学科服务 更多>>

关键词： Supervised clustering   MCMC   risk adjustment   metropolis hastings   ICD-10   large feature space  

摘要： There is a well-established tradition within the statistics literature that explores different techniques for reducing the dimensionality of large feature spaces. The problem is central to machine learning and it has been largely explored under the unsupervised learning paradigm. We introduce a supervised clustering methodology that capitalizes on a Metropolis Hastings algorithm to optimize the partition structure of a large categorical feature space tailored towards minimizing the test error of a learning algorithm. This is a general methodology that can be applied to any supervised learning problem with a large categorical feature space. We show the benefits of the algorithm by applying this methodology to the problem of risk adjustment in competitive health insurance markets. We use a large claims data set that records ICD-10 codes, a large categorical feature space. We aim at improving risk adjustment by clustering diagnostic codes into risk groups suitable for health expenditure prediction. We test the performance of our methodology against common alternatives using panel data from a representative sample of twenty three million citizens in Colombian Healthcare System. Our results outperform common alternatives and suggest that it has potential to improve risk adjustment.

关键词： Social life cycle impact assessment   Multivariate analysis   Exploratory factor analysis   Structural equation modeling   Type II   S-LCA   Life expectancy at birth  

摘要： Purpose This research aims to propose the SMiLe method (Social Metric for Life Cycle), developed to identify potential impact pathways from correlations between the subcategories and their effects on the human health endpoint. We used multivariate data analysis in the S-LCA context to obtain the impact pathways and obtain the characterization model. Method The proposed method was developed in three stages: (I) area of protection (AoP) definition;(II) estimation technique selection;(III) development of the cause-effect chain and characterization model. For stage I, AoP was well-being and the endpoint category, human health, indicated by the life expectancy at birth (LEX). As estimation techniques, the exploratory factor analysis (EFA), covariance-based SEM (CB-SEM), and partial least squares SEM (PLS-SEM) were respectively used in the exploratory step (analysis of the relationships between indicators representing the subcategories), confirmatory stage (validation of impact pathways), and predictive step (obtaining the characterization model). The three estimation techniques used socioeconomic indicators from 189 countries to represent the subcategories. Results and discussion As a result of the data collection, it was possible to develop a database of 21 indicators, representing 15 subcategories related to four stakeholders, including data from various international sources, such as the World Bank and the International Labour Organization. The EFA and CB-SEM results showed that it was possible to identify and confirm that the subcategories used in this study were organized in two factors (social dimensions) related to "Economy and competitiveness" and "Access to water, sanitation and conflict prevention". Finally, through the PLS-SEM, there was a strong correlation between these social dimensions and the life expectancy at birth. Conclusions Using social indicators related to subcategories and multivariate techniques, such as EFA and SEM, it was possible to identify

关键词： fine needle aspiration cytology   histopathology   immunocytochemistry   ovarian mass   ovarian neoplasm   risk of malignancy   risk stratification  

摘要： Introduction The present study was undertaken to assess the accuracy of fine needle aspiration cytology (FNAC) and cell-block immunocytochemistry, and to estimate the risk of malignancy, using a categorical reporting system, in the diagnosis of ovarian masses. Methods This was a 5-year retrospective study of FNAs of ovarian masses. The cytological diagnoses were categorised as inadequate, non-neoplastic, benign neoplasms, indeterminate for malignancy, suspicious for malignancy and malignant neoplasms. The cytology was correlated with the corresponding histopathology to assess the diagnostic accuracy and risk of malignancy associated with each diagnostic category. Results Of a total of 66 703 FNAs performed during the study period, 580 (0.9%) were performed on ovarian masses. Of these, 40 (6.9%) were reported as non-neoplastic;76 (13.1%) as benign neoplasms;14 (2.4%) as indeterminate for malignancy, 48 (8.3%) as suspicious for malignancy, 337 (58.1%) as malignant neoplasms and 65 (11.2%) as inadequate for interpretation. Immunocytochemistry (ICC) was performed on 99 (17%) aspirates. Subsequent histopathology was available in 208 (35.8%) cases. On cyto-histopathological correlation, 183 (88%) were concordant and 25 (12%) were discordant. The overall sensitivity, specificity, positive and negative predictive values and diagnostic accuracy for diagnosing ovarian malignancy were 88.4%, 85.7%, 96.8%, 60.0% and 88% respectively. Risk of malignancy for each category was 80%, 0%, 4.5%, 66.7%, 88.5% and 98.5% respectively. Conclusions Ultrasound-guided FNAC has high specificity and diagnostic accuracy for preoperative diagnosis of ovarian malignancies and hence is a valid diagnostic procedure in certain clinical situations. Reporting using a categorical system imparts uniformity and also provides the clinicians with an associated risk of malignancy to guide further management.

关键词： biopsy   fine‐   needle aspiration   inter‐   observer variability   intra‐   observer variability   thyroid  

摘要： Aim To examine the intra- and inter-observer variability for non-benign thyroid cytological subcategories according to the Bethesda classification system after the second review. Methods Between November 2018 and May 2019, thyroid fine needle aspiration biopsies of 381 nodules were retrospectively evaluated. Among them, 74 non-benign (category III-VI) thyroid biopsies, analyzed according to the Bethesda system (pathologist 1:40 vs pathologist 2:34) by two independent pathologists, were reassessed by the same pathologists and by a cytopathologist. In this observer-blinded study, weighted Cohen's kappa was used to assess the intra-observer agreement, and Krippendorff's alpha was used to assess the inter-observer agreement. Results At the first and second evaluations of pathologists 1 and 2, the percentage agreement was 62.5% for pathologist 1 and 58.8% for pathologist 2. The intra-observer agreement was substantial (kappa = 0.705) for pathologist 1, and moderate (kappa = 0.447) for pathologist 2. In the second evaluation of pathologist 1 and 2, which was compared with the cytopathologist, the agreement percentage of pathologist 1 with the cytopathologist was 50.0%, and that of pathologist 2 was 56.8%. The inter-observer agreement was below the lowest acceptable limit for an overall agreement (alpha = 0.634) among the three raters. The inter-observer agreement was only acceptable between the cytopathologist and the second pathologist, while it was low between the other raters. In the evaluation of the non-benign nodules, the mean category score of the cytopathologist was 3.22 and lower than both pathologists (3.73 and 3.58, respectively). Conclusions The intra-observer agreement of pathologists was moderate-to-substantial in the evaluation of non-benign thyroid biopsies according to the Bethesda reporting system. However, the inter-observer agreement was below the lowest acceptable limit when the cytopathologist was taken as a reference.

关键词： Unsupervised domain adaptation   Categorical domain adaptation   Vicinal domain adaptation   Cross-domain weighting   Domain augmentation  

摘要： Unsupervised domain adaptation aims to learn a task classifier that performs well on the unlabeled target domain, by utilizing the labeled source domain. Inspiring results have been acquired by learning domain invariant deep features via domain-adversarial training. However, its parallel design of task and domain classifiers limits the ability to achieve a finer category-level domain alignment. To promote categorical domain adaptation (CatDA), based on a joint category-domain classifier, we propose novel losses of adversarial training at both domain and category levels. Since the joint classifier can be regarded as a concatenation of individual task classifiers respectively for the two domains, our design principle is to enforce consistency of category predictions between the two task classifiers. Moreover, we propose a concept of vicinal domains whose instances are produced by a convex combination of pairs of instances respectively from the two domains. Intuitively, alignment of the possibly infinite number of vicinal domains enhances that of original domains. We propose novel adversarial losses for vicinal domain adaptation (VicDA) based on CatDA, leading to Vicinal and Categorical Domain Adaptation (ViCatDA). We also propose Target Discriminative Structure Recovery (TDSR) to recover the intrinsic target discrimination damaged by adversarial feature alignment. We also analyze the principles underlying the ability of our key designs to align the joint distributions. Extensive experiments on several benchmark datasets demonstrate that we achieve the new state of the art. (c) 2021 Elsevier Ltd. All rights reserved.

关键词： Cystic fibrosis   generalized partially linear model   polynomial spline   rate proportion   weighted estimating equation  

摘要： Proportional rates models are frequently used for the analysis of recurrent event data with multiple event categories. When some of the event categories are missing, a conventional approach is to either exclude the missing data for a complete-case analysis or employ a parametric model for the missing event type. It is well known that the complete-case analysis is inconsistent when the missingness depends on covariates, and the parametric approach may incur bias when the model is misspecified. In this paper, we aim to provide a more robust approach using a rate proportion method for the imputation of missing event types. We show that the log-odds of the event type can be written as a semiparametric generalized linear model, facilitating a theoretically justified estimation framework. Comprehensive simulation studies were conducted demonstrating the improved performance of the semiparametric method over parametric procedures. Multiple types of Pseudomonas aeruginosa infections of young cystic fibrosis patients were analyzed to demonstrate the feasibility of our proposed approach.

关键词： Lung cancer   Screening   Lung-RADS   National Lung Screening Trial  

摘要： Introduction: To determine the clinical significance of category 3 (CAT3) abnormalities and the necessity of a 6-month follow-up computed tomography (CT). We also explored features associated with increased lung cancer risk. Methods: From the National Lung Screening Trial database, we identified participants with CAT3 lesions at prevalence screen. Rates of lung cancer and lung cancer-specific deaths (LSDs) were compared between those who underwent first follow-up CT before 6 months (early diagnostic group) and those who underwent annual screening (annual diagnostic group). We estimated the change in LSD if the 6-month CT was eliminated. Regression analysis was performed to identify features associ-ated with participants with CAT3 who developed lung cancer. Results: A total of 1763 CAT3s were identified (6.6% of all participants who had low-dose CT), with 108 lung cancers (6.1%) and 41 LSDs (2.3%) in a 7-year period. Rates of lung cancer (7.5% versus 3.1%) and LSD (4.0% versus 1.0%) were higher in the early diagnostic group than in the annual diagnostic group. We estimated an increase in LSD of 0.6% of all participants with CAT3 (24.4% of all LSDs) if the 6-month CT was not performed. Multivariate regression analysis found that increased age, emphysema, and a part-solid nodule greater than 5 mm were associated with par-ticipants with CAT3 who developed lung cancer. Conclusions: CAT3 lesions are uncommon, and eliminating the 6-month CT would potentially increase LSD by 0.6% of all patients with CAT3. Age, emphysema, and part-solid nodule greater than 5 mm may be useful in risk predic-tion models to determine which participants with CAT3 are more likely to develop lung cancer and suggest which pa-tients may need more intense follow-up. (c) 2021 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved.

关键词： attention   category‐   aware   embedding   gated recurrent unit   next POI recommendation  

摘要： With the continuous accumulation of users' check-in data, we can gradually capture users' behavior patterns and mine users' preferences. Based on this, the next point-of-interest (POI) recommendation has attracted considerable attention. Its main purpose is to simulate users' behavior habits of check-in behavior. Then, different types of context information are used to construct a personalized recommendation model. However, the users' check-in data are extremely sparse, which leads to low performance in personalized model training using recurrent neural network. Therefore, we propose a category-aware gated recurrent unit (GRU) model to mitigate the negative impact of sparse check-in data, capture long-range dependence between user check-ins and get better recommendation results of POI category. We combine the spatiotemporal information of check-in data and take the POI category as users' preference to train the model. Also, we develop an attention-based category-aware GRU (ATCA-GRU) model for the next POI category recommendation. The ATCA-GRU model can selectively utilize the attention mechanism to pay attention to the relevant historical check-in trajectories in the check-in sequence. We evaluate ATCA-GRU using a real-world data set, named Foursquare. The experimental results indicate that our ATCA-GRU model outperforms the existing similar methods for next POI recommendation.

关键词： Diagnosis   Ultrasonography   Hepatocellular carcinoma   Cirrhosis   Sensitivity and specificity  

摘要： Objectives Liver Imaging Reporting and Data System (LI-RADS) for hepatocellular carcinoma (HCC) diagnosis in high-risk patients is a dynamic system, which was lastly updated in 2018. We aimed to evaluate the accuracy for HCC diagnosis of LI-RADS v2018 with magnetic resonance imaging (MRI) with extracellular contrast for solitary nodules <= 20 mm detected during ultrasound (US) surveillance in cirrhotic patients, with particular interest in those observations categorized as LI-RADS 3. Methods Between November 2003 and February 2017, we included 262 consecutive cirrhotic patients with a newly US-detected solitary <= 20-mm nodule. A LI-RADS (LR) v2018 category was retrospectively assigned. The diagnostic accuracy for each LR category was described, and the main MRI findings associated with HCC diagnosis were analyzed. Results Final diagnoses were as follows: 197 HCC (75.2%), 5 cholangiocarcinoma (1.9%), 2 metastasis (0.8%), and 58 benign lesions (22.1%);0/15 (0%) LR-1, 6/26 (23.1%) LR-2, 51/74 (68.9%) LR-3, 11/12 (91.7%) LR-4, 126/127 (99.2%) LR-5, and 3/8 (37.5%) LR-M were HCC. LR-5 category displayed a sensitivity and specificity of 64% (95% CI, 56.8-70.7) and 98.5% (95% CI, 91.7-100), respectively. Considering also LR-4 as diagnostic for HCC, the sensitivity slightly increased to 69.5% (95% CI, 62.6-75.9) with minor impact on specificity (96.2%;95% CI, 89.3-99.6). Regarding LR-3 observations, 51 out of 74 were HCC, 2 were non-HCC malignancies, and 20 out of 21 LR-3 nodules > 15 mm (95.2%) were finally categorized as HCC. Conclusions The high probability of HCC in US-detected LR-3 observations (68.9%) justifies triggering an active diagnostic work-up if intended to diagnose HCC at a very early stage.