Prístupy radiomics a strojového učenia u pacientok s endometriózou

Abstract

This work deals with the use of radiomic features extracted from MRI images for the classification of endometriotic lesions and ovarian cysts using machine learning methods. The theoretical part focuses on the issue of endometriosis, its diagnostic possibilities, the principles of radiomics and selected classification algorithms. The practical part of the work includes the processing of MRI data of patients from the Ostrava University Hospital, preprocessing of image data, extraction of radiomic features from manually segmented lesions and selection of significant features based on statistical analysis. These features were subsequently used as inputs for three classification models – logistic regression, Random Forest and Support Vector Machine. The models were trained, validated and tested in the Python using the PyCaret library. The best classification results were achieved by the Random Forest model with an AUC=0,99, which indicates the high potential of the combination of radiomics and machine learning for non-invasive distinction between two types of gynecological pathologies. In the final part, the results, benefits and limitations of this work are discussed.