The aim is to evaluate mediastinal lymph node assessment using deep learning without needing detailed pixel-level annotations.
Developed models using deep learning techniques.
Divided the assessment problem into subtasks for better management.
Integrated prior knowledge to enhance model performance.
Models showed improved or comparable performance without pixel-level annotations.
Integration of prior knowledge led to significant enhancements in assessment accuracy.
Abstract
The division of the problem setting into subtasks as well as the integration of prior knowledge enables better or comparable performance of models trained with and without segmentation masks.