Background and ObjectiveObstetricians use Cardiotocography (CTG), which is the continuous recording of fetal heart rate and uterine contraction, to assess fetal health status. Deep learning models for intelligent fetal monitoring trained on extensively labeled and identically distributed CTG records have achieved excellent performance. However, creation of these training sets requires excessive time and specialist labor for the collection and annotation of CTG signals. Previous research has demonstrated that multicenter studies can improve model performance. However, models trained on cross-domain data may not generalize well to target domains due to variance in distribution among datasets. Hence, this paper conducted a multicenter study with Deep Semi-Supervised Domain Adaptation (DSSDA) for intelligent interpretation of antenatal CTG signals. This approach helps to align cross-domain distribution and transfer knowledge from a label-rich source domain to a label-scarce target domain.MethodsWe proposed a DSSDA framework that integrated Minimax Entropy and Domain Invariance (DSSDA-MMEDI) to reduce inter-domain gaps and thus achieve domain invariance. The networks were developed using GoogLeNet to extract features from CTG signals, with fully connected, softmax layers for classification. We designed a Dynamic Gradient-driven strategy based on Mutual Information (DGMI) to unify the losses from Minimax Entropy (MME), Domain Invariance (DI), and supervised cross-entropy during iterative learning.ResultsWe validated our DSSDA model on two datasets collected from collaborating healthcare institutions and mobile terminals as the source and target domains, which contained 16,355 and 3,351 CTG signals, respectively. Compared to the results achieved with deep learning networks without DSSDA, DSSDA-MMEDI significantly improved sensitivity and F1-score by over 6%. DSSDA-MMEDI also outperformed other state-of-the-art DSSDA approaches for CTG signal interpretation. Ablation studies were performed to determine the unique contribution of each component in our DSSDA mechanism.ConclusionsThe proposed DSSDA-MMEDI is feasible and effective for alignment of cross-domain data and automated interpretation of multicentric antenatal CTG signals with minimal annotation cost.

Intelligent classification of cardiotocography (CTG) based on machine learning (ML), a useful tool to improve the accuracy of fetal abnormality detection, can assist obstetricians with clinical decisions. With the advancement of information technologies and medical devices, there are development opportunities for multicenter clinical research and obtaining more digital CTG signals. However, most of the existing clinical multicenter CTG datasets are partially annotated and have discrepancies which do not satisfy the ML condition of independent identical distribution. Therefore, this paper focuses on an unsupervised domain adaptation (UDA) algorithm to realize cross-domain intelligent classification of multimodal CTG signals. We propose a method dubbed domain adversarial training of neural network for multicenter CTG (DANNMCTG), which mainly consists of a label classifier, a feature extractor and a domain discriminator. To match different distribution of fetal heart rate (FHR), uterine contraction (UC) and fetal movement (FetMov) signals, we condition the domain alignment on label predictions by defining the multi-linear map. For analysis, two datasets from the hospital central station and home monitoring devices were considered as the source and target domains. The results showed that the accuracy value, F1 value and area under the curve (AUC) value of the DANNMCTG were 71.25%, 76.08% and 0.7705, respectively. This method significantly improved the performance of the deep learning models without exploiting any information in the target domain, and outperformed the state-of-the-art UDA algorithms for CTG classification. In summary, the DANNMCTG can effectively mitigate the influence of domain shift for multicenter intelligent prenatal fetal monitoring.

随着大数据时代的到来,掌握数据科学的相关知识和方法将成为医工融合专业学生的一项核心竞争能力和必备能力.结合医学大数据与人工智能本科生创新实验室的建设,文章探索并实践如何把"本科生进实验室"与"毕业设计论文"有机结合的培养模式,充分利用大学四年时间培养医工融合专业学生的数据科学能力,以期满足"医工结合"背景下的数据科学素质能力要求.

