Articles from rubric: «Artificial intelligence in health care»
-
2020 № 4 Moscow experiment on computer vision in radiology: involvement and participation of radiologists
B a c k g r o u n d . In 2019, the Moscow Government decided to conduct a large-scale scientific research – an the Experiment on the use of innovative computer vision technologies for medical image analysis and subsequent applicability in the healthcare system of Moscow (www.mosmed.ai).
O b j e c t i v e – analyze engagement, attitudes and feedback from doctors-radiologists in frame of the Experiment.
M a t e r i a l s a n d m e t h o d s . The Experiment is a prospective research approved by the Independent Ethics Committee and registered with Clinicaltrails.gov (ID NCT04489992). Patients signed informed voluntary consent. On the date 01.10.2020, ten services are involved in the Experiment, they providing automated analysis of chest computed tomography and x-ray, mammography. The study includes quantitative indicators of the Experiment from 06/18/2020 to 10/01/2020. Methods of social survey, descriptive statistics, assessment of diagnostic accuracy metrics were used.
R e s u l t s a n d d i s c u s s i o n . During the first four months of the active phase of the Experiment, ten computer vision services were integrate into Unified Radiology Service of Moscow. More then 497 thousand studies have been successfully analyzed. Analyzes is carried out for 884 diagnostic devices in 293 medical organizations, 272 of them are actively involved. The involvement of medical organizations is 82%. The median time for automatic analysis of 1 study is 8 minutes. Overall, 63% of studies were analyzed in less than 15 minutes. At the beginning of the Experiment, 538 doctors had access to the system; in four months this number increased to 899. The involvement of doctors was 24%, which is slightly higher than the global indicators. According to the results of a sociological survey, the attitude to AI technologies of Moscow radiologists can be characterize as expectant, moderately optimistic. Radiologists
have determined that the results of computer vision services are fully consistent with the real situation in 64% of cases. In 36% cases some inconsistencies were recorded; of this number, significant discrepancies took place in 6%, insignificant – in 23%.
C o n c l u s i o n . Results of the Experiment’s first four months can be consider as successful. A high level of involvement of radiologists is define. Special measures will be implement to increase the involvement of radiologists, as well as a comprehensive comparative assessment of the work of services at the further stages of the Experiment. -
2020 № 4 Predictive analytics technologies in the management of the COVID‑19 pandemic
Recently, a new coronavirus infection, or COVID‑19, caused by the pathogen SARS-CoV‑2, has been continuing to
spread around the world rapidly. According to the World Health Organization (WHO), which declared this outbreak a pandemic, COVID‑19 is a serious public health problem of international concern. Due to the lack of proven effective treatment and vaccination against COVID‑19, precautions are considered by WHO to be strategic goals and a primary response to the pandemic. It is recommended that country guidelines adopt national health care programs aimed at assessing and reducing the risk of infection spread. Predictive analytics have begun to be actively used to compile population and personal forecasts of the progression of morbidity, mortality, assess the severity of the course of the disease, etc. This article provides an overview of available developments and publications on the use of predictive analytics in the management of COVID‑19 pandemic. -
2020 № 4 Increased accuracy of prediction of fragmentation duration of urinary stones based on multifactorial regression models
The regression models for prediction of contact holmium lithotripsy duration are given. Models are obtained on the
basis of calculated and experimental data on duration of different stages of laser lithotripsy. They allow, based on the volume and radiological density of urinary stones and taking into account the anatomical characteristics of the patient, to calculate the expected time of complete fragmentation of the stones with a higher accuracy than on the factor of additional costs the known model based -
2020 № 3 The first Russian breast pathology histologic images data set
Data set of annotated histology images on breast pathology is provided, containing more than 40 thousand images
from 104 microscopic slides and 92 patients and additional clinical data (age, TNM, grade, WHO type). The data set is prepared
in compliance with relevant procedures for clinical research at Burnasyan Federal Medical Biophysical Center Of Federal
Medical Biological Agency. The data set is accessible at GitHub for research and educational purposes. -
2020 № 1 Modern methods of analysis and forecasting of time series and use in medicine
The article is a review of Russian and foreign scientific publications related to the use of methods of analysis and
forecasting of time series in medicine. 112 major publications over the past 5 years, located on the Internet resources e-library
and PubMed, are considered. Examples of the application of such methods as exponential smoothing, regression analysis,
the ARIMA method and their variants for time series analysis are shown. Various approaches to mathematical modeling of
the time series are presented. The results of the article can be used to select a method of analysis and forecasting time series
depending on the tasks. -
2019 № 4 Prediction of recurrence in patients with Cushing’s disease after successful endoscopic transnasal adenomectomy: neural network model and its software implementation
Introduction. Due to the high frequency of recurrences in patients with Cushing’s disease after endoscopic transnasal adenomectomy (up to 55% in the 5 year period), it is important to develop a method for predicting recurrence of the disease based on a combination of factors. Мaterials and methods. The study included 219 patients who underwent endoscopic transnasal adenomectomy in 2007–2014.Over 3 years, remission persisted in 172 patients; relapse developed in 47 patients. The construction of artificial neural networks of various topologies was performed in the Statistica v. 13, and then software was developed for the best network.
