选;最终,得到反应产物。门控图卷积神经网络拥有三个权重参数矩阵并通过门控对信息加以融合,与传统的图卷积神经网络相比,它能获取更加丰富的原子隐藏特征信息。通过主动采样的方式进行训练,使得该模型能够兼顾较差样本和普通样本的分析能力。实验结果表明,所提模型对化学反应产物的Top-1预测准确率可达87.2%,对比Weisfeiler-Lehman差分网络（WLDN）模型提高了1.6个百分点,可见模型能够更准确地预测有机化学反应产物。     

基于全局时空特性的城市路网交通速度预测模型

城市路网交通速度预测是智能交通系统中的重要组成部分,其可为出行者提供实时的交通信息,对提升道路通行能力具有重要意义。现有基于图卷积网络的预测模型一定程度上加强了对一阶相邻路段间空间关联程度的挖掘,但在非一阶相邻路段关联度大于一阶相邻路段关联度的情况下,如果仍输入原始的邻接矩阵,会遗失一些相对重要的路段空间信息,无法得到较好的预测结果。为准确挖掘城市路网中的时空特性,提出一种基于全局图卷积和门控循环单元的城市路网交通速度预测模型G-GCGRU。考虑全局路网下非一阶相邻路段间的空间影响程度,利用相关性分析方法计算得到路段间的关联度矩阵,并作为新的卷积方式进一步加深对空间特征的挖掘,在此基础上,采用门控循环单元方法提取路网时间特征。使用深圳市罗湖区城市路网车速数据进行实验,结果表明,该模型预测性能优于图卷积网络（GCN）、门控循环单元（GRU）和GCN-GRU混合模型,以均方根误差为评价指标,预测精度分别提高25.3%、4.7%和2.1%。     

Effect of weightlifting upon left ventricular function and markers of cardiomyocyte damage

The purpose of this study was to assess left ventricular (LV) function and biochemical markers of myocyte after prolonged weightlifting activity. Seventeen male subjects (age range 20-34 years) performed a 90-min bout of weightlifting exercise consisting of three sets of 8-10 repetitions at 70% one-repetition maximum. Body mass, heart rate, systolic blood pressure (SBP) and echocardiographically determined indices of LV loading (LV internal diameter during diastole, LV meridonial wall stress), systolic function (stroke volume (SV), ejection fraction (EF), end-systolic pressure volume relationship; SBP/ESV) and diastolic filling (ratio of early to late; E:A) were obtained pre-exercise, immediately after and 24 h post-exercise. A 5-ml venous blood sample was obtained for the assessment of cardiac troponin T (cTnT) via third generation electrochemiluminescence assay. Data were assessed via one-way ANOVA and Pearson's correlation. Although SV declined (80.9 +/- 18.3 vs. 66.9 +/- 17.2, p < 0.05) there was no alteration in LV contractility (EF 62 +/- 6 vs. 59 +/- 7; SBP/ESV 3.51 +/- 1.4 vs. 3.51 +/- 1.4, p > 0.05). The E:A ratio was significantly decreased following exercise (1.78 +/- 0.41 vs. 1.33 +/- 0.37, p < 0.05). This decrease was not fully explained by loading conditions (r2 = 0.05 to 0.24). All values returned to baseline 24 h post-exercise. No cTnT was reported in any of the blood samples. In conclusion, there was no significant evidence of any LV contractile depression and no cTnT was observed post exercise. The small reduction in diastolic filling could not be explained by changes in haemodynamic loading or the post-exercise elevation in heart rate.     

