Fuzzy masks: boosting radiomic reliability in head and neck tumors amid delineation uncertainty

Key Points

• This research aims to improve radiomic reliability in head and neck tumors despite contour delineation uncertainties.
• Developed fuzzy mask models to address tumor contour uncertainty.
• Compared performance of fuzzy masks with traditional binary masks.
• Assessed feature reliability and clustering of predictive outputs.
• Fuzzy masks yielded 29 more reliable features than binary masks.
• Gradient weighting achieved 2% more independent feature clusters.
• The intraclass correlation coefficient of predictive outputs reached 0.99.