关键词： Artificial intelligence   Computed tomography   Hematoma   Algorithms   Decision curve analysis  

摘要： Purpose The objective of this study was to predict hematoma expansion (HE) by radiomic models based on different machine learning methods and determine the best radiomic model through the comparison. Method A total of 108 patients with intracerebral hemorrhage were retrospectively evaluated. Images of baseline non-contrast computed tomography (NCCT) and follow-up NCCT scan within 24 h were retrospectively reviewed. An HE was defined as a volume increase of more than 33% or an increase greater than 12.5 mL from the volume of the baseline NCCT. Texture parameters of the baseline NCCT images were selected by the least absolute shrinkage and selection operator (LASSO) regression. We used support vector machine (SVM), decision tree (DT), conditional inference trees (CIT), random forest (RF), k-nearest neighbors (KNN), back-propagation neural network (BPNet) and Bayes to build models. Receiver operating characteristic (ROC) analysis and decision curve analysis (DCA) was performed and compared among models. Results Every model had a relatively high AUC (all > 0.75), SVM and KNN had the highest AUC of 0.91. There were significant differences between SVM and CIT (Z > 2.266, p = 0.02345), KNN and CIT (Z = 2.4834, p = 0.01301), RF and CIT (Z = 2.6956, p = 0.007027), KNN and BPNet (Z = 2.0122, p = 0.0442), RF and BPNet (Z = 1.9793, p = 0.04778). There was no significant difference among SVM, DT, RF, KNN and Bayes (p > 0.05). The SVM obtained the largest net benefit when the threshold probability was less than 0.33, while KNN obtained the largest net benefit when the threshold probability was greater than 0.33. Combined with ROC and DCA, SVM and KNN performed better in all the models for predicting HE. Conclusion Radiomic models based on different machine learning methods can be used to predict HE and the models generated by SVM and KNN performed best.

关键词： Parkinson's disease   wearable technology   sensors   internet of things   artificial intelligence   machine learning   deep learning   remote monitoring   smart personalized healthcare  

摘要： Parkinson's disease is a chronic neurodegenerative disease that affects a large portion of the population, especially the elderly. It manifests with motor, cognitive and other types of symptoms, decreasing significantly the patients' quality of life. The recent advances in the Internet of Things and Artificial Intelligence fields, including the subdomains of machine learning and deep learning, can support Parkinson's disease patients, their caregivers and clinicians at every stage of the disease, maximizing the treatment effectiveness and minimizing the respective healthcare costs at the same time. In this review, the considered studies propose machine learning models, trained on data acquired via smart devices, wearable or non-wearable sensors and other Internet of Things technologies, to provide predictions or estimations regarding Parkinson's disease aspects. Seven hundred and seventy studies have been retrieved from three dominant academic literature databases. Finally, one hundred and twelve of them have been selected in a systematic way and have been considered in the state-of-the-art systematic review presented in this paper. These studies propose various methods, applied on various sensory data to address different Parkinson's disease-related problems. The most widely deployed sensors, the most commonly addressed problems and the best performing algorithms are highlighted. Finally, some challenges are summarized along with some future considerations and opportunities that arise.

关键词： 滑坡   机器学习   滑坡易发性   三峡库区   深度学习  

摘要： 中国山区多、地形复杂,构造发育、地质灾害隐患分布广泛。滑坡作为山区最具灾难性的地质灾害之一,严重威胁着人民群众的生命及财产安全。构建滑坡易发性模型能够量化滑坡发生的可能性,对制定防灾措施、减少潜在风险具有重要作用。由于经验驱动模型难以量化,且往往依赖主观判断,近年来,滑坡易发性模型的精度与准确度在从经验驱动和统计理论模型向新兴机器学习方向发展的过程中得到提升。对目前滑坡易发性评价常用的机器学习模型进行综合评述,并针对三峡库区的案例研究,对不同的机器学习技术进行广泛分析和比较。机器学习模型通过结合实地调查资料和历史数据,可绘制滑坡易发性地图,辅助制定滑坡减缓策略。根据滑坡易发性预测模型的准确性和效率,评价几种常用算法的优势和局限性。结果表明,与一些常用的滑坡易发性制图方法相比,基于树结构的集成算法模型性能更好。此外,高质量的数据库十分重要,深度学习算法的更多应用还有待进一步研究探索。

关键词： Additive manufacturing   Machine learning   Grain structure   Neural networks   Cellular automaton method  

