PARASIDE: An automatic paranasal sinus segmentation and structure analysis tool for magnetic resonance imaging

Area of Science:

• Medical Imaging
• Radiology
• Computational Anatomy

Background:

• Chronic rhinosinusitis (CRS) affects 5%-12% of the population, significantly impacting quality of life.
• Objective assessment of CRS severity is challenging, with current methods like the Lund-Mackay score (LMS) being subjective and manual.
• Existing scoring systems lack objective quantification of sinus inflammation.

Purpose of the Study:

• To introduce PARASIDE, an automated tool for paranasal sinus segmentation on T1-weighted MRI.
• To quantify sinus features (volume, thickness, intensity) for objective disease assessment.
• To accurately predict the Total Lund-Mackay Score (TLMS) using automated quantitative features.

Main Methods:

• Automated segmentation of air and soft tissue volumes in maxillary, frontal, sphenoid, and ethmoidal sinuses using T1-weighted MRI.
• Quantification of extracted features like volume, thickness, and intensity.
• Regression analysis to predict TLMS based on segmented features, compared against Quantitative Opacification Score (QOS) and Quantitative Lund-Mackay Score (QLMS) baselines.

Main Results:

• PARASIDE achieved a mean-squared error (MSE) of 2.444 and mean absolute error (MAE) of 1.181 for TLMS prediction.
• PARASIDE outperformed the QOS/QLMS baseline, which had an MSE of 3.784 and MAE of 1.445.
• The segmentation accuracy was high, with a mean Dice similarity coefficient of 0.882 ± 0.138 and an average symmetric surface distance (ASSD) of 0.311 ± 0.354 mm.

Conclusions:

• PARASIDE provides the first automated whole-paranasal sinus segmentation for T1-weighted MRI.
• The system accurately predicts CRS severity using quantitative features, surpassing existing volumetric scoring methods.
• PARASIDE offers a reproducible, objective tool for clinical workflows, potentially reducing variability and accelerating reporting.