The performance of medical image classification has been enhanced by deep convolutional neural networks(CNNs),which are typically trained with cross-entropy(CE)loss.However,when the label presents an intrinsic ordinal...The performance of medical image classification has been enhanced by deep convolutional neural networks(CNNs),which are typically trained with cross-entropy(CE)loss.However,when the label presents an intrinsic ordinal property in nature,e.g.,the development from benign to malignant tumor,CE loss cannot take into account such ordinal information to allow for better generalization.To improve model generalization with ordinal information,we propose a novel meta ordinal regression forest(MORF)method for medical image classification with ordinal labels,which learns the ordinal relationship through the combination of convolutional neural network and differential forest in a meta-learning framework.The merits of the proposed MORF come from the following two components:A tree-wise weighting net(TWW-Net)and a grouped feature selection(GFS)module.First,the TWW-Net assigns each tree in the forest with a specific weight that is mapped from the classification loss of the corresponding tree.Hence,all the trees possess varying weights,which is helpful for alleviating the tree-wise prediction variance.Second,the GFS module enables a dynamic forest rather than a fixed one that was previously used,allowing for random feature perturbation.During training,we alternatively optimize the parameters of the CNN backbone and TWW-Net in the meta-learning framework through calculating the Hessian matrix.Experimental results on two medical image classification datasets with ordinal labels,i.e.,LIDC-IDRI and Breast Ultrasound datasets,demonstrate the superior performances of our MORF method over existing state-of-the-art methods.展开更多
This letter presents a new discriminative model for Information Retrieval (IR), referred to as Ordinal Regression Model (ORM). ORM is different from most existing models in that it views IR as ordinal regression probl...This letter presents a new discriminative model for Information Retrieval (IR), referred to as Ordinal Regression Model (ORM). ORM is different from most existing models in that it views IR as ordinal regression problem (i.e. ranking problem) instead of binary classification. It is noted that the task of IR is to rank documents according to the user information needed, so IR can be viewed as ordinal regression problem. Two parameter learning algorithms for ORM are presented. One is a perceptron-based algorithm. The other is the ranking Support Vector Machine (SVM). The effec- tiveness of the proposed approach has been evaluated on the task of ad hoc retrieval using three English Text REtrieval Conference (TREC) sets and two Chinese TREC sets. Results show that ORM sig- nificantly outperforms the state-of-the-art language model approaches and OKAPI system in all test sets; and it is more appropriate to view IR as ordinal regression other than binary classification.展开更多
The adjacent-categories, continuation-ratio and proportional odds logit-link regression models provide useful extensions of the multinomial logistic model to ordinal response data. We propose fitting these models with...The adjacent-categories, continuation-ratio and proportional odds logit-link regression models provide useful extensions of the multinomial logistic model to ordinal response data. We propose fitting these models with a logarithmic link to allow estimation of different forms of the risk ratio. Each of the resulting ordinal response log-link models is a constrained version of the log multinomial model, the log-link counterpart of the multinomial logistic model. These models can be estimated using software that allows the user to specify the log likelihood as the objective function to be maximized and to impose constraints on the parameter estimates. In example data with a dichotomous covariate, the unconstrained models produced valid coefficient estimates and standard errors, and the constrained models produced plausible results. Models with a single continuous covariate performed well in data simulations, with low bias and mean squared error on average and appropriate confidence interval coverage in admissible solutions. In an application to real data, practical aspects of the fitting of the models are investigated. We conclude that it is feasible to obtain adjusted estimates of the risk ratio for ordinal outcome data.展开更多
Food insecurity is a global issue,and households in a society can experience food insecurity at different levels that could range from being mildly food in-secure to severely food insecure.The severity of food insecur...Food insecurity is a global issue,and households in a society can experience food insecurity at different levels that could range from being mildly food in-secure to severely food insecure.The severity of food insecurity is an ordinal categorical variable in nature and different types of ordinal logistic regression models could be used to model such variables.The purpose of this study is to identify the socioeconomic and demographic factors associated with house-hold food insecurity in Namibia by fitting an ordinal logistic regression model using the 2015/2016 Namibia Household Income and Expenditure Survey.The proportional odds model(POM)and the partial proportional odds model(PPOM)were fitted and the performance of the two models was also com-pared.The PPOM was found to be the better model and based on the PPOM result,the study found factors such as the age of the household head,the household size,the source of income of a household,the annual income of the household,the education level attained by a household head and the geo-graphical location of a household to be significant factors associated with se-verity of household food insecurity in Namibia.展开更多