Purpose: To enable the in-home diagnosis of heart failure (HF) based on morphological features of high quality ballistocardiography (BCG) signals and respiratory effort. Methods: Non-contact vital signs including BCG and respiratory effort signals from 25 subjects (11 HF, 14 non-heart failure (Non-HF)) were collected using a force sensor-based medical equipment. By assessing the recorded BCG signals w.r.t signal quality indexes, a steady-state BCG template is modeled by using consecutive high quality BCG signals, from which morphological features including the amplitude, time, area and energy features of signal wave groups are extracted to distinguish the HF and Non-HF subjects. Results: It is validated that a total 13 morphological features of BCG and respiratory effort signals showed differences between HF and Non-HF subjects. Using typical classifiers for discriminating HF and Non-HF subjects yields the accuracy, sensitivity and specificity of 92%, 80% and 100%. Conclusion: The acquisition and analysis of high quality BCG signals has the potential of identifying HF disease.

文章以生物医学工程专业《信号与系统》课程教学为例,结合教学实践经验和教学效果评价分析,将O-PIRTAS教学方法引入到课程教学的过程中.该教学方法将教学分为七个必要的环节:教学目标、课前准备、教学视频、视频回顾、知识测试、活动探究和总结提升,其兼具理论合理性和实践可行性.实践结果表明,此模式有效地调动了学生的学习热情,充分地激发了学生的学习积极性,提升了教学效果,实现了课程的知识技能、过程方法和情感态度价值观的三维目标.

Late fetal growth restriction (FGR) is a common complication of pregnancy characterized by chronic hypoxia. However, late FGR is in a dilemma of the high incidence but low detection rate. Depending on the non-invasiveness and convenient operation, the routine cardiotocography (CTG) allows continuous monitoring fetal heart rate (FHR) to assess fetal intrauterine stockpiling ability. In this paper, we aimed to explore the FHR pattern of late FGR in routine CTG. For analysis, the FHR features were acquired using routine CTG in a population of 160 healthy and 102 late FGR fetuses published in IEEE Dataport. First, we explored the relationships among FHR features and their importance on late FGR assessment by utilizing hypothesis testing, principal component analysis (PCA) and Spearman correlation analysis. Second, we presented a regression coefficient-based backward-stepwise-selection of association rules analysis (ARA) called backward-stepwise Max-R 2 Apriori ARA, to find the optimum itemset that helps diagnose late FGRs from healthy fetuses. The hypothesis testing, PCA and Spearman correlation analysis found eight FHR features were highly relevant to the late FGR. Moreover, the backward-stepwise Max-R2 Apriori ARA validated the correlation and interpretation about FHR features of late FGR. In conclusion, the analysis results are consistent with clinical knowledge on late FGR and help screen late FGR in antepartum fetal monitoring.

《机器学习导论》作为生物医学工程专业本科阶段的一门重要的专业课,涉及的先修知识广泛,要求学生具备一定的专业知识和综合应用能力.单单基于传统讲授法,很难取得良好的教学效果.针对该课程在传统教学模式下存在的主要问题,该文提出一种基于O-PIRTAS翻转课堂的教学新模式,将生物医学工程方面的人工智能应用与该课程教学重点有机结合,旨在通过增加学生的学习胜任力,激发其内在的学习动机,提高教学效果.

Heart rate measurement through Ballistocardiogram (BCG) signal is an efficient method for long-term cardiac activity monitoring in real-time, especially for patients with cardiovascular and cerebrovascular disease. In this study, we propose a one-dimensional (1D) U-net++ to identify the position of J-peak in BCG signals automatically. The proposed 1D U-net++ is based on a 1D convolution neural network through dense skip connection backward transfer data features. The low-level and high-level data features of the BCG signals are combined with the last layer features of 1D U-net++ to shorten the semantic gap when the encoder and decoder feature skip connection. The BCG signals of eight healthy subjects were collected for experimental verification, and the accuracy and precision of J-peak detection reached 99.4% and 99.3%, respectively. The experimental results demonstrate that our proposed method can effectively identify J-peak in BCG signal.

