The Menze Lab at DQBM contributed to the development of PARASIDE, an automated segmentation and analysis framework for paranasal sinuses in T1-weighted MRI, now published in Computers in Biology and Medicine . The tool segments air and soft tissue compartments across all major sinus structures and extracts quantitative features to predict the Total Lund–Mackay Score.
PARASIDE achieved a mean Dice score of 0.882 and outperformed established volumetric baselines in disease severity regression, reaching a mean absolute error of 1.181. The framework enables reproducible, explainable, and radiation-free sinus assessment and supports large-scale imaging studies and clinical workflows.
Publication: https://doi.org/10.1016/j.compbiomed.2026.111511