Background:Urinary schistosomiasis has been a major public health problem in Zambia for many years.However,the disease profile may vary in different locale due to the changing ecosystem that contributes to the risk of...Background:Urinary schistosomiasis has been a major public health problem in Zambia for many years.However,the disease profile may vary in different locale due to the changing ecosystem that contributes to the risk of acquiring the disease.The objective of this study was to quantify risk factors associated with the intensity of urinary schistosomiasis infection in school children in Lusaka Province,Zambia,in order to better understand local transmission.Methods:Data were obtained from 1912 school children,in 20 communities,in the districts of Luangwa and Kafue in Lusaka Province.Both individual-and community-level covariates were incorporated into an ordinal logistic regression model to predict the probability of an infection being a certain intensity in a three-category outcome response:0=no infection,1=light infection,and 2=moderate/heavy infection.Random effects were introduced to capture unobserved heterogeneity.Results:Overall,the risk of urinary schistosomiasis was strongly associated with age,altitude at which the child lived,and sex.Weak associations were observed with the normalized difference vegetation index,maximum temperature,and snail abundance.Detailed analysis indicated that the association between infection intensities and age and altitude were category-specific.Particularly,infection intensity was lower in children aged between 5 and 9 years compared to those aged 10 to 15 years(OR=0.72,95%CI=0.51-0.99).However,the age-specific risk changed at different levels of infection,such that when comparing children with light infection to those who were not infected,age was associated with a lower odds(category 1 vs category 0:OR=0.71,95%CI:0.50-0.99),yet such a relation was not significant when considering children who were moderately or heavily infected compared to those with a light or no infection(category 2 vs category 0:OR=0.96,95%CI:0.45-1.64).Overall,we observed that children living in the valley were less likely to acquire urinary schistosomiasis compared to those living in plateau areas(OR=0.48,95%CI:0.16-0.71).However,category-specific effects showed no significant association in category 1(light infection),whereas in category 2(moderate/high infection),the risk was still significantly lower for those living in the valley compared to those living in plateau areas(OR=0.18,95%CI:0.04-0.75).Conclusions:This study demonstrates the importance of understanding the dynamics and heterogeneity of infection in control efforts,and further suggests that apart from the well-researched factors of Schistosoma intensity,various other factors influence transmission.Control programmes need to take into consideration the varying infection intensities of the disease so that effective interventions can be designed.展开更多
Longitudinal data with ordinal outcomes commonly arise in clinical and social studies,where the purpose of interest is usually quantile curves rather than a simple reference range.In this paper we consider Bayesian no...Longitudinal data with ordinal outcomes commonly arise in clinical and social studies,where the purpose of interest is usually quantile curves rather than a simple reference range.In this paper we consider Bayesian nonlinear quantile regression for longitudinal ordinal data through a latent variable.An efficient Metropolis–Hastings within Gibbs algorithm was developed for model fitting.Simulation studies and a real data example are conducted to assess the performance of the proposed method.Results show that the proposed approach performs well.展开更多
基金This work was supported in part by the Natural Science Foundation of Shanghai(21ZR1403600)the National Natural Science Foundation of China(62176059)+3 种基金Shanghai Municipal Science and Technology Major Project(2018SHZDZX01)Zhang Jiang Laboratory,Shanghai Sailing Program(21YF1402800)Shanghai Municipal of Science and Technology Project(20JC1419500)Shanghai Center for Brain Science and Brain-inspired Technology.
文摘The performance of medical image classification has been enhanced by deep convolutional neural networks(CNNs),which are typically trained with cross-entropy(CE)loss.However,when the label presents an intrinsic ordinal property in nature,e.g.,the development from benign to malignant tumor,CE loss cannot take into account such ordinal information to allow for better generalization.To improve model generalization with ordinal information,we propose a novel meta ordinal regression forest(MORF)method for medical image classification with ordinal labels,which learns the ordinal relationship through the combination of convolutional neural network and differential forest in a meta-learning framework.The merits of the proposed MORF come from the following two components:A tree-wise weighting net(TWW-Net)and a grouped feature selection(GFS)module.First,the TWW-Net assigns each tree in the forest with a specific weight that is mapped from the classification loss of the corresponding tree.Hence,all the trees possess varying weights,which is helpful for alleviating the tree-wise prediction variance.Second,the GFS module enables a dynamic forest rather than a fixed one that was previously used,allowing for random feature perturbation.During training,we alternatively optimize the parameters of the CNN backbone and TWW-Net in the meta-learning framework through calculating the Hessian matrix.Experimental results on two medical image classification datasets with ordinal labels,i.e.,LIDC-IDRI and Breast Ultrasound datasets,demonstrate the superior performances of our MORF method over existing state-of-the-art methods.
基金Supported by the High Technology Research and Devel-opment Program of China (No.2006AA01Z150)the Key Project of the National Natural Science Foundation of China (No.60373101)+1 种基金the Natural Science Foundation of Heilongjiang Province (No.F2007-14)the Project of Heilongjiang Outstanding Young University Teacher (No. 1151G037).
文摘This letter presents a new discriminative model for Information Retrieval (IR), referred to as Ordinal Regression Model (ORM). ORM is different from most existing models in that it views IR as ordinal regression problem (i.e. ranking problem) instead of binary classification. It is noted that the task of IR is to rank documents according to the user information needed, so IR can be viewed as ordinal regression problem. Two parameter learning algorithms for ORM are presented. One is a perceptron-based algorithm. The other is the ranking Support Vector Machine (SVM). The effec- tiveness of the proposed approach has been evaluated on the task of ad hoc retrieval using three English Text REtrieval Conference (TREC) sets and two Chinese TREC sets. Results show that ORM sig- nificantly outperforms the state-of-the-art language model approaches and OKAPI system in all test sets; and it is more appropriate to view IR as ordinal regression other than binary classification.
