Arkadiusz Sycz
supervisor: Artur Przelaskowski
The problem I address in my research concerns the identification of patients with blood cancers based on information collected in hospital information systems. The basic assumption of the experiment is the ability to improve the diagnosis of blood cancers with the help of expert knowledge supported by artificial intelligence. The raw data of the experiment are laboratory results, textual descriptions of patients, their diagnoses as well knowledge about state-of-art in diagnostics. The data are electronic health records sampled from databases of Polish hospitals. The challenge is to overcome data flaws, data mining and interpretation of results, performed under the supervision of a domain expert - a hematologist. We aim to conduct research using statistical modeling tools combined with domain knowledge engineering with the intention of confirming hypotheses, feature engineering, detecting new patterns hidden in the data and using them to build a reliable explainable predictive model in circumstances of uncertainty and incompleteness of real-world data, assessing model efficiency and utility. To achieve these goals, structure learning (networks, trees), probabilistic estimation of cause-and-effect relationships will be used.