基于集成学习的肿瘤药物敏感性预测研究

肿瘤药物敏感性预测对个性化精准用药具有重要意义。本文基于GDSC数据库通过Boosting集成学习构建了面向RNA-seq基因表达和癌症药物敏感性数据的预测模型。先将183种药物集分别做归一化处理和基因特征降维,接着用AdaBoost集成SVM的方法建模,并采用十折交叉验证。实验结果表明构建的预测模型具有较高的预测精度,13种药物的AUC大于0.95,108种大于0.9,174种大于0.8。对比验证实验中,AdaBoost+SVM相比单学习器模型在整体药物集的综合评价指标中约提高4%,与其他集成模型相比提高2%。同时本文探讨了药物特异性,通过特征选择和富集分析对药物作用通路进行验证,从生物学角度提供了模型可解释性,证明其应用于临床用药指导的价值。     

基于SFS-SVM的乳腺癌预测模型的构建

目的:构建基于序列前向选择算法(SFS)与支持向量机算法(SVM)分类器融合的乳腺癌预测模型,提高计算机辅助诊断技术对乳腺癌细针穿刺细胞病理的准确率。方法:对456组乳腺肿瘤病理数据作为训练集,利用SFS-SVM算法对30个特征进行筛选,得到最优的特征组合,再用112组乳腺肿瘤病理数据作为测试集验证,构建乳腺癌预测模型。该模型的预测精度通过5折交叉验证进行评价。评价指标包括:受试者工作特性曲线(ROC)下面积(AUC)、准确率(ACC)、敏感度和特异度。结果:构建了基于SFS-SVM的乳腺癌预测模型,该模型(AUC为98.39%,ACC为97.35%)相对于单独SVM算法(AUC为97.00%, ACC为92.42%)有一定的提高。结论:基于SFS特征选择的SVM分类器乳腺癌预测模型能较好地对乳腺癌进行辅助诊断。     

嵌入欠采样技术的支持向量机集成分类算法的MicroRNA靶标预测

针对MicroRNA(miRNA)靶基因样本数据不平衡导致阳性样本预测准确率低和整体分类效果不佳的问题,提出一种基于欠采样技术的集成学习算法——支持向量机(SVM)-嵌入下采样和权重平滑(IUSW)集成学习算法。算法采用SVM作为基学习算法,以AdaBoost为集成框架,迭代过程中嵌入基于聚类的欠采样以降低阴阳样本数据分布不平衡程度,同时在自适应样本权重调整过程中,以样本权重平滑机制剔除阴性样本中的异常点以避免过学习,最终以带权重的投票机制组合多个弱分类器预测结果作为miRNA集成分类器的预测结果。实验表明,在不平衡数据集上SVM-IUSW算法和其他算法相比,不但有效提高了阳性靶标的预测准确率和整体分类效果,还增强了miRNA靶标分类器的泛化能力。     

基于特权信息集成学习的精神分裂症单模态神经影像计算机辅助诊断

神经影像技术目前已经应用于精神分裂症的诊断。为了提升基于单模态神经影像的精神分裂症计算机辅助诊断(CAD)的性能,本文提出一种基于特权信息学习(LUPI)分类器的集成学习算法。该算法首先对单模态数据采用极限学习机-自编码器(ELM-AE)进行特征二次学习,然后通过随机映射算法将高维特征随机分成多个子空间,并进行两两组合形成源领域和目标领域数据对,用于训练多个支持向量机+(SVM+)弱分类器,最终通过集成学习获得一个强分类器,实现有效的模式分类。本算法在公开的精神分裂症神经影像数据库中进行了实验,包括结构磁共振成像和功能磁共振成像数据。结果表明该算法取得了最优的诊断结果,其在基于结构磁共振成像诊断的分类精度、敏感性和特异性分别可以达到72.12%±8.20%、73.50%±15.44%和70.93%±12.93%,而基于功能磁共振成像诊断的分类精度、敏感性和特异性分别为72.33%±8.95%、68.50%±16.58%、75.73%±16.10%。本文算法的主要创新点在于克服了传统的LUPI分类器需要额外的特权信息模态的不足,可以直接应用于单模态数据分类问题,而且还提升了分类性能,因此具有较为广泛的应用前景。     

基于多粒度级联森林的骨质疏松性骨折预测研究

目的骨质疏松性骨折(osteoporotic fracture,OF)的预测对于骨折防范具有重要的临床指导意义。针对传统logistic回归预测模型存在的精度不高和未考虑遗传因子问题,本文引入多粒度级联森林(multi-grained cascade forest,gcForest)并结合遗传因子来预测OF。方法首先基于 t 分布邻域嵌入( t -distributed stochastic neighbor embedding, t -SNE)算法对OF关联基因位点进行非线性降维,降维后的基因位点与临床因素构成特征组。然后构建gcForest模型对OF进行预测。最后通过10次十折分层交叉验证与logistic、梯度提升决策树、随机森林进行对比。结果基于gcForest的模型分类精度为0.892 7,AUC值为0.92±0.05,泛化性能最优。结论在考虑遗传因素的条件下,gcForest分类效果优于其他模型,验证了本文方法的高效性和实用性。     

