Technical Approach to Converting Medical Information System (MIS) "ESCULAP" Archives for Artificial Intelligence Tasks: Experience of State Institution of Science “Center of Innovative Healthcare Technologies” State Administrative Department

The digital transformation of healthcare requires converting accumulated medical data into formats suitable for analysis. A significant portion of medical information in Ukraine is stored in archives of legacy systems, including the medical information system (MIS) “Esculap”. These data rely on the dBase structure (.DBF and .FPT files), which prevents their direct use in modern artificial intelligence and machine learning applications. This paper describes a technical approach to converting such archives. An algorithm was developed in Python using the libraries dbfread, pandas, and numpy. The proposed method enables extraction and systematization of depersonalized patient data, diagnoses, and treatment histories. Particular attention is given to resolving text encoding issues and processing MEMO fields stored in .FPT files. The result of the study is the transformation of relational tables in legacy formats into CSV files compatible with contemporary analytical tools. The resulting datasets can be used for machine learning tasks, neural network development, and statistical research in public health. The proposed approach was tested on archival data of the State Institution of Science “Center of Innovative Healthcare Technologies” of the State Administrative Department. The developed algorithm ensures accurate preparation of longitudinal medical data for further morbidity analysis. The software tool enables annual updates of research databases by converting newly generated archival records into a structured and analysis-ready format. The practical implementation of the algorithm included the inventory of 195 archival files, identification of complete .DBF/.FPT pairs, and validation of MEMO field integrity. The conversion process accounted for Windows-1251 encoding specifics, the presence of corrupted or incomplete tables, and the risk of automatic data type alteration when opening CSV files in spreadsheet editors. This approach minimized information loss and preserved logical links between patients, clinical episodes, and textual medical records