Measurement of specific movement power application: evaluation of weight lifters

A new dynamometer was developed for the measurement of specific movement power (SMP) exerted by mono- or multi-articular movements. To determine the reproducibility of SMP, two identical test protocols were carried out on separate days for six movement types on seven adult males. The movement types were arm pull (AP), leg rise (LR), knee extension (KE), elbow flexion (EF), hip extension (HE) and squat movement (SQ). Variations in peak power obtained in two tests ranged from 0.7% (AP) to 9.6% (SQ). Coefficients of the test-re-test correlation in peak power ranged from 0.805 (SQ) to 0.961 (AP) and standard errors ranged from 4 W (EF) to 14 W (SQ). SMP in upper extremities increased from 166 W (EF) to 307 W (AP) resulting from the increase in velocity. However, in the movements of lower extremities, SMP increased from 506 W (KE) to 1351 W (SQ) as a result of the increase in force. To evaluate the validity of the SMP, a pull movement in weightlifting was tested and related to the athletic performance on weightlifters. Positive linear correlation (r = 0.862, p<0.001) was observed between SMP and the total weight best records. It is concluded that this newly developed dynamometer has enough reproducibility and validity for evaluating the SMP, which is developed by various joint movement patterns related to the sport. The feasibility of applying this measuring protocol to the testing and training programmes for improving the daily living activities and athletic performances should now be assessed.     

Measurement of specific movement power application: evaluation of weight lifters

A new dynamometer was developed for the measurement of specific movement power (SMP) exerted by mono- or multi-articular movements. To determine the reproducibility of SMP, two identical test protocols were carried out on separate days for six movement types on seven adult males. The movement types were arm pull (AP), leg rise (LR), knee extension (KE), elbow flexion (EF), hip extension (HE) and squat movement (SQ). Variations in peak power obtained in two tests ranged from 0.7% (AP) to 9.6% (SQ). Coefficients of the test- re-test correlation in peak power ranged from 0.805 (SQ) to 0.961 (AP) and standard errors ranged from 4 W (EF) to 14 W (SQ). SMP in upper extremities increased from 166 W (EF) to 307 W (AP) resulting from the increase in velocity. However, in the movements of lower extremities, SMP increased from 506 W (KE) to 1351 W (SQ) as a result of the increase in force. To evaluate the validity of the SMP, a pull movement in weightlifting was tested and related to the athletic performance on weightlifters. Positive linear correlation (r = 0.862, p < 0.001) was observed between SMP and the total weight best records. It is concluded that this newly developed dynamometer has enough reproducibility and validity for evaluating the SMP, which is developed by various joint movement patterns related to the sport. The feasibility of applying this measuring protocol to the testing and training programmes for improving the daily living activities and athletic performances should now be assessed.     

A neural network-based approach for calculating dissolved oxygen profiles in reservoirs

A Neural Network (NN) modelling approach has been shown to be successful in calculating pseudo steady state time and space dependent Dissolved Oxygen (DO) concentrations in three separate reservoirs with different characteristics using limited number of input variables. The Levenberg–Marquardt algorithm was adopted during training. Pre-processing before training and post processing after simulation steps were the treatments applied to raw data and predictions respectively. Generalisation was improved and over-fitting problems were eliminated: Early stopping method was applied for improving generalisation. The correlation coefficients between neural network estimates and field measurements were as high as 0.98 for two of the reservoirs with experiments that involve double layer neural network structure with 30 neurons within each hidden layer. A simple one layer neural network structure with 11 neurons has yielded comparable and satisfactorily high correlation coefficients for complete data set, and training, validation and test sets of the third reservoir.     

A Comparison of the Accuracy and Reproducibility of Digital Three-Dimensional Coronary Artery Reconstructions Using Edge Detection or Videodensitometry

Global quantitative three-dimensional measurements of coronary arteries may be helpful in determining the functional significance of various forms of coronary pathology. A computerized system has been developed that is capable of performing 3-D reconstruction of digitized images obtained from multiple coronary angiographic views using either automated edge detection (AED) or videodensitometric (VD) techniques. To compare the accuracy and reproducibility of measurements obtained from this system using either technique, stationary and moving coronary aluminum 3-D phantoms, each with 13 branches (diameter 0.58–6.35 mm, length 21.5–64.5 mm), were imaged and reconstructed 10 separate times each. Individual branch lengths and diameters were calculated and compared to each other and to known values. Diameter measurements were compared using either AED or VD. Intraclass correlation coefficients between observed values (ICC) for vessel length werer= 0.89 for the stationary andr= 0.97 for the moving phantom. ICCs for vessel diameter werer= 0.93 (AED) andr= 0.95 (VD) for the stationary andr= 0.98 (AED) andr= 0.97 (VD) for the moving phantom. Mean differences (±SD) between true and observed values [MDTO(±SD)] for vessel length were −1.0 ± 3.9 mm for the stationary and −3.5 ± 3.2 mm for the moving phantom. MDTO(±SD) for vessel diameter were −0.10 ± 0.52 mm (AED) and +0.03 ± 0.30 mm (VD) for the stationary and −0.21 ± 0.44 mm (AED) and −0.12 ± 0.33 (VD) for the moving phantom. We conclude that the quantitative accuracy and reproducibility of measurements obtained by computerized 3-D reconstruction of coronary model phantoms is of high enough quality to warrant further clinical evaluation. VD appears to be more accurate than AED for measuring vessel diameter.     