摘要： Directed energy deposition (DED) is an advanced additive manufacturing technology for the fabrication of near -net-shape metal parts with complex geometries and high performance metrics. Studying the grain structure evolution during the process is pivotal to evaluating and tailoring the as-built products' mechanical properties. However, it is time-consuming to simulate the multi-layer deposition process using the physics-based numerical model to optimize the process parameters for achieving the desired microstructure. In this paper, a physics-informed machine learning algorithm to predict the grain structure in the DED process is proposed. To generate training data for the machine learning algorithm, we use an experimentally validated cellular automaton finite volume method (CAFVM) for DED Inconel 718, where CA is applied to model the grain structure and FVM to simulate the heat transfer. We develop a neural network model to identify the correlation between the local thermal features and their corresponding grain structure characteristics. The inputs and outputs of the neural network (NN) model are selected based on the governing physics, and a novel way to extract them is proposed. The NN model can quickly predict the grain structure characteristics with the local thermal data for thin-wall builds, and the predictions are in good agreement with the numerical simulation results. We expect the proposed method can benefit other metal additive manufacturing technologies to formulate efficient and accurate process-structure relationships and in-process feedback control.

关键词： 信息赋权能力   全球减贫   互联网普及率   机器学习方法   贫困  

摘要： 伴随信息通讯技术的不断发展,如何通过提升信息赋权能力进而有效减少贫困是实践者和研究者关注的焦点,但关于信息赋权能力促进减贫的内在机制依然尚无共识。基于1990—2017年124个国家(地区)的面板数据,通过机器学习中的随机森林方法实现对控制变量的客观选取,不同的实证模型表明:信息赋权能力虽然对不同国家(地区)降低贫困发生率具有显著的正向影响,但这一影响呈现出两个临界拐点的非线性特征。进一步分析发现信息赋权能力对减贫的长、短期动态效应存在显著差异,同时这一影响机制既展现在信息赋权能力对贫困发生的外部环境改善,形成减贫的外生刺激机制,还呈现在信息赋权能力降低减贫过程中对内生环境的路径依赖机制。利用工具变量等不同稳健性检验方法验证上述结果的一致性。这一研究发现为后发国家通过信息赋权能力提升来实现发展的弯道超车提供理论与实证支撑,也为中国反贫困实践的理论提炼和经验的国际推广提供了新视角。最后讨论了本研究的理论、方法和政策层面启迪。

关键词： 智能电表   机器学习   数据预处理   融合算法   故障预测  

摘要： 针对智能电表故障数据规模大、维度高、存在错误及异常数据的特点,提出了一种融合多分类机器学习模型的智能电表故障预测方法。采用正态分布补全及箱型图方法,对原始数据集进行缺失值填补及异常值替换;通过计算特征属性与故障类型的相关系数,消除不相关特征,形成特征子集;构建混合采样策略,解决故障数据不平衡问题。计算三种典型机器学习算法处理智能电表故障数据的预测准确率,构建混淆矩阵;计及各分类器的预测能力,构建多分类器融合决策函数。最后,分别采用公共数据集与实际用电数据作为样本,验证了所提方法的有效性。

关键词： 高密度   水位监测   概念模型   地下水   补给边界   排泄边界  

摘要： 随着我国地下水监测工作的高速发展,高频率高密度水位监测数据的出现催生了对其进行深入信息挖掘的需求。在传统地下水模型研究中,地下水水位监测值常位于模型构建过程的下游,当水位监测的时空密度逐渐增大时,新增信息无法有效传导至模型的规划阶段并指导概念模型的修订。文章提出了一种地下水系统补排边界的识别方法,在不建立地下水数值模型的前提下,以监测井空间位置为节点,按照德劳内原则建立三角网格。在此网格系统中,首先定义一个水力梯度变换函数gradF,以求取网格中任意位置的水力梯度;借鉴机器学习领域的优化算法,使用水力梯度场驱动含水层中随机分布质点的运行轨迹,并以此推断和识别区域内地下水补给和排泄边界。在环境地学计算平台EnviFusion-CGS中实现,并构建了详细工作流程。以山东省青岛市大沽河中下游含水层为示范区,对含水系统的补给区和排泄区的空间分布及其动态变化进行了分析,取得了良好效果。本研究为构建和修订已有含水层概念模型提供了新思路。

