Automatic classification and prediction models for early Parkinson’s disease diagnosis from SPECT imaging

Early and accurate diagnosis of Parkinson’s disease (PD) is important for early management, proper prognostication and for initiating neuroprotective therapies once they become available. Recent neuroimaging techniques such as dopaminergic imaging using single photon emission computed tomography (SPECT) with ¹²³I-Ioflupane (DaTSCAN) have shown to detect even early stages of the disease. In this paper, we use the striatal binding ratio (SBR) values that are calculated from the ¹²³I-Ioflupane SPECT scans (as obtained from the Parkinson’s progression markers initiative (PPMI) database) for developing automatic classification and prediction/prognostic models for early PD. We used support vector machine (SVM) and logistic regression in the model building process. We observe that the SVM classifier with RBF kernel produced a high accuracy of more than 96% in classifying subjects into early PD and healthy normal; and the logistic model for estimating the risk of PD also produced high degree of fitting with statistical significance indicating its usefulness in PD risk estimation. Hence, we infer that such models have the potential to aid the clinicians in the PD diagnostic process.     

Relationship between chemical degradation and thickness loss of an amine-cured epoxy coating exposed to different UV environments

The relationship between chemical degradation and thickness loss of an unpigmented, non UV-stabilized, crosslinked amine-cured epoxy coating exposed to three UV conditions was investigated. Spin-coated samples having a thickness of approximately 7 μm on an Si substrate were prepared from a stochiometric mixture of a bisphenol A epoxy resin and a tetra-functional amine curing agent. Samples were exposed outdoors and to two accelerated laboratory UV environments. Chemical degradation and thickness loss were measured by transmission Fourier transform infrared spectroscopy (FTIRS) and laser scanning confocal microscopy (LSCM), respectively. In addition, surface roughness and morphological changes were measured by atomic forcemicrosocopy (AFM) and LSCM. Substantial chemical degradation, thickness loss, and morpholocal changes occurred in the exposed films, and the rate of chemical degradation was greater than that due to the thickness loss. This additional chemical loss was attributed to an inhomogeneous degradation process in which nanoscale localized depressions initiate at certain sites on the surface, which then enlarge and deepen with exposure time. The results of this study provide a better understanding of the degradation mechanism and should lead to the development of scientific-based models for predicting the service life of crosslinked amine-cured epoxy coatings. Presented at the 82nd Annual Meeting of the Federation of Societies for Coatings Technology, October 27–29, 2004, in Chicago, IL     

基于前向预测策略的块正交匹配追踪算法研究

在块正交匹配追踪算法中,候选集匹配原子块的选择策略对信号重建具有重要作用,但是该算法在迭代选择原子块的过程中,每次选择当次最优迭代的原子块,并不能保证最终迭代性能是最优的。为此,提出一个基于前向预测的最优匹配原子块选择策略的块正交匹配追踪算法,通过预测原子块在未来迭代过程中的性能选择最优匹配的原子块。实验结果表明,与块正交匹配追踪算法相比,该算法的重建误差随前向预测参数的增大而减小。     

一种HEVC帧内预测快速算法

HEVC作为新一代的视频编码标准,比现有H.264标准的压缩效率提高近一倍,但其存在复杂度较高的问题。为此,针对HEVC中帧内预测最耗时的模块,即编码单元块划分模块和帧内预测模式选择模块,提出一种适合HEVC帧内预测的快速算法。该算法将率失真(RD)代价作为阈值参数,利用候选模式集中预测模式被选中概率快速递减的规律,基于RD代价进行帧内预测块划分和帧内预测模式选择。实验结果表明,该算法在相同编码质量条件下可减少59%的HM10.0帧内预测模块复杂度,相应比特率的增加幅度小于1.34%。     

基于GAMLSS-GLO及随机森林算法的土石坝渗流监测模型

根据土石坝渗流原型观测,厘清水位、降水等因素的影响,模拟大坝渗流的真实状态,合理评估其渗流监测结果,是土石坝安全监控亟待解决的关键问题。基于此,根据数理统计原理,采用随机森林算法构建无降水条件下渗流量与上下游水位的回归模型;考虑渗流量受前期累积降水的综合影响,引入广义可加模型（GAMLSS-GLO）,模拟降水影响下土石坝渗流监测值的波动区间,并将其与渗流-水位回归模型叠加,预测土石坝渗流监测的可靠区间;最后,将该方法应用于糯扎渡心墙堆石坝的渗流监测。结果表明：所提模型方法对渗流的水位、降水响应表现出良好的适用性,显著提高了渗流模拟预测质量。同时求解了渗流量置信区间,有利于土石坝的运行工况判断及安全监控。     

