Seminar: Dawid Zamojski
Dawid Zamojski gave a talk at a Department seminar entitled "Using machine learning methods to support the planning of the infertility diagnosis process".
During the seminar, he presented a research project focused on applying machine learning methods to support the infertility diagnostic process. The work is based on the analysis of diagnostic data from several areas of reproductive medicine, including genetics, cytogenetics, microbiology, cytology, and immunology.
The aim of the project is to investigate whether clinical data can support early pregnancy-outcome risk assessment and help plan more personalized diagnostic pathways for patients. The study also addresses several challenges typical of real-world medical data, such as missing values, class imbalance, limited feature sets, and the need for careful interpretation of predictive model performance.
The presentation covered exploratory UMAP analysis, missing-data imputation, and the comparison of several modeling strategies, including Random Forest, optimized XGBoost, and XGBoost combined with SMOTE.
One of the most interesting findings was that cytogenetic data showed particularly promising predictive potential in the analyzed dataset. At the same time, the project highlights that AI models in medicine require not only promising metrics, but also robust validation, bias control, and responsible clinical interpretation.
This marks an important stage in Dawid’s PhD research, combining reproductive medicine, clinical data analysis, and modern artificial intelligence methods.