Results. A highly efficient neural network (3-layer perceptron) was constructed, which allows predicting recurrence within 3 years or remission for at least 3 years. The sensitivity of the model is 74%, the specificity 97%, the positive predictive value 85%, the negative predictive value 93%. The predictors of the model are sex, age, duration of the disease, MRI type of adenoma, levels of adrenocorticotropic hormone and cortisol in blood in early postoperative period. Web-calculator was developed and is available to doctors for free practical use on http://medcalc.appspot.com/.
Сonclusion. The software implementing neural network is a quite effective tool for predicting recurrence and it will allow to perform personalized approach to management of patients who underwent neurosurgical treatment for the Cushing’s disease. -
2019 № 3 Development of algorithm for searching of clinically homogeneous patients from semistructured text data of oncological electronic health record
The growth in the number of patients with malignant neoplasms in Russia significantly increases the load on a specialized network of oncological institutions and oncologists. It is most likely that this trend will continue in the coming years. One of the ways to improve the efficiency of medical activity is the extraction knowledge from medical data arrays, using modern data analysis methods, by clustering patients into groups of clinically homogeneous (similar) patients from electronic health records. The aim of the study is to develop an algorithm for finding clinically homogeneous patients according to the electronic health records of the oncological dispensary, with follow-up possibility of integration into the clinical decision support system (CDSS). The use of such CDSS in practical medicine and in the field of medical education will allow us to analyze both semistructured and unstructured arrays of information, which will require further implementation and improvement of information systems at all levels of
medical care. The homogeneity of patients was determined by machine learning by cosine distance in the space of vector representations of electronic health records. An experiment on 20 randomly selected electronic health records of patients of Krasnodar Regional Oncological Dispensary showed high efficiency of the algorithm in creating clusters of clinically homogeneous patients. -
2019 № 3 Prospects for the using of machine learning methods for predicting cardiovascular disease
Morbidity and mortality from cardiovascular diseases (CVD) has remained the leading rate in recent decades worldwide. Primary prevention methods based on the management of cardiovascular risk factors are most effective in reducing the burden of CVD. In preventive medicine for risk management of CVD use the riskometers – scales that was obtained as a result of long prospective studies.But the practical application of the developing scales has showed the limitations in the forecast accuracy. Machine learning makes it possible to improve the accuracy of cardiovascular risk prediction due to nonlinear relationships of their deeper adjustment between risk factors and disease outcomes.
2236 patients’ data were used. We trained the model on the features used in the Framingham scale construction. We compared the resulting model and the Framingham scale for the accuracy of the cardiovascular event prediction. Thus, according to the ROC analysis for the Framingham scale, the indicators are as follows: precision Accuracy: 70,0%, the AUC: 0.59. At the same time for the model obtained using machine learning similar indicators were: Accuracy: 78,8%, AUC: 0.84. Thus, the use of machine learning algorithms including deeplearning algorithms can significantly improve the accuracy of cardiovascular risk prediction of trained models. -
2019 № 3 Experience in developing and implementing an oncological formations searching system using artificial intelligence with the example of X‑ray computed tomography of the lungs
The experience of creating and implementing AI system Botkin.AI for identifying pulmonary nodules according to CT
data is considered. The main parameters of the model are described, the results of pilot projects of platform practical application in several regions of the Russian Federation are presented. Examples of a platform application for identifying pulmonary nodules with various size and localization are given. During pilot projects in the regions, 7 patients with high suspicion of pulmonary malignancy were identified. The results and experience show that the applying of the Botkin.AI system can be used both for the implementation of regional lung cancer screening programs and as an additional tool to increase the lung cancer detections with introducing automatic revision of chest CT data, regardless of these studies indications. -
2019 № 2 Application of a neural network machine learning method to complication diseases treatment in hemodialysis patients
-
2018 № itm Forecasting the stage of adenomiosis with neural networks
Adenomyosis is a widespread gynecological disease, which is often accompanied by infertility. There are problems with the diagnosis of the disease, since the disease has various clinical manifestations, including often asymp- tomatic course of the disease. Due to the complexity of the diagnosis and according to different sources, its frequency varies from 5% to 70%. An equally difficult problem is determining the stage of the disease. The stage determines the tactics and strategy of treating patients. A sample of 84 patients with adenomyosis, using the Spearman rank correla- tion coefficient, revealed indicators that are interrelated with the stages of the disease. In this work, the application of the heuristic procedure to neural networks for predicting the laboratory-clinical indicators of the adenomyosis stage is considered. A software application has been developed that allows you to predict the stage of adenomyosis without resorting to hysterectomy. The methodological value of the work is that, using the example of a common gynecological disease, it is shown that the use of modern data analysis tools opens up wide possibilities for solving prognostic prob- lems of determining patients’ belonging to certain classes according to the stages or types of the disease. Software applications that automate the procedure for classifying patients can form the basis of various systems of support for making medical decisions.