基于自步学习的多经验核映射集成分类器在乳腺癌超声计算机辅助诊断上的应用

特征表达和分类器的性能是决定计算机辅助诊断(CAD)系统性能的重要因素。为了提升基于超声成像的乳腺癌CAD系统的性能,本文提出了一种基于自步学习(SPL)的多经验核映射(MEKM)排他性正则化机(ERM)集成分类器算法,能同时提升特征表达和分类器模型的性能。该算法首先通过MEKM映射得到多组特征,以增强特征表达能力,并嵌入到ERM作为多个支持向量机的核变换;然后采用SPL策略自适应地选择样本,由易到难地逐步训练ERM集成分类器模型,从而提升分类器的性能。该算法分别在乳腺癌B型超声数据库和弹性超声数据库上进行了验证,结果显示B型超声的分类准确率、敏感度和特异性分别为(86.36±6.45)%、(88.15±7.12)%和(84.52±9.38)%,而弹性超声的分类准确率、敏感度和特异性分别为(85.97±3.75)%、(85.93±6.09)%和(86.03±5.88)%。实验结果表明,本文所提出算法能有效提升乳腺超声CAD的性能,具有投入实用的潜能。     

基于机器学习的骨质疏松性骨折预测研究

复杂疾病的预测是遗传学研究的一个重要课题。本文引入机器学习的方法,将临床变量与遗传变量作为特征,对骨质疏松性骨折进行预测研究。对临床表型和遗传变异数据进行特征选择后分别使用Logistic回归分析法、XGBoost算法对临床因子特征变量、临床因子+遗传因子特征变量进行预测;最后,使用十折交叉验证法,对预测结果进行验证。实验结果表明,相较单独使用临床因子进行预测,加入遗传因子变量,XGBoost、Logistic方法的预测准确率均得到提高;另外,XGBoost方法较Logistic回归模型预测效果更好。     

骨质疏松性压缩性骨折经皮椎体成形术治疗后非手术椎体再骨折发生情况及影响因素分析

目的探讨骨质疏松性压缩性骨折经皮椎体成形术治疗后非手术椎体再骨折发生情况及影响因素分析。方法选取我院治疗的138例骨质疏松压缩性骨折病例进行回顾性分析,据术后随访结果,将患者分为是否发生非手术椎体骨折分为非骨折组71例、骨折组67例。对可能会导致经皮椎体成形术治疗后非手术椎体再骨折情况的相关资料进行单因素比较,对造成影响的单因素进行多因素回归分析。结果 138例骨质疏松性压缩性骨折患者经皮椎体成形术治疗后有67例患者发生非手术椎体再骨折,发生率占据总人数的48.55%;经多因素分析显示,骨密度、骨折椎体部位(胸腰段)均是影响骨质疏松性压缩性骨折患者经皮椎体成形术治疗后发生非手术椎体骨折的独立危险因素(P0.001)。结论骨质疏松性压缩性骨折患者经皮椎体成形术治疗后发生非手术椎体再骨折的机率高,骨密度、骨折椎体部位是影响骨质疏松性压缩性骨折患者经皮椎体成形术治疗后发生非手术椎体骨折的危险因素。     

基于深度神经网络的冠脉造影图像的血管狭窄自动定位及分类预测

提出一套基于深度神经网络与监督学习的算法,用于对冠状动脉图像中的血管狭窄特征进行自动检测和分类。主要利用冠脉造影定量分析(QCA)作为标签进行监督学习,将冠脉狭窄的严重程度分为正常(<25%狭窄分数)、狭窄(>25%狭窄)类别,并实现图像中的狭窄定位检测。利用inception模型作为基础分类器,对图像级狭窄进行初步分类;随后结合多层次池化结构,对多视角造影图像进行联合预测,以获取左动脉/右动脉/患者级狭窄预测。在分类器的基础上进一步提取特征,分别利用监督学习/非监督学习模型,实现图像中的狭窄定位。在235例临床研究共计13 744张图像上,用所述方法进行训练及交叉验证。结果表明,在图像级狭窄分类上,该算法可以达到85%的准确率和0.91的AUC分数;在多视图联合预测实验中,针对左/右/患者级的狭窄进行分类预测,分别达到0.94/0.90/0.96的灵敏度与0.87/0.88/0.86的AUC分数。在狭窄定位实验中,针对左/右动脉狭窄检测的灵敏度分别为0.70/0.68;在512像素×512像素的图像中,均方误差分别为37.6/39.3像素。实验证明,该算法可实现从图像到病人的辅助诊断预测潜力,具有较高的精确度;不仅能提供冠脉造影过程中的初步筛选能力,而且为更精确和自动化的计算机辅助诊断奠定基础。     