基于机器学习的肺部CT图像非刚性配准误差预测方法

配准误差评估通常由人工完成,耗时费力;常用的Dice测度只关注组织边缘的配准误差,难以评估组织内部配准结果。针对以上问题,提出一种基于机器学习的肺部CT图像非刚性配准误差预测方法(PREML)。该方法首先构建形变场统计特征、形变场物理保真度特征和图像相似性特征三类特征,然后通过池化方法扩充特征数量,最后使用随机森林回归方法预测非刚性配准误差,并且使用自适应随机扰动方法模拟肺部配准误差空间分布,进一步提升形变场统计特征的配准误差表征能力。在三个肺部CT图像数据集上进行训练与测试,其配准误差预测结果与金标准之间的平均绝对差异为1.245±2.500 mm,预测性能优于基线方法。结果表明,PREML方法具有预测精度高、鲁棒性强的特点,可提升配准算法在临床应用的有效性和安全性。     

2000-2019年新疆水-能源-粮食系统耦合协调发展研究

水、能源和粮食是区域可持续发展的关键要素。以新疆维吾尔自治区为研究区,构建水-能源-粮食（W-E-F）系统综合评价指标体系,利用耦合协调度模型和灰色GM（1,1）模型,分析新疆2000-2019年W-E-F系统耦合协调发展水平,并对2020-2031年的耦合协调度进行预测。结果表明：2000-2019年新疆W-E-F系统的综合评价指数由0.39增大至0.61,整体上呈现上升趋势;耦合协调度从0.59增大至0.77,3个子系统之间存在着密切的联系,目前整体表现为中级协调发展水平,并预测在2025年将过渡为高级协调发展阶段。水资源子系统仍是制约W-E-F系统协调发展的关键,农业用水占比过大,导致新疆总体上结构型缺水严重。未来应积极调整用水结构,提高水资源利用效率,优化粮食消费结构,以提升系统内部的协调性。     

硝胺推进剂和双基系推进剂燃速预估模型进展

综述硝胺推进剂和双基系推进剂稳态燃速预估模型的最新进展，对下列6个模型的计算原理。和优缺点进行了介绍和评价：HMX／AP（1：1）燃速计算模型、“双区”稳态燃烧模型、复合多火焰模型、神’经网络模型、自由基裂解模型和双基系推进剂半经验预估模型。     

多粒度时序特征在离网预测中的应用

电信运营商为了发现可能离网的客户,针对不同的场景研究开发了多种离网预测模型。目前的离网预测模型首先选择一种时间粒度抽取特征,之后使用机器学习算法对抽取的数据建模。这类方法只考虑了模型对分类性能的影响,没有充分考虑数据的作用。针对上述问题,提出一种使用多种时间粒度抽取特征的方法,并尝试在模型训练的不同阶段对不同粒度的特征进行融合。实验结果表明,使用多种粒度抽取特征训练出来的模型性能会明显优于使用单一粒度抽取特征的模型。     

有向动态网络中基于模体演化的链路预测方法

以往传统的链路预测方法大多数针对无向网络,而实际上大多数社交网络是有向的,并且没有考虑网络中同一节点对之间的重复边以及微观演化信息,因此不能较好地解决有向动态网络中的链路预测问题。针对有向网络,将节点对之间的重复边信息转换为该节点对之间连边的权值;接着采用了基于三元组模体的演化模型,对滑动窗口中相邻时间片的模体转换概率进行统计后,采用指数加权滑动平均法对其进行时序分析得到不同模体转换概率的预测矩阵,进而使用该矩阵对网络中的链边进行预测。这不仅充分利用了网络微观演化信息,而且解决了动态网络中重复边的问题。最后对实验结果进行分析发现,在高全局聚类系数高平均度的网络中AUC相比Triad Transition Matrix方法提高了近0.01,而相比Common Neighbor方法提高更多。因此,所提方法能够较好地应用网络微观演化信息进行链路预测。