文摘The adjacent-categories, continuation-ratio and proportional odds logit-link regression models provide useful extensions of the multinomial logistic model to ordinal response data. We propose fitting these models with a logarithmic link to allow estimation of different forms of the risk ratio. Each of the resulting ordinal response log-link models is a constrained version of the log multinomial model, the log-link counterpart of the multinomial logistic model. These models can be estimated using software that allows the user to specify the log likelihood as the objective function to be maximized and to impose constraints on the parameter estimates. In example data with a dichotomous covariate, the unconstrained models produced valid coefficient estimates and standard errors, and the constrained models produced plausible results. Models with a single continuous covariate performed well in data simulations, with low bias and mean squared error on average and appropriate confidence interval coverage in admissible solutions. In an application to real data, practical aspects of the fitting of the models are investigated. We conclude that it is feasible to obtain adjusted estimates of the risk ratio for ordinal outcome data.
基金The publication of this article was supported financially by the Department of Mathematics,Statistics and Actuarial Science at Namibia University of Science and Technology(NUST).
文摘Food insecurity is a global issue,and households in a society can experience food insecurity at different levels that could range from being mildly food in-secure to severely food insecure.The severity of food insecurity is an ordinal categorical variable in nature and different types of ordinal logistic regression models could be used to model such variables.The purpose of this study is to identify the socioeconomic and demographic factors associated with house-hold food insecurity in Namibia by fitting an ordinal logistic regression model using the 2015/2016 Namibia Household Income and Expenditure Survey.The proportional odds model(POM)and the partial proportional odds model(PPOM)were fitted and the performance of the two models was also com-pared.The PPOM was found to be the better model and based on the PPOM result,the study found factors such as the age of the household head,the household size,the source of income of a household,the annual income of the household,the education level attained by a household head and the geo-graphical location of a household to be significant factors associated with se-verity of household food insecurity in Namibia.
基金The first author(CS)received a travel award from the Danish Bilharziasis Laboratory,now the DBL-Centre for Health Research and Development,University of Copenhagen,DenmarkThe second author’s(LNK)efforts were partly funded by the University of Namibia.
文摘Background:Urinary schistosomiasis has been a major public health problem in Zambia for many years.However,the disease profile may vary in different locale due to the changing ecosystem that contributes to the risk of acquiring the disease.The objective of this study was to quantify risk factors associated with the intensity of urinary schistosomiasis infection in school children in Lusaka Province,Zambia,in order to better understand local transmission.Methods:Data were obtained from 1912 school children,in 20 communities,in the districts of Luangwa and Kafue in Lusaka Province.Both individual-and community-level covariates were incorporated into an ordinal logistic regression model to predict the probability of an infection being a certain intensity in a three-category outcome response:0=no infection,1=light infection,and 2=moderate/heavy infection.Random effects were introduced to capture unobserved heterogeneity.Results:Overall,the risk of urinary schistosomiasis was strongly associated with age,altitude at which the child lived,and sex.Weak associations were observed with the normalized difference vegetation index,maximum temperature,and snail abundance.Detailed analysis indicated that the association between infection intensities and age and altitude were category-specific.Particularly,infection intensity was lower in children aged between 5 and 9 years compared to those aged 10 to 15 years(OR=0.72,95%CI=0.51-0.99).However,the age-specific risk changed at different levels of infection,such that when comparing children with light infection to those who were not infected,age was associated with a lower odds(category 1 vs category 0:OR=0.71,95%CI:0.50-0.99),yet such a relation was not significant when considering children who were moderately or heavily infected compared to those with a light or no infection(category 2 vs category 0:OR=0.96,95%CI:0.45-1.64).Overall,we observed that children living in the valley were less likely to acquire urinary schistosomiasis compared to those living in plateau areas(OR=0.48,95%CI:0.16-0.71).However,category-specific effects showed no significant association in category 1(light infection),whereas in category 2(moderate/high infection),the risk was still significantly lower for those living in the valley compared to those living in plateau areas(OR=0.18,95%CI:0.04-0.75).Conclusions:This study demonstrates the importance of understanding the dynamics and heterogeneity of infection in control efforts,and further suggests that apart from the well-researched factors of Schistosoma intensity,various other factors influence transmission.Control programmes need to take into consideration the varying infection intensities of the disease so that effective interventions can be designed.
基金supported in part by the National Key Research and Development Plan(No.2016YFC0800100)National Natural Science Foundation of China Grant 11671374 and 71631006.
文摘Longitudinal data with ordinal outcomes commonly arise in clinical and social studies,where the purpose of interest is usually quantile curves rather than a simple reference range.In this paper we consider Bayesian nonlinear quantile regression for longitudinal ordinal data through a latent variable.An efficient Metropolis–Hastings within Gibbs algorithm was developed for model fitting.Simulation studies and a real data example are conducted to assess the performance of the proposed method.Results show that the proposed approach performs well.