基于判别混合高斯模型的信息积累方法及在脑机接口中的应用

设计有效的学习算法快速准确地对脑电信号(eelectroencephalogram,EEG)进行连续预测是脑机接口(brain-computer interface,BCI)研究的关键之一.本文提出了一种新颖的基于判别混合高斯模型(discriminative gaussian mixture model,DGMM)的信息积累方法.该方法通过区分度权值对分类器在各时段的输出进行积累,从而达到提高脑电信号分类精度的作用.在两个运动想象数据集上的实验结果表明该方法能够提高BCI系统的性能,具有较好的实用性.     

IntelliHealth: A medical decision support application using a novel weighted multi-layer classifier ensemble framework

Accuracy plays a vital role in the medical field as it concerns with the life of an individual. Extensive research has been conducted on disease classification and prediction using machine learning techniques. However, there is no agreement on which classifier produces the best results. A specific classifier may be better than others for a specific dataset, but another classifier could perform better for some other dataset. Ensemble of classifiers has been proved to be an effective way to improve classification accuracy. In this research we present an ensemble framework with multi-layer classification using enhanced bagging and optimized weighting. The proposed model called “HM-BagMoov” overcomes the limitations of conventional performance bottlenecks by utilizing an ensemble of seven heterogeneous classifiers. The framework is evaluated on five different heart disease datasets, four breast cancer datasets, two diabetes datasets, two liver disease datasets and one hepatitis dataset obtained from public repositories. The analysis of the results show that ensemble framework achieved the highest accuracy, sensitivity and F-Measure when compared with individual classifiers for all the diseases. In addition to this, the ensemble framework also achieved the highest accuracy when compared with the state of the art techniques. An application named “IntelliHealth” is also developed based on proposed model that may be used by hospitals/doctors for diagnostic advice.     

Comparison of logistic and Bayesian classifiers for evaluating the risk of femoral neck fracture in osteoporotic patients

Femoral neck fracture prediction is an important social and economic issue. The research compares two statistical methods for the classification of patients at risk for femoral neck fracture: multiple logistic regression and Bayes linear classifier. The two approaches are evaluated for their ability to separate femoral neck fractured patients from osteoporotic controls. In total, 272 Italian women are studied. Densitometric and geometric measurements are obtained from the proximal femur by dual energy X-ray absorptiometry. The performances of the two methods are evaluated by accuracy in the classification and receiver operating characteristic curves. The Bayes classifier achieves an accuracy approximately 1% higher than that of the multiple logistic regression. However, the performances of the two methods, evaluated by the area under the curves, are not statistically different. The study demonstrates that the Bayes linear classifier can be a valid alternative to multiple logistic regression in the classification of osteoporotic patients.     

糖尿病视网膜病变的风险揭示与关键因素分析

目的:通过构建组合模型对糖尿病并发视网膜病变（DR）的患病风险进行预测,为DR的预防和诊断提供参考。方法:基于3 000例糖尿病患者的生化检测数据,运用互信息作为评价标准筛选出与DR有关的特征因素,将其作为入模变量构建5种常见的模型,以准确率、精确率、召回率和AUC作为评价标准筛选出预测能力较优的3种模型,并运用Stacking方法构建组合模型。结果:通过互信息筛选出39个特征因素,发现随机森林模型、SVM模型以及Logistic回归模型这3种模型表现较优;构建的3种组合模型中,发现以SVM、Logistic为初级分类器,随机森林为次级分类器的组合模型预测能力最好,其AUC高达0.877。结论:组合模型相比单一模型具有更好的DR风险预测能力,更有助于DR的临床诊断。     

骨水泥型股骨柄人工股骨头置换治疗老年骨质疏松股骨颈骨折的可行性：随机对照临床试验方案

BACKGROUND: Osteogenesis is a common problem after surgery for femoral neck fracture in elderly patients. Internal fixation for the treatment of femoral neck fracture should be performed to optimize bone remodeling and strengthen fractured bone trabeculae, with the aim of achieving strong fixation from the perspective of biomechanics. Percutaneous internal fixation with cannulated compression screws has become a preferred treatment method of osteoporotic femoral neck fracture in elderly patients, but the insufficient holding power of the screws used in the femoral neck does not lead to strong fixation. An alternative, joint prosthesis, is recommended for the repair of femoral neck fracture in elderly patients. However, its long-term therapeutic effects in the treatment of osteoporotic fracture of proximal femoral neck remain poorly understood in patients with avascular necrosis of the femoral head.     