Prediction of ground reaction forces and moments during sports-related movements

When performing inverse dynamic analysis (IDA) of musculoskeletal models to study human motion, inaccuracies in experimental input data and a mismatch between the model and subject lead to dynamic inconsistency. By predicting the ground reaction forces and moments (GRF&Ms) this inconsistency can be reduced and force plate measurements become unnecessary. In this study, a method for predicting GRF&Ms was validated for an array of sports-related movements. The method was applied to ten healthy subjects performing, for example, running, a side-cut manoeuvre, and vertical jump. Pearson’s correlation coefficient (\(r\)) and root-mean-square deviation were used to compare the predicted GRF&Ms and associated joint kinetics to the traditional IDA approach, where the GRF&Ms were measured using force plates. The main findings were that the method provided estimates comparable to traditional IDA across all movements for vertical GRFs (\(r\) ranging from 0.97 to 0.99, median 0.99), joint flexion moments (\(r\) ranging from 0.79 to 0.98, median 0.93), and resultant joint reaction forces (\(r\) ranging from 0.78 to 0.99, median 0.97). Considering these results, this method can be used instead of force plate measurements, hereby, facilitating IDA in sports science research and enabling complete IDA using motion analysis systems that do not incorporate force plate data.     

基于门控机制和卷积神经网络的中文文本情感分析模型

针对中文数据的特殊性导致判别时容易产生噪声信息,使用传统卷积神经网络（CNN）无法深度挖掘情感特征信息等问题,提出了一种结合情感词典的双输入通道门控卷积神经网络（DC-GCNN-SL）模型。首先,使用情感词典的词语情感分数对句子中的词语进行标记,从而使网络获取情感先验知识,并在训练过程中有效地去除了输入句子的噪声信息;然后,在捕获句子深度情感特征时,提出了基于GTRU的门控机制,并通过两个输入通道的文本卷积运算实现两种特征的融合,控制信息传递,有效地得到了更丰富的隐藏信息;最后,通过softmax函数输出文本情感极性。在酒店评论数据集、外卖评论数据集和商品评论数据集上进行了实验。实验结果表明,与文本情感分析的其他模型相比,所提模型具有更好的准确率、精确率、召回率和F1值,能够有效地获取句子的情感特征。     

基于神经网络模型的产品属性情感分析

针对基于词向量的神经网络模型在产品属性情感分析中效果不佳的问题,提出一种集成离散特征和词向量特征的开关递归神经网络模型。首先,通过直接循环图为语句建模,采用开关递归神经网络模型完成产品属性情感分析任务;然后,在开关递归神经网络模型中集成离散特征和词向量特征;最后,分别在流水线、联合、折叠三种任务模型中完成属性提取和情感分析任务。以宏观F1分数作为评估指标,在SemEval-2014的笔记本电脑和餐馆评论数据集上做实验。开关递归神经网络模型的F1分数为：48.21%和62.19%,超过普通递归神经网络模型近1.5个百分点,因而开关递归神经网络能够有效捕获复杂特征,提升产品属性情感分析的效果。而集成离散特征和词向量特征的神经网络模型的F1分数为：49.26%和63.31%,均超过基线结果0.5到1个百分点,表明离散特征和词向量特征互相促进,另一方面,也表明仅仅基于词向量的神经网络模型仍有提升空间。三种任务模型中,流水线模型的F1分数最高,表明应将属性提取和情感分析任务分开完成。     

