The existing decision support systems used in healthcare for analyzing and processing medical data is considered in the article, their functionality is discussed. The solution for developing decision support systems to diagnose bronchopulmonary diseases, allowing to establish the patient's primary diagnosis treatments based on the integration of intelligent information processing, machine learning, pattern recognition, and extraction knowledge is proposed. The scheme of the proposed clinical decision support system for the diagnosis of respiratory diseases is discussed. Developing the clinical decision support system with the list of proposed capabilities will, on one hand, significantly improve the quality of medical care, since it reduces risks of human factors due to the use of computer-based information processing, and, on the other hand, increase the level of digitalization in medical institutions as well as their economic efficiency.
Keywords-clinical decision support system, data processing, machine learning, pattern recognitionI.
The article considers the basic algorithm of text and table data extraction used in the developed system of clinical decision-making in diagnosis of respiratory diseases, methods of formation of the data structure of an individual patient, a set of data from all patients for further application in models of machine learning as well as construction of ML models which provide detection of disease in patients. Data extraction and generation processes are performed in the Python programming language using additional libraries: "docx" and "pandas" for data processing and "sklearn", "lightgbm" and "catboost" for building machine learning models. The relevance of the task is due to large volumes of unstructured data received by the CDSS input and necessary for its effective functioning. The novelty of development lies in application of a set of existing and development of new algorithms of extraction and primary processing of text and table information.
This article considers the basic algorithm of graphic data extraction used in the developed system of clinical decision-making in diagnosis of respiratory diseases, methods for processing files in JPEG and DICOM formats, visualizing and preprocessing images as well as constructing neural network models based on a convolutional neural network that provide detection symptoms of pneumonia in patients based on radiographs. Data extraction and processing are performed in the “Python” programming language using additional libraries: “pydicom” for processing DICOM files. “Pillow” for visualization, “Keras” for building a convolutional neural network model. The relevance of the task is due to the large volumes of graphical data supplied to the input of the CDSS and necessary for its effective functioning. The novelty of this development lies in the application of a set of existing and development of new algorithms for extracting and processing graphic information.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.