Improving Chinese electronic medical record retrieval by field weight assignment, negation detection, and re-ranking.

The technique of information retrieval has been widely used in electronic medical record (EMR) systems. It's a pity that most existing methods have not considered the structures and language features of Chinese EMRs, which affects the performance of retrieval. To improve accuracy and comprehensiveness, we propose an improved algorithm of Chinese EMR retrieval. First, the weights of fields in Chinese EMRs are assigned based on the corresponding importance in clinical applications. Second, negative relations in EMRs are detected, and the retrieval scores of negative terms are adjusted accordingly. Third, the retrieval results are re-ranked by expansion terms and time information to enhance the recall without decreasing precision. Experiment results show that the improved algorithm increases the precision and recall significantly, which shows that the algorithm takes a full account of the characteristics of Chinese EMRs and fits the needs for clinical applications.

Process mining framework with time perspective for understanding acute care: a case study of AIS in hospitals.

BACKGROUND: Acute care for critical illness requires very strict treatment timeliness. However, healthcare providers usually cannot accurately figure out the causes of low efficiency in acute care process due to the lack of effective tools. Besides, it is difficult to compare or conformance processes from different patient groups. METHODS: To solve these problems, we proposed a novel process mining framework with time perspective, which integrates four steps: standard activity construction, data extraction and filtering, iterative model discovery, and performance analysis. RESULTS: It can visualize the execution of actual clinical activities hierarchically, evaluate the timeliness and identify bottlenecks in the treatment process. We take the acute ischemic stroke as a case study, and retrospectively reviewed 420 patients' data from a large hospital. Then we discovered process models with timelines, and identified the main reasons for in-hospital delay. CONCLUSIONS: Experiment results demonstrate that the framework proposed could be a new way of drawing insights about hospitals' clinical process, to help clinical institutions increase work efficiency and improve medical service.

Using predictive process monitoring to assist thrombolytic therapy decision-making for ischemic stroke patients.

BACKGROUND: Although clinical guidelines provide the best practice for medical activities, there are some limitations in using clinical guidelines to assistant decision-making in practical application, such as long update cycle and low compliance of doctors with the guidelines. Driven by data of actual cases, process mining technology provides the possibility to remedy these shortcomings of clinical guidelines. METHODS: We propose a clinical decision support method using predictive process monitoring, which could be complementary with clinical guidelines, to assist medical staff with thrombolytic therapy decision-making for stroke patients. Firstly, we construct a labeled data set of 1191 cases to show whether each case actually need thrombolytic therapy, and whether it conform to the clinical guidelines. After prefix extraction and filtering the control flow of completed cases, the sequences with data flow are encoded, and corresponding prediction models are trained. RESULTS: Compared with the labeled results, the average accuracy of our prediction models for intravenous thrombolysis and arterial thrombolysis on the test set are 0.96 and 0.91, and AUC are 0.93 and 0.85 respectively. Compared with the recommendation of clinical guidelines, the accuracy, recall and AUC of our predictive models are higher. CONCLUSIONS: The performance and feasibility of this method are verified by taking thrombolytic decision-making of patients with ischemic stroke as an example. When the clinical guidelines are not applicable, doctors could be provided with assistant decision-making by referring to similar historical cases using predictive process monitoring.

Modeling clinical activities based on multi-perspective declarative process mining with openEHR's characteristic.

BACKGROUND: It is significant to model clinical activities for process mining, which assists in improving medical service quality. However, current process mining studies in healthcare pay more attention to the control flow of events, while the data properties and the time perspective are generally ignored. Moreover, classifying event attributes from the view of computers usually are difficult for medical experts. There are also problems of model sharing and reusing after it is generated. METHODS: In this paper, we presented a constraint-based method using multi-perspective declarative process mining, supporting healthcare personnel to model clinical processes by themselves. Inspired by openEHR, we classified event attributes into seven types, and each relationship between these types is represented in a Constrained Relationship Matrix. Finally, a conformance checking algorithm is designed. RESULTS: The method was verified in a retrospective observational case study, which consists of Electronic Medical Record (EMR) of 358 patients from a large general hospital in China. We take the ischemic stroke treatment process as an example to check compliance with clinical guidelines. Conformance checking results are analyzed and confirmed by medical experts. CONCLUSIONS: This representation approach was applicable with the characteristic of easily understandable and expandable for modeling clinical activities, supporting to share the models created across different medical facilities.

Discover high-risk factor combinations using Bayesian network from cohort data of National Stoke Screening in China.