-
2018 № 4 Black box problem overcoming in medical applications of machine learning
A new interface for machine learning predicting models is proposed. Approach is based on optimal valid partitioning (OVP) technique and the modified method of statistically weighted syndromes (LSWR). The interface allows you to overcome the problem of “black box” illustrating prediction process by scatter plots, ROC curves and informative indicators ranking
-
2018 № 3 The basic recommendations for the creation and development of information systems in health care based on artificial intelligence
Artificial intelligence is becoming one of the main drivers in solving serious problems of medicine and health, such as inadequate resources, further improving efficiency, quality and speed of work. All over the world, more and more solutions are being developed in this area. However, the more new products appear, the more questions and problems arise.
The work analyzes some foreign publications and research results, which studied the main problems associated with the creation and implementation of artificial intelligence in health care. As a result of the analysis, a number of practical recommendations were formulated that will help increase the likelihood of successful creation and introduction of such products in the practical link of health. -
2017 № 4 Compliance control of the hospital ad- verse events prevention protocols with the medical information system
The paper presents the results of using the hospital’s information system (HIS) to analyze the compliance with the requirements of standard operating procedures (SOPs) for the hospital adverse events (HAE) prevention. Templates were embedded into medical record form to register the inpatient’s risk level for venous thromboembolism, falls and pressure ulcer, as well as to prescribe corresponding preventive measures. The data uploaded from the HIS were used to calculate the pro- cess and outcome indicators for HAE prevention, and to correlate them with the HAEs. This approach allows to establish the effective control of compliance with requirements of public oversight authorities for healthcare quality and safety assurance.
-
2017 № 4 Experience of the Development of the Software Package for Neural Network Diagnosis and Prediction of Diseases of Hepatopancreatoduodenal Zone
The article presents the experience of the internal development of a software package for diagnosis and forecasting diseases of hepatopancreatoduodenal zone based on the artificial neural network of multilayer perceptron type with hyperbol- ic tangent taken as an activation function. The article includes the characteristics of the analyzed data which is the set of risk factors for the development of peptic ulcer, cholecystitis and pancreatitis and substantiates the necessity for the application of automated control systems acting on the principles of artificial neural networks. The methods of operating of a multilayer per- ceptron are given, and there are proposed modifications intended to optimize the development of the software package and to solve a number of problems that arise during practical implementation of the system and during data preparation. A set of possible input and output parameters of the network, intended for its training, is proposed. The article contains the description of the practically developed user interface, intended to create, configure, train and clinically apply the artificial neural network, as well as to construct its graphs and statistically control its functioning.
-
2017 № 3 Prospects for neural networks and deep machine learning in creating health solutions
The paper gives an overview of the prospects of using neural networks and deep machine learning in the creation of artificial intelligence systems for healthcare. The definition and explanations on the technologies of machine learning and neural networks are given. The review of already implemented artificial intelligence projects is presented, as well as the forecast of the most promising directions of development in the near future
-
2017 № 2 Determination of risk factors of cardiovascular mortality in institutions of the penitentiary system using the machine learning methods.
The article presents the results of the first clinical and epidemiological studies to identify risk factors for fatal cardiovascular disease in patients of medical institutions of the penitentiary system. In a research the Machine learning methods based on creation of optimal partitioning signs space and recognition methods were applied. It is with high reliability has allowed to determine predictive factors of hospital mortality in cardiac patients, which became: the use of a strong tonic drink «chifir», age, weight, growth, systolic and diastolic blood pressure, hemoglobin level, heart rate, left ventricular ejection fraction, end-systolic and end-diastolic left-ventricular dimension, hypertension and the number of criminal record.