Search

The study analyzed scientific texts based on a manually created database of synopses of theses in dentistry. The main goal was to structure medical texts into various topics by means of natural language processing techniques (topic modeling). Furthermore, a dynamic topic modeling showed the most popular in the field of dentistry over almost the last thirty years.

Learning treatment methods and disease progression is significant part of medicine. Graph representation of data provides wide area for visualization and optimization of structure. Present work is dedicated to suggest method of data processing for increasing information interpretability. Graph compression algorithm based on maximum clique search is applied to data set with acute coronary syndrome treatment trajectories. Results of compression are studied using graph entropy measures.

Endometrial cancer (EC) is the most common gynecological tumor in high-income countries, and its incidence has increased over time. The most critical risk factor for EC is the long-term unopposed exposure to increased estrogens both exogenous and endogenous. Machine learning can be used as a promising tool to resolve longstanding challenges and support identification of the risk factors and their correlations before the clinical trials and make them more focused. In this paper we present the results of the research of the correlation analysis of Endometrial cancer risk factors. The study was performed with EC patients of the Almazov center in Saint-Petersburg, Russia. All women involved in the current study underwent radical surgical intervention due to EC. After initial cancer treatment, they were referred to the Almazov center outpatient specialists for follow-up visits. Many of them were readmitted of the inpatient clinic due to relapse. We extracted a variety of parameters related to lifestyle, dietary habits, socioeconomic, and reproductive features from the inpatient and outpatient databases of Almazov center. The medical records of the women with enough data were included in the study. Prediction of Progression-free survival (PFS) and overall survival (OS) were analyzed respectively. The AUC of ROC was calculated for PFS = 0.93 and for OS = 0.94.

Business process modeling aims to construct digital representations of processes being executed in the company. However, models derived from the event logs of their execution tend to overcomplicate the desired representation, making them difficult to apply. The most accurate recovery of the business process model requires a comprehensive study of the various artifacts stored in the company's information system. This paper, however, aims to explore the possibility to automatically obtain the most accurate model of business process, using mutual optimization of models recovered from a set of event logs. Further, the obtained models are executed in multi-agent simulation model of company, and the resulting event logs are examined to determine patterns that are specific to distinct employees and those that generally characterize business process.

The relevance of this study lies in improvement of machine learning models understanding. We present a method for interpreting clustering results and apply it to the case of clinical pathways modeling. This method is based on statistical inference and allows to get the description of the clusters, determining the influence of a particular feature on the difference between them. Based on the proposed approach, it is possible to determine the characteristic features for each cluster. Finally, we compare the method with the Bayesian inference explanation and with the interpretation of medical experts [1].

The present study is devoted to interpretable artificial intelligence in medicine. In our previous work we proposed an approach to clustering results interpretation based on Bayesian Inference. As an application case we used clinical pathways clustering explanation. However, the approach was limited by working for only binary features. In this work, we expand the functionality of the method and adapt it for modelling posterior distributions of continuous features. To solve the task, we apply BEST algorithm to provide Bayesian t-testing and use NUTS algorithm for posterior sampling. The general results of both binary and continuous interpretation provided by the algorithm have been compared with the interpretation of two medical experts.

Developing predictive modeling in medicine requires additional features from unstructured clinical texts. In Russia, there are no instruments for natural language processing to cope with problems of medical records. This paper is devoted to a module of negation detection. The corpus-free machine learning method is based on gradient boosting classifier is used to detect whether a disease is denied, not mentioned or presented in the text. The detector classifies negations for five diseases and shows average F-score from 0.81 to 0.93. The benefits of negation detection have been demonstrated by predicting the presence of surgery for patients with the acute coronary syndrome.

The main goal of this paper is to develop a spell checker module for clinical text in Russian. The described approach combines string distance measure algorithms with technics of machine learning embedding methods. Our overall precision is 0.86, lexical precision - 0.975 and error precision is 0.74. We develop spell checker as a part of medical text mining tool regarding the problems of misspelling, negation, experiencer and temporality detection.

We study the way dynamics affects modelling in chronic heart failure (CHF) tasks. By dynamics we understand the patient history and the appearance of new events, states and variables changing in time. The goal is to understand what impact past data has on prediction quality. Three different experiments have been conducted: CHF episode results prediction (better, worse, no change), CHF stage classification and heart rate value prediction. For modelling we use clinical data of CHF patients. For each task the groups of static and dynamic features are selected and analyzed. For each task 3 machine learning algorithms were trained: XGBoost, Logistic Regression, and Random Forest for multi classification and Linear Regression, Decision Tree and XGBoost for the regression task. Different combinations of features were examined from both groups applying forward feature selection algorithm. The results confirm that the highest predictions quality is reached with combination of static and dynamic features.

This study proposes a graph-based method for representing the dynamics of chronic diabetes as a complex process with different characteristics. The study was based on the case histories of 6864 patients with diabetes mellitus, 90% of whom suffer from type 2 diabetes. Our method allows to predict the sequence of events during the development of type 2 diabetes for each patient. Typical developmental trajectories of the disease were investigated, their clustering was carried out, the trajectory patterns were identified and studied. Based on the constructed directed graph reflecting transitions between different conditions of the patients, the clustering of diabetic statuses was carried out using the Modularity Class method; 8 clusters were selected, each of them was interpreted and studied. The method of the disease developmental trajectories creation by means of machine learning methods was described. Unlike static models of a disease course, this method considers complete past information on the patient and his or her previous events, using each event of the course of disease to predict the next event.

The study is aimed towards the assessment and improvement of complex informational, healthcare, and medical processes existing in the distributed environment during monitoring and providing healthcare services to patients with chronic diseases. The study is performed within a framework of a project for development a complex unified medical information system capable to work on large scale (on population-level). The presented experimental studies show that predictive modelling enables unified detection of remote patients' state using self-measuring and distributed electronic health records. In addition, an experimental mobile application is presented as a tool for data collection and user-system interaction during chronic disease monitoring and treatment.

This article describes the study results of echocardiographic (ECHO) test data for 4P medicine applied to cardiovascular patients. Data from more than 145,000 echocardiographic tests were analyzed. One of the objectives of the study is the possibility to identify patterns and relationships in patient characteristics for more accurate appointment procedures based on the history of the disease and the individual characteristics of the patient. This is achieved by using classifications models based on machine learning methods. Early detection of disease risks and "accurate" appointment of diagnostic procedures makes a significant contribution to value-based medicine. Moreover, it was also possible to identify the classes and characteristics of patients for whom repeated diagnostic procedures are well founded. Calculation of personal risks from empirical retrospective data helps to detect the disease in early stages. Identifying patients with high risk of disease complications allow physicians to make right decisions about timely treatment, which can significantly improve the quality of treatment, and help to avoid diseases complications, optimize costs and improve the quality of medical care.