Background Leprosy is an infectious disease caused by Mycobacterium leprae and remains a source of preventable disability if left undetected.Case detection delay is an important epidemiological indicator for progress ...Background Leprosy is an infectious disease caused by Mycobacterium leprae and remains a source of preventable disability if left undetected.Case detection delay is an important epidemiological indicator for progress in interrupting transmission and preventing disability in a community.However,no standard method exists to effectively analyse and interpret this type of data.In this study,we aim to evaluate the characteristics of leprosy case detection delay data and select an appropriate model for the variability of detection delays based on the best fitting distribution type.Methods Two sets of leprosy case detection delay data were evaluated:a cohort of 181 patients from the post exposure prophylaxis for leprosy(PEP4LEP)study in high endemic districts of Ethiopia,Mozambique,and Tanzania;and self-reported delays from 87 individuals in 8 low endemic countries collected as part of a systematic literature review.Bayesian models were fit to each dataset to assess which probability distribution(log-normal,gamma or Weibull)best describes variation in observed case detection delays using leave-one-out cross-validation,and to estimate the effects of individual factors.Results For both datasets,detection delays were best described with a log-normal distribution combined with covariates age,sex and leprosy subtype[expected log predictive density(ELPD)for the joint model:-1123.9].Patients with multibacillary(MB)leprosy experienced longer delays compared to paucibacillary(PB)leprosy,with a relative difference of 1.57[95%Bayesian credible interval(BCI):1.14-2.15].Those in the PEP4LEP cohort had 1.51(95%BCI:1.08-2.13)times longer case detection delay compared to the self-reported patient delays in the systematic review.Conclusions The log-normal model presented here could be used to compare leprosy case detection delay datasets,including PEP4LEP where the primary outcome measure is reduction in case detection delay.We recommend the application of this modelling approach to test different probability distributions and covariate effects in studies with similar outcomes in the field of leprosy and other skin-NTDs.展开更多
针对聚类中的多视角和可解释的问题,提出多视角生成模型的可解释性聚类算法(interpretable clustering with multi-view generative model,ICMG).ICMG能够产生多个视角的聚类划分,并通过视角的语义信息对聚类结果进行定性和定量地解释....针对聚类中的多视角和可解释的问题,提出多视角生成模型的可解释性聚类算法(interpretable clustering with multi-view generative model,ICMG).ICMG能够产生多个视角的聚类划分,并通过视角的语义信息对聚类结果进行定性和定量地解释.首先,构建一种多视角生成模型(multi-view generative model,MGM),该模型使用贝叶斯程序学习(Bayesian program learning,BPL)和嵌入多视角因素的贝叶斯案例模型(multi-view Bayesian case model,MBCM)生成多个视角.其次,基于视角的匹配度进行聚类得到多种聚类方案.最后使用视角的原型和子空间所附带的语义信息定性和定量地解释聚类结果.实验结果表明:ICMG能够得到多种可解释的聚类结果,相比于传统多视角聚类算法具有较明显的优势.展开更多
基金the European Union awarded to NLR/LM(grant number RIA2017NIM-1839-PEP-4LEP),and the Leprosy Research Initiative(LRIwww.lepro syres earch.org)awarded to NLR/LM(grant number 707.19.58.).
文摘Background Leprosy is an infectious disease caused by Mycobacterium leprae and remains a source of preventable disability if left undetected.Case detection delay is an important epidemiological indicator for progress in interrupting transmission and preventing disability in a community.However,no standard method exists to effectively analyse and interpret this type of data.In this study,we aim to evaluate the characteristics of leprosy case detection delay data and select an appropriate model for the variability of detection delays based on the best fitting distribution type.Methods Two sets of leprosy case detection delay data were evaluated:a cohort of 181 patients from the post exposure prophylaxis for leprosy(PEP4LEP)study in high endemic districts of Ethiopia,Mozambique,and Tanzania;and self-reported delays from 87 individuals in 8 low endemic countries collected as part of a systematic literature review.Bayesian models were fit to each dataset to assess which probability distribution(log-normal,gamma or Weibull)best describes variation in observed case detection delays using leave-one-out cross-validation,and to estimate the effects of individual factors.Results For both datasets,detection delays were best described with a log-normal distribution combined with covariates age,sex and leprosy subtype[expected log predictive density(ELPD)for the joint model:-1123.9].Patients with multibacillary(MB)leprosy experienced longer delays compared to paucibacillary(PB)leprosy,with a relative difference of 1.57[95%Bayesian credible interval(BCI):1.14-2.15].Those in the PEP4LEP cohort had 1.51(95%BCI:1.08-2.13)times longer case detection delay compared to the self-reported patient delays in the systematic review.Conclusions The log-normal model presented here could be used to compare leprosy case detection delay datasets,including PEP4LEP where the primary outcome measure is reduction in case detection delay.We recommend the application of this modelling approach to test different probability distributions and covariate effects in studies with similar outcomes in the field of leprosy and other skin-NTDs.
文摘针对聚类中的多视角和可解释的问题,提出多视角生成模型的可解释性聚类算法(interpretable clustering with multi-view generative model,ICMG).ICMG能够产生多个视角的聚类划分,并通过视角的语义信息对聚类结果进行定性和定量地解释.首先,构建一种多视角生成模型(multi-view generative model,MGM),该模型使用贝叶斯程序学习(Bayesian program learning,BPL)和嵌入多视角因素的贝叶斯案例模型(multi-view Bayesian case model,MBCM)生成多个视角.其次,基于视角的匹配度进行聚类得到多种聚类方案.最后使用视角的原型和子空间所附带的语义信息定性和定量地解释聚类结果.实验结果表明:ICMG能够得到多种可解释的聚类结果,相比于传统多视角聚类算法具有较明显的优势.