Population-based Analysis for Risk of Suicide Death in Elderly Patients after Osteoporotic Fracture: a Nested Case-Control Study

BackgroundThe purpose of study was to investigate the incidence rate of suicide in elderly patients with osteoporotic fractures in a nested case-control model and to analyze the change in the risk of suicide death over time after each osteoporotic fracture.MethodsWe used the National Health Insurance Service-Senior cohort of South Korea. Suicide cases and controls were matched based on sex and age at the index date. Controls were randomly selected at a 1:5 ratio from the set of individuals who were at risk of becoming a case at the time when suicide cases were selected. Conditional logistic regression analysis was performed to evaluate the association between each type of osteoporotic fracture and the risk of suicide death.ResultsThree thousand seventy suicide cases and 15,350 controls were identified. Patients with hip fracture showed an increased risk of suicide death within 1 year of fracture (adjusted odds ratio [aOR] = 2.64; 95% confidence interval [CI], 1.57–4.46; P < 0.001) compared to controls. However, the increased risk of suicide death in patients with hip fracture lasted up to 2 years (aOR = 1.59; 95% CI, 1.04–2.41; P = 0.031). Spine fracture increased the risk of suicide deaths for all observation periods. There was no evidence that humerus fracture increased the risk of suicide death during the observational period. Radius fracture increased only the risk of suicide death within 2 years of fracture (aOR = 1.43; 95% CI, 0.74–2.77; P = 0.282).ConclusionThere were noticeable differences in both degree and duration of increased suicide risks depending on the type of osteoporotic fracture. Mental stress and suicide risk in elderly patients after osteoporotic fracture should be assessed differently depending on the types of fracture.     

老年骨质疏松性股骨转子间骨折植入物置入固定并综合治疗的特征分析

背景：对股骨转子间骨折患者的植入物治疗的特点需要进行深入的分析。 目的：分析老年骨质疏松性股骨转子间骨折植入物置入固定复合综合治疗方法的特点。 方法：回顾性分析328例老年股骨转子间骨折病例,统计骨质疏松性股骨转子间骨折受伤原因、合并骨折及内科合并症情况,并对治疗方式和结局进行分析。 结果与结论：所有股骨转子间骨折患者中,年龄≥65岁的骨质疏松性转子间骨折患者比例达72.6%,其中9.5%合并其他部位骨折,46.3%具内科合并症。84.5%获得植入物置换治疗,优良率分别为：动态髋螺钉内固定85.6%、股骨近端抗旋髓内钉或γ钉固定90.9%、全髋或人工股骨头置换86.7%。可见老年骨质疏松性股骨转子间骨折患者发病率较高,治疗以髓内固定的效果最佳,治疗时应综合考虑其年龄、骨折类型、骨质疏松和内科合并症严重程度。     

Decision tool for the early diagnosis of trauma patient hypovolemia

We present a classifier for use as a decision assist tool to identify a hypovolemic state in trauma patients during helicopter transport to a hospital, when reliable acquisition of vital-sign data may be difficult. The decision tool uses basic vital-sign variables as input into linear classifiers, which are then combined into an ensemble classifier. The classifier identifies hypovolemic patients with an area under a receiver operating characteristic curve (AUC) of 0.76 (standard deviation 0.05, for 100 randomly-reselected patient subsets). The ensemble classifier is robust; classification performance degrades only slowly as variables are dropped, and the ensemble structure does not require identification of a set of variables for use as best-feature inputs into the classifier. The ensemble classifier consistently outperforms best-features-based linear classifiers (the classification AUC is greater, and the standard deviation is smaller, p < 0.05). The simple computational requirements of ensemble classifiers will permit them to function in small fieldable devices for continuous monitoring of trauma patients.     

Ensemble framework based real-time respiratory motion prediction for adaptive radiotherapy applications

Successful treatment of tumors with motion-adaptive radiotherapy requires accurate prediction of respiratory motion, ideally with a prediction horizon larger than the latency in radiotherapy system. Accurate prediction of respiratory motion is however a non-trivial task due to the presence of irregularities and intra-trace variabilities, such as baseline drift and temporal changes in fundamental frequency pattern. In this paper, to enhance the accuracy of the respiratory motion prediction, we propose a stacked regression ensemble framework that integrates heterogeneous respiratory motion prediction algorithms. We further address two crucial issues for developing a successful ensemble framework: (1) selection of appropriate prediction methods to ensemble (level-0 methods) among the best existing prediction methods; and (2) finding a suitable generalization approach that can successfully exploit the relative advantages of the chosen level-0 methods. The efficacy of the developed ensemble framework is assessed with real respiratory motion traces acquired from 31 patients undergoing treatment. Results show that the developed ensemble framework improves the prediction performance significantly compared to the best existing methods.