The sparse distribution of targets in monitored areas is an important prior for device-free localization (DFL) with radio tomography networks. In this article, our goal is to develop an enhanced sparse representation-based DFL method that takes the full potential of sparsity for location reconstruction. An expanded sensing matrix spanning the concatenation of a sampling matrix and a unit error-correcting base is proposed for modelling the measurement process. The sampling matrix can either be composed of the ellipse model from calibrated networks or the received signal strength (RSS) fingerprint-based model induced by training samples with one person at predefined locations. Thus, the sparsity of targets is enhanced under the expanded sensing matrix and the ℓ 1 -minimization-based approximations are derived for the recovery of locations. Experimental studies in an open outdoor scenario, in a line-of-sight (LOS) indoor scenario, and in a non-line-of-sight (NLOS) indoor scenario, are conducted to verify the efficacy of the proposed method.

实践教学环节是培养具有实践能力与创新人才的重要基础,大学生实践动手能力的提高已经成为当前我国高等教育亟待加强的重要任务之一.高校实验室作为开展实践教学和科学研究的重要平台,其传统的建设和管理模式弊端愈发突显,亟待改进.提出一种基于物联网技术的应用于高等院校的创新实验室架构,探讨该架构下各功能模块的设计.基于此创新实验室,开展多层次的学生创新实践能力培养活动,真正做到以学生为中心,有效促进学生实践能力的提升.

本文讨论了如何在程序设计基础课程中实现交互式的课堂教学.通过使用Jupyter Notebook和Python Tutor等工具,提高课堂教学效率,激发学生的学习热情.本文提出的教学设计,也适用于其它编程语言的教学.

Infrared radiation changes (IRC) induced by human motion can provide important clue for motion classifica-tion. This paper presents a hidden Markov model (HMM)-based compressive infrared classification method to recognize human motions. In order to solve the problem of self-occlusion, an orthogonal-view based compressive infrared sensing system is implemented for pro jecting the IRC to two orthogonal planes in the infrared radiation field. Then, a double-layer feature model using HMM classifier is trained to carry out motion recognition with the compressive measurements. Experimental results show that the mean correct classification rate with double-layer feature is 95.71%, which is better than that with main-layer feature. This method provides a new approach to classification of human motions for ambient assisted system.

Indoor human tracking and activity recognition are fundamental yet coherent problems for ambient assistive living. In this paper, we propose a method to address these two critical issues simultaneously. We construct a wireless sensor network (WSN), and the sensor nodes within WSN consist of pyroelectric infrared (PIR) sensor arrays. To capture the tempo-spatial information of the human target, the field of view (FOV) of each PIR sensor is modulated by masks. A modified partial filter algorithm is utilized to decode the location of the human target. To exploit the synergy between the location and activity, we design a two-layer random forest (RF) classifier. The initial activity recognition result of the first layer is refined by the second layer RF by incorporating various effective features. We conducted experiments in a mock apartment. The mean localization error of our system is about 0.85 m. For five kinds of daily activities, the mean accuracy for 10-fold cross-validation is above 92%. The encouraging results indicate the effectiveness of our system.

实现室内人体定位跟踪与动作智能识别在人口老龄化社会具有重要的现实意义。本文提出了一种通过构造无线传感器网络(Wireless sensor network, WSN)同时解决这两个相关问题的方法。在WSN中，每个热释电红外(Pyroelectric Infrared, PIR)传感器的视场(Field of View, FOV)通过两个自由度(Degrees of freedom, DOF)分割来实现调制，通过位置信息的编码解码实现了人体目标的粗略定位。通过相邻两个传感器节点的数据融合扩大了监测区域，同时提高了人体定位的精确度。动作的持续时间是动作识别的一个关键特征，为此本文构造了一个两层的随机森林(Random Forest, RF)分类器。第一层随机森林用于识别每个数据帧的动作类型，第二层随机森林利用相同动作的持续时间作为有效的特征判断最终的动作类型。实验在真实的室内环境中进行，5种日常动作的10折交叉验证平均准确率高于93%。实验结果表明本文提出的方法可以同时有效地实现人体定位跟踪与日常动作识别。 Human locomotion tracking and activity recognition in the indoor environment are fundamental problems for healthy aging. In this paper, we propose a method to deal with these two coherent problems simultaneously by constructing a wireless sensor network (WSN). In the WSN, the Field of View (FOV) of each Pyroelectric Infrared (PIR) sensor is modulated by two degrees of freedom (DOF) segmentation, which provides coarse location information of the human target. Data fusion of the adjacent sensor nodes enlarges the monitored region and improves the human localization accuracy. To incorporate the activity lasting time as a crucial cue for activity recognition, we build a two-layer Random Forest (RF) classifier. The first layer is utilized to label the activity type for each data frame, and the second layer will utilize the lasting time of the same activity as a useful feature for the final activity classification. We conducted experiments in a mock apartment, and the average mean accuracy for 10-fold cross validation of 5 kinds of daily activities is above 93%. The encouraging results show that our method can achieve human tracking and daily activity recognition simultaneously and effectively.

