(2023) Kucheruk, Dariia; Hlybovets, Andriy; Brudno, Michael
This thesis proposes a multi-label segmentation model for optimizing the segmentation of neonatal brain MRI with partially annotated data. A multi-label segmentation model that addresses the challenges of limited annotated data by modifying the preprocessing, loss function, and postprocessing of the original multi-class label segmentation was developed. The proposed approach aims to improve the accuracy and efficiency of neonatal brain MRI segmentation by leveraging partially annotated data. We evaluate our method on a unique dataset of neonatal brain MRI and demonstrate its effectiveness compared to the models trained on fully annotated data.