Backscattering properties of marine phytoplankton Prorocentrum micans

Backscattering coefficients of Prorocentrum micans were determined in the laboratory using the Hydroscat-6 instrument. Ancillary parameters measured included absorption and scattering coefficients, chlorophyll-a concentration, cell size, and cell concentration. Spectral variability was found in the case of the backscattering coefficients, and the maximum backscattering coefficient was obtained at 488 nm, which deviated from the theory because of pigment absorption. We were unable to detect the effect of chlorophyll fluorescence on the shape of the backscattering coefficient spectra at 700 nm. However, the particulate backscattering coefficient was found to be well correlated with chlorophyll-a concentration at the six chosen wavelengths. At the same time, the backscattering ratio was accurately arranged in the range 0.012 to 0.019 at 442 nm and 0.013 to 0.021 at 620 nm. A positive correlation between particle density and backscattering ratio was established with a nonlinear regression model and the correlation coefficient was as high as 0.96 at 442 nm, this providing a good foundation for improving the accuracy of identifying the red tide alga, P. micans, for water colour remote sensing.     

醇类化合物色谱保留指数与结构参数定量关系的研究

采用拓扑指数(~mQ)、定位基参数(S_(ox))与醇在6种固定相上的气相色谱保留指数值(RI)进行了相关分析,发现RI与上述参数间存在良好的相关性,其关系式可表示为：RI=d (a~0)Q-(b~1)Q cS_(ox),其中a、b、c、d为系数。相关系数均大于0．98。继以留一法(Leave-one-out,LOO)进行交互检验,相关系数RCV均大于0．97。说明所建定量结构-保留关系(QSRR)模型具有良好的稳定性和预测能力,较好地揭示了醇类物质在不同固定相上气相色谱保留指数的变化规律。     

基于语义关系图的跨模态张量融合网络的图像文本检索

跨模态图像文本检索的难点是如何有效地学习图像和文本间的语义相关性。现有的大多数方法都是学习图像区域特征和文本特征的全局语义相关性或模态间对象间的局部语义相关性,而忽略了模态内对象之间的关系和模态间对象关系的关联。针对上述问题,提出了一种基于语义关系图的跨模态张量融合网络（CMTFN-SRG）的图像文本检索方法。首先,采用图卷积网络（GCN）学习图像区域间的关系并使用双向门控循环单元（Bi-GRU）构建文本单词间的关系;然后,将所学习到的图像区域和文本单词间的语义关系图通过张量融合网络进行匹配以学习两种不同模态数据间的细粒度语义关联;同时,采用门控循环单元（GRU）学习图像的全局特征,并将图像和文本的全局特征进行匹配以捕获模态间的全局语义相关性。将所提方法在Flickr30K和MS-COCO两个基准数据集上与多模态交叉注意力（MMCA）方法进行了对比分析。实验结果表明,所提方法在Flickr30K测试集、MS-COCO1K测试集以及MS-COCO5K测试集上文本检索图像任务的Recall@1分别提升了2.6%、9.0%和4.1%,召回率均值（mR）分别提升了0.4、1.3和0.1个百分点,可见该方法能有效提升图像文本检索的精度。     

基于BiGRU深度神经网络的心肌梗死检测

针对目前智能医疗诊断领域的研究现状,结合心电信号的时序性和多导联关联性特点,为降低心肌梗死疾病的误诊率,提出一种基于双向门控循环单元神经网络(Bidirectional Gated Recurrent Unit,BiGRU)和多导联心电图(electrocardiogram,ECG)信号的深度神经网络学习算法。对原始心电信号进行去噪处理,分割成心拍序列;将心拍序列送入深度神经网络训练模型学习分类;采用Physikalisch-Technische Bundesanstalt(PTB)心电数据库验证多导联BiGRU算法。算法对心梗检测的灵敏度为99.93%、特异性为99.72%、准确率为99.89%。实验结果表明,该算法的检测效果明显优于其他文献的检测算法,对提高心肌梗死的正确诊断率具有重要意义。     

Reconstruction of the electrocardiogram during heart surgery.