Objectives: Banxia Baizhu Tianma Decoction (BBTD) is widely used to treat vertebrobasilar insufficiency vertigo (VBIV) in China, but its efficacy remains largely unexplored. We systemically summarized relevant evidence from randomized controlled trials (RCTs) to assess the therapeutic effect of BBTD.Methods: Seven electronic databases were searched for relevant electronic studies published before July 2016. We evaluated RCTs that compared BBTD, anti-vertigo drugs and a combination of BBTD and anti vertigo drugs. We performed a meta-analysis in accordance with the Cochrane Collaboration criteria. The outcomes were clinical efficacy (CE), blood flow velocity of the vertebrobasilar artery by transcranial Doppler (TCD), and adverse effects.Results: Twenty-seven studies with a total of 2796 patients were identified. Compared with anti-vertigo drugs, BBTD showed slight effects on CE (n = 350; RR, 1.09; 95% CI, 1.01-1.18; p = 0.03; I-2 = 0%); however, BBTD plus anti-vertigo drugs (BPAD) significantly improved the clinical efficacy (n = 2446; RR, 1.20; 95% CI, 1.16-1.24; p < 0.00001; I-2 = 0%) and accelerated the blood flow velocity of the left vertebral artery (LVA) (n = 1444; WMD, 5.21 cm/s; 95% CI, 3.72-6.70 cm/s; p < 0.00001; I-2 = 91%), the blood flow velocity of the right vertebral artery (RVA) (n = 1444; WMD, 5.45 cm/s; 95% CI, 4.02-6.88 cm/s; p < 0.00001; I-2 = 89%), and the blood flow velocity of the basilar artery (BA) (n = 1872; WMD, 5.20 cm/s; 95% CI, 3.86-6.54 cm/s; p < 0.00001; I-2 = 90%). Adverse effects were mentioned in six studies.Conclusions: The current evidence indicates that BPAD is effective for the treatment of VBIV, but the efficacy and safety of BBTD is uncertain because of the limited number of trials and low methodological quality. Hence, high-quality and adequately powered RCTs are warranted. (C) 2017 Elsevier Ltd. All rights reserved.

Radio tomographic networks based imaging is bringing significant impact in activity sensing. In this article, we proposed an abnormal activity detection method without any computed recovery imaging. By organizing a vertically arranged profile-aware network, the critical state feature of abnormal activity is encoded into data stream of received signal strengths (RSSs). Then, the new coming sensor data is compared with the instantaneous state feature already recorded, and abnormal detection is performed according to similarity. To validate the efficacy of our method, we defined walking as normal activity and fall as abnormal activity in indoor environments. Experiments give the encouraging results.

Segmenting leaf images with complex background is currently a difficult and focal point of research.In this paper,we propose a method which combines simple man-machine interaction with marker-based watershed segmentation to solve the problem of leaf image segmentation effectively.First,the method lets the user mark the peripheral and exterior points of a leaf in odd-even sequence on the leaf image under complex background;after a series of processing the marked image is produced.Secondly,the method transforms the leaf image under complex background into greyscale image and L* a* b* image,then applies the marker-based watershed segmentation to greyscale image,a* component image and b* component image respectively with the previous marked image as parameters.Finally,the final segmentation result is achieved comprehensively by means of voting.The segmentation experiments on 200 leaf images in 20 categories under complex backgrounds indicate that this method can realise precise segmentation,and can preserve the detailed parts of a leaf.There are tiny mistakes in some segmentations,but have slight impact on leaf shapes.

