Precision and Reliability of Enhanced Cone Beam Computed Tomography for Head and Neck Cancer Radiotherapy

Summary: Enhanced cone beam computed tomography (eCBCT) offers improved image quality, potentially revolutionizing tumor response monitoring and normal tissue changes during radiation therapy (RT). This study assesses eCBCT's efficacy in accurately segmenting normal tissues and tumor volumes by comparing observer agreement and accuracy against traditional CT simulations (CTsim) in head and neck cancer (HNC) patients. We acquired paired CTsim and first fraction eCBCT images for five HNC patients undergoing definitive RT. A primary observer manually segmented 17 regions of interest (ROIs) per image on both modalities. Four independent observers also segmented eCBCTs. Deformable image registration software co-registered paired CTsim/eCBCT images. Agreement between eCBCT and CTsim segmentation volumes was calculated using volume overlap and surface distance metrics. The Mann-Whitney U test compared performance between different ROI types, and the Intraclass Correlation Coefficient (ICC) calculated inter-observer variability. Among 425 segmented ROIs, the median Dice similarity coefficient (DSC) was 0.8, mean surface distance (MSD) was 1.5 mm, and maximum Hausdorff distance (Max HD) was 9.3 mm. Mandibular bone ROIs performed best (average DSC 0.88, MSD 0.89 mm), followed by muscles of mastication and salivary glands. GTV ROIs