关键词： 农业干旱   机器学习   SPEI   MODIS  

摘要： 频繁的旱灾对甘肃省经济、农业生产造成严重危害,因此利用先进方法建立准确可靠的农业干旱监测模型对该省防旱减灾十分重要。本文基于随机森林(RF)、BP神经网络和支持向量机(SVM)3种机器学习方法,利用甘肃省2002—2019年4—10月多源遥感数据得到的植被状态指数(VCI)、温度状态指数(TCI)、植被供水指数(VSWI)、降水状态指数(PCI)以及DEM、土壤有效含水量(AWC)和气候类型作为自变量,气象站点以3个月时间尺度的标准化降水蒸发指数(SPEI_3)为因变量,构建3种不同的农业干旱监测模型,分析比较出适用于监测甘肃省农业干旱的最佳模型,同时进一步探究了机器学习方法构建的模型在不同环境下的适用性。结果表明:构建的3种机器学习模型中,随机森林模型的R2平均值高(0.86)且误差小(RMSE为0.40,MAE为0.31),农业干旱的监测效果要优于BP神经网络和支持向量机模型;干燥和湿润两种环境下分别构建的3种机器学习模型在湿润环境中监测能力表现均更优异(R^(2)>0.82),而随机森林模型在两种环境中监测干旱的表现比其他两种模型强。研究结果为甘肃省的农业干旱监测与评估提供了新的科学方法,对农业干旱研究具有重要意义。

关键词： 干眼综合征   机器学习   眨眼  

摘要： 目的探讨干眼患者自发瞬目模式的特点。方法横断面研究。连续纳入2019年1至12月在首都医科大学附属北京同仁眼科中心角膜病专科门诊就诊的干眼患者357例作为干眼组,其中男性102例,女性255例;年龄(46.2±13.3)岁;同时纳入健康志愿者152名作为对照组,其中男性32例,女性120例;年龄(48.1±13.9)岁。所有患者进行问卷调查眼表疾病评分指数、瞬目视频获取、泪膜破裂时间(BUT)检查、角结膜荧光素染色、基础泪液分泌试验。将瞬目视频的单帧图片输入UNet分割算法与ResNet分类算法建立的模型进行分析,获取睑裂高度百分比绘制瞬目波。将完全瞬目分为A、B、C型,不完全瞬目分为Ⅰ、Ⅱ、Ⅲ型,进而分析干眼患者的自发瞬目特征,并使用独立样本t检验和Wilcoxon秩和检验分析其与对照组差异的统计学意义。结果本研究建立的分割模型与分类模型准确度分别为96.3%与96.0%,与人工分析的一致性为97.9%。干眼组患者瞬目频率为30(18,42)次/min,显著高于对照组的20(9,46)次/min(U=18132.50,P=0.002),但完全瞬目次数为6(3,24)次/min,明显低于对照组的12(3,33)次/min(U=12361.00,P=0.016),不完全瞬目次数为15(6,27)次/min,明显高于对照组的3(0,10)次/min(U=22839.00,P<0.001)。完全瞬目中,干眼患者A型瞬目占比显著高于对照组[53.7%(2796/5177)和39.3%(633/1698);χ^(2)=101.83,P<0.001];不完全瞬目中,干眼患者Ⅱ型瞬目占比显著高于对照组[36.0%(2334/6477)和29.6%(126/426);χ^(2)=6.99,P=0.008]。干眼患者平均瞬目间期为1.5 s,与对照组2.2 s比较,差异无统计学意义(U=15230.00,P=0.093);干眼患者眼睑完全闭合期为0.8 s,明显短于对照组1.3 s(U=16291.50,P=0.006)。闭眼期、睁眼前期、睁眼末期与对照组比较,差异均无统计学意义(均P>0.05)。结论干眼患者不完全瞬目次数增加、完全瞬目次数减少,眼睑完全闭合时间明显缩短;其瞬目模式以闭合幅度减少的Ⅱ型不完全瞬目及闭合时间缩短的A型完全瞬目为主。

关键词： Sewage sludge   Coffee industry residue   Co-combustion   TG-FTIR   Artificial neural network   Principal component analysis  

摘要： Faced with the problems of the environment and resources, the co-combustion of sewage sludge has gradually become a trend. In this paper, the combination of thermogravimetric and Fourier transform infrared spectroscopy (TG-FTIR) was used to study the co-combustion of sewage sludge (SS) and coffee industry residues (CIR). SS and CIR were mixed according to the mass ratios of 1:0, 9:1, 7:3, 5:5, 3:7, 1:9, and 0:1, and the temperature was programmed to be heated to 800 degrees C in the atmosphere with an air flow rate of 50 mL/min. The percentages of mass loss during combustion of SS and CIR were 55.8% and 96.8%, respectively. As the mass percentage of CIR increased, the comprehensive combustion index (CCI) of the sample improved. The Kissinger-Akahira-Sunose (KAS), Flynn-Wall-Ozawa (FWO), and Starink methods were used to evaluate the activation energy (E-alpha) of the reaction. Principal component analysis (PCA) and artificial neural network (ANN) were used to determine the main reaction and predict experimental data of co-combustion of SS and CIR, respectively.