目的:系统评价中医辨证论治联合化疗方案治疗晚期非小细胞肺癌(non-small cell lung cancer,NSCLC)的临床效果以及安全性.方法:计算机检索The Cochrane Library,Pubmed,Embase,中国科技期刊全文数据库(VIP),万方数据库,中国期刊全文数据库(CNKI),中国生物医学文献数据库(CBM)等,检索时间限定为建库至2015年7月5日,检索所有中医辨证论治联合化疗方案治疗晚期非小细胞肺癌的随机对照试验(randomized controlled trials,RCTs),并追索纳入研究的参考文献.由两位评价者独立对纳入研究的质量进行严格评价和资料提取后,采用RevMan 5.3软件进行Meta分析.结果:最终纳入12项RCT,共1 341例患者.Meta分析结果显示:与单纯化疗相比,中医辨证论治联合化疗方案可有效提高临床近期疗效[OR=1.58,95％ CI(1.03,2.42),P=0.03],改善患者生存质量[OR =4.38,95％ CI(3.17,6.05),P＜0.000 01].不良反应方面,中医辨证论治联合化疗方案能减少化疗引起的骨髓抑制:白细胞下降[OR =0.21,95％ CI(0.10,0.44),P＜0.000 1],血红蛋白下降[OR =0.29,95％ CI(0.13,0.68),P=0.004],血小板下降[OR =0.26,95％ CI(0.10,0.69),P=0.007],减少消化道不良反应[OR=0.23,95％ CI(0.09,0.63),P=0.004],但在减少肝功能损害[OR =0.37,95％ CI(0.08,1.66),P=0.19],肾功能损害[OR =0.51,95％ CI(0.23,1.14),P=0.10]方面两组差异均无统计学意义.结论:中医辨证论治联合化疗方案治疗晚期非小细胞肺癌能有效提高临床近期疗效和改善患者生存质量,减少化疗引起的严重骨髓抑制和消化道不良反应,但在减少肝肾功能损害方面和单纯使用化疗方案相比无显著性差异.鉴于本研究纳入RCTs数量和质量有限,中医辨证论治联合化疗方案治疗晚期非小细胞肺癌的临床疗效以及安全性仍需要严格的、大样本的随机双盲试验加以验证.

TCM medical cases in records are free text with much valuable data and clinical terms, how to recognize and extract these clinical terms automatically is a valuable work. TCM medical records obtained from Guangdong Provincial Hospital of Chinese Medicine are segmented to single word and labeled with five labeling features(words in sentence, grammatical property of words, words in clinical dictionary, set phrases acting on neighbor context, and set phrases acting on far distance.), and divided into training sets and testing sets. Training sets are also handled with outputted labeling (labeling of symptoms or signs, TCM diagnosis, TCM syndrome type, Chinese medicines (drug), and Names of TCM prescriptions.). In order to evaluate abilities of labeling features on improving clinical terms recognition with CRF, three indicators (recognition Precision (P), recognition Recall (R) and F-score (F)) are defined, and three comparisons are given: comparisons of individual labeling features, comparisons of combined labeling features, and comparisons of combined features in different diseases. The results show that, "grammatical property of words" is the best labeling features in all individual labeling features. Multi-combined features have higher scores than individual labeling features on improving clinical terms recognition. The combined mode of "grammatical property of words", "words in sentence", and "words in clinical dictionary" may be the most suitable labeling features. Multi-combined labeling features can improve term recognition with CRF model for text mining in TCM medical cases.

Human daily activity recognition is the foundation of automatic Ambient Assisted Living (AAL) system. In the paper, we propose a sensing model which can capture the discriminative spatio-temporal feature of human motion in an efficient way. The object space is separated into distinct discrete sampling cells by reference structure, and the ceiling mounted Pyroelectric infrared (PIR) sensors are used to capture the time-varying signal induced by human motion. The GMM-HMM model is utilized to classify different human activities. We use a self-developed PIR sensor node mounted on the ceiling to conduct experiments in a real office environment. Promising experimental results confirm the validity of our model.