AI-powered data analysis means new opportunities for patients and clinical research

Hospital systems contain vast amounts of data that, in the current technical environment, are mainly used to record and store important information during medical care. This data is collected from a wide range of subsystems and contains dozens to hundreds of parameters that the treating physician works with. He often has to go through extensive medical records to find the context he is looking for. PatientFinder, which is currently being implemented by St. Anne’s University Hospital Brno (FNUSA), is a new technology platform designed to enable more efficient work with information from these systems.

The groundwork for the project to implement a software tool to analyze hospital data was laid in 2021, when FNUSA began selecting the most suitable implementer. Key aspects of the selection were data security, exclusive control over data flows, installation of the hardware and software environment within the hospital infrastructure, guarantee of use for research purposes and opening up new opportunities for patient participation in experimental treatment projects and clinical trials. During 2022, we have negotiated the technical and legal framework for cooperation with the supplier, and since March 2023, activities related directly to the deployment of the software, which bears the name PatientFinder, are underway. This will use AI (artificial intelligence) modules in two key functions. The first is pseudonymisation, anonymisation and data cleansing, and the second is the identification of parameters from medical records.

In the first phase, the data is cleansed of primary and secondary identifiers, and the AI recognizes and then deletes over 20 groups of attributes (phone numbers, names, family members, addresses, birth dates, etc.) in the text so that the data already cleansed can be processed in subsequent operations. This means that patients do not have to worry about leaking personal and sensitive information. In the second phase, the analysis and identification of parameters from the text is done, where the unstructured data is converted into a structured database. Thanks to AI, more than 300 parameters (e.g. blood pressure, medication or lab results) are identified in the medical records. The standardization of these records is done by using Natural Language Processing (NLP) methods, which can be very simply translated as automated recognition of information from text using AI. It is AI that helps analyse huge amounts of data in a short time, correcting typos, formats, incorrect or incomplete entries and units, or eliminating erroneous information, while looking for relationships between data. However, for AI to work properly, the system needs to be trained specifically for the Czech environment and the Czech language. This “training” will be covered by a team of doctors and researchers from the International Clinical Research Center (ICRC) in the coming months. The goal is to have the software “trained” by the end of summer 2023.

The expected launch of the system in autumn 2023 will benefit patients and the scientific community. Patients will have greater opportunities for involvement in clinical trials and experimental treatments. The physician will be able to more easily locate a group of patients who fit the criteria for a given type of clinical trial. In addition, patient outreach will be faster. Knowing the structure of the patient base, key attributes and factors, it will also be possible to obtain a type of study for Brno that will directly address the needs of patients. For research, this platform is a valuable tool that will enable more accurate results to be generated as it can process large volumes of data. Cohort studies so far work with hundreds of patients who are included on purpose. In the case of this tool, hundreds of thousands of records can be safely processed. Research and recommendations for clinical practice can thus be built on more representative data.

Within the Czech Republic, the implementation of PatientFinder is a pilot project, but it is a proven concept used by many hospitals abroad. In addition, FNUSA envisages a broader plan for the use of advanced AI features. Our hospital is involved in other international grants, and in cooperation with foreign institutes and universities we are developing special tools that will help the digitalization of healthcare itself in the future.

Author: Mgr. Michal Janota, Vice-head for Operations, ICRC,

Contact for media: Ing. Jiří Erlebach. Head of PR and Marketing, Spokesperson, FNUSA, +420 543 182 006.

ICRC is a joint workplace of FNUSA and LF MU.