On an Approach of the Solution of Machine Learning Problems Integrated with Data from the Open-Source System of Electronic Medical Records: Application for Fractures Prediction

The purpose of the work is to develop mathematical and software background for the development of ML models in medical research, which is based on the application of open-source EMR systems and ML tools. The flowchart includes basic steps of ML model development, including import and preparing the clinical data, the statement of task, the choice of method (learner), setting its parameters and model assessment. The problems dealing with dimension reduction which arise often in medical research are highlighted and solved with the help of modified principal component analysis (PCA) method. We analyze the problem of export-import data from EMR system and offer the ways of its solution within the most known open-source systems (OpenEMR and OpenMRS). The special attention is paid to the application of free open-source software in ML in medical research with the purpose of development of methodologies of prophylaxis and treatment. As an example, we consider the problem of development of classifier for fractures prediction where we describe all the presented steps of ML model development. With the help of benchmark of learners in the package mlr we compare different methods of ML when applying them in medical research.