Electrocardiograms (ECG) recorded during arrhythmia surgery are used for identification of arrhythmias of different morphology. However, the interpretation of an intraoperative ECG is difficult because some leads cannot be recorded and the signals of the remaining leads often differ from those of a preoperative recording because of the sternotomy. Therefore, a method for reconstruction of a complete intraoperative ECG, which resembles a preoperatively recorded ECG, was studied in 24 patients undergoing heart surgery. The reconstruction method involves calculating coefficients for a transformation matrix, using a preoperative ECG recording and a first intraoperative ECG recording. Once this matrix has been established, further intraoperative recordings can be transformed into an ECG which strongly resembles a preoperative ECG. The correlation between reconstructed intraoperative leads and the corresponding preoperative leads was high in the leads Vx and Vy (median correlation coefficient 0.98 and 0.97) and slightly smaller in lead Vz (0.94). Further studies will prove if the method can be useful in arrhythmia surgery.     

Inter-rater reliability of output measures for a posture matching assessment approach: a pilot study with food service workers

Despite the ongoing health problem of repetitive strain injuries, there are few tools currently available for ergonomic applications evaluating cumulative loading that have well-documented evidence of reliability and validity. The purpose of this study was to determine the inter-rater reliability of a posture matching based analysis tool (3DMatch, University of Waterloo) for predicting cumulative and peak spinal loads. A total of 30 food service workers were each videotaped for a 1-h period while performing typical work activities and a single work task was randomly selected from each for analysis by two raters. Inter-rater reliability was determined using intraclass correlation coefficients (ICC) model 2,1 and standard errors of measurement for cumulative and peak spinal and shoulder loading variables across all subjects. Overall, 85.5% of variables had moderate to excellent inter-rater reliability, with ICCs ranging from 0.30-0.99 for all cumulative and peak loading variables. 3DMatch was found to be a reliable ergonomic tool when more than one rater is involved.     

Inter-rater reliability of output measures for a posture matching assessment approach: a pilot study with food service workers

Despite the ongoing health problem of repetitive strain injuries, there are few tools currently available for ergonomic applications evaluating cumulative loading that have well-documented evidence of reliability and validity. The purpose of this study was to determine the inter-rater reliability of a posture matching based analysis tool (3DMatch, University of Waterloo) for predicting cumulative and peak spinal loads. A total of 30 food service workers were each videotaped for a 1-h period while performing typical work activities and a single work task was randomly selected from each for analysis by two raters. Inter-rater reliability was determined using intraclass correlation coefficients (ICC) model 2,1 and standard errors of measurement for cumulative and peak spinal and shoulder loading variables across all subjects. Overall, 85.5% of variables had moderate to excellent inter-rater reliability, with ICCs ranging from 0.30–0.99 for all cumulative and peak loading variables. 3DMatch was found to be a reliable ergonomic tool when more than one rater is involved.     

Supporting executive functions during children's preliteracy learning with the computer

The present study examined how embedded activities to support executive functions helped children to benefit from a computer intervention that targeted preliteracy skills. Three intervention groups were compared on their preliteracy gains in a randomized controlled trial design: an experimental group that worked with software to stimulate early literacy with embedded executive‐functioning support (EL + EF group), an experimental group that worked with the same early literacy software without the explicit EF‐support (EL group) and a control group that played with language discovery games from the same software. These groups were contrasted on their preliteracy progress over time and on their learning behaviours while working with the software. Results showed that children in the two experimental groups progressed more in preliteracy skills on the long‐term, with the EL + EF group showing the highest learning effects. These results were mainly shown for the children with lower levels of EFs. Moreover, children in the EL + EF group showed more independent behaviour and played more games than children in the EL group. These results indicate that young children can show more effective learning over time during computer interventions when their executive functions are kept active.     

A latent class modeling approach to detect network intrusion

This study presents a latent class modeling approach to examine network traffic data when labeled abnormal events are absent in training data, or such events are insufficient to fit a conventional regression model. Using six anomaly-associated risk factors identified from previous studies, the latent class model based on an unlabeled sample yielded acceptable classification results compared with a logistic regression model based on a labeled sample (correctly classified: 0.95 vs. 0.98, sensitivity: 0.99 vs. 0.99, and specificity: 0.77 vs. 0.97). The study demonstrates a great potency for using the latent class modeling technique to analyze network traffic data.