BACKGROUND: In recent years, the increasing incidence and prevalence of stroke has brought a heavy economic burden on families and society in China. The Ministry of Health of the Peoples' Republic of China initiated the national stroke screening and intervention program in 2011 for stroke prevention and control. In the screening, only those who have been classified to "potential high-risk" group in preliminary screening need further examination and physician confirmation to determine the risk level of stroke in rescreening. However, at the beginning of the program, the "potential high-risk" classification method in the preliminary screening are determined by experts based on their experience. The primary aim of this study is to study the causality of stroke and risk factors in middle-aged population using the cohort data, and to explore whether the stroke screening and intervention program should include more precise "potential high-risk" evaluation criteria for this age group in preliminary screening. METHOD: We use the cohort data of screening between 2013 and 2017 in this study. After data cleaning, the cohort consists of 48,007 people aged from 40 to 59 who are free of stroke at baseline. We use Bayesian networks to develop models. RESULT: The results show that the stroke incidence in middle-aged population with certain two risk factors is higher than some of that with three factors, which is in keeping with our previous study results. We can take the ratio of the stroke incidence with combinations of risk factors and incidence without any of the risk factors as a variable threshold. By adjusting the threshold, we can get precise stroke preliminary screening criteria to achieve a balance between economy and efficiency. CONCLUSION: We find that the criteria used in preliminary screening are not reasonable enough. There is a need for national stroke screening and intervention program to further include some more important risk factors or combinations of two risk factors as classification criteria in the preliminary screening. The results of the study can directly guide stroke screening program in China to make the screening more accurate and efficient.

Automatic Articular Cartilage Segmentation Based on Pattern Recognition from Knee MRI Images.

An automatic method for cartilage segmentation using knee MRI images is described. Three binary classifiers with integral and partial pixel features are built using the Bayesian theorem to segment the femoral cartilage, tibial cartilage and patellar cartilage separately. First, an iterative procedure based on the feedback of the number of strong edges is designed to obtain an appropriate threshold for the Canny operator and to extract the bone-cartilage interface from MRI images. Second, the different edges are identified based on certain features, which allow for different cartilage to be distinguished synchronously. The cartilage is segmented preliminarily with minimum error Bayesian classifiers that have been previously trained. According to the cartilage edge and its anatomic location, the speed of segmentation is improved. Finally, morphological operations are used to improve the primary segmentation results. The cartilage edge is smooth in the automatic segmentation results and shows good consistency with manual segmentation results. The mean Dice similarity coefficient is 0.761.

Study on industrial carbon emissions in China based on GDIM decomposition method and two decoupling effects.

The driving factors of China's industrial carbon emissions are decomposed by generalized Divisia index method (GDIM), so as to study the reasons for the change of China's industrial carbon emissions. The decoupling effect of China's industrial carbon emissions and economic growth is examined by speed decoupling and quantity decoupling. The speed decoupling is calculated by Tapio decoupling elasticity and emission reduction effort function, and the quantity decoupling is measured by environmental Kuznets curve (EKC). The results show that the positive driving factors are output size effect > industrial energy consumption effect > population size effect, and the negative driving factors are investment carbon emission effect > output carbon intensity effect > per capita output effect > economic efficiency effect > energy intensity effect. The elasticity of emission reduction is basically greater than that of energy conservation, indicating that there is still abundant room for efforts in emission reduction. The overall decoupling effect of carbon emissions is undecoupling-strong decoupling-undecoupling. Quadratic EKC shape is "U" shape, and the inflection point is 11.0987; the shape of cubic EKC is "N," and the inflection points are - 0.0137 and 2.4069, respectively, which satisfies the hypothesis of EKC curve.

Predicting Recurrence for Patients With Ischemic Cerebrovascular Events Based on Process Discovery and Transfer Learning.

The recurrence of Ischemic cerebrovascular events (ICE) often results in a high rate of mortality and disability. However, due to the lack of labeled follow-up data in hospitals, prediction methods using traditional machine learning are usually not available or reliable. Therefore, we propose a new framework for predicting the long-term recurrence risk in patients with ICE after discharge from hospitals based on process mining and transfer learning, to point out high-risk patients for intervention. First, process models are discovered from clinical guidelines for analyzing the similarity of ICE population data collected by different medical institutions, and the control flow found are taken as added characteristics of patients. Then we use the in-hospital data (target domain) and the national stroke screening data (source domain), to develop risk prediction models applying instance filter and weight-based transfer learning method. To verify our method, 205 cases from a tertiary hospital and 2954 cases from the screening cohort (2015-2017) are tested. Experimental results show that our framework can improve the performance of three instance-based transfer algorithms. This study provides a comprehensive and efficient approach for applying transfer learning, to alleviate the limitation of insufficient labeled follow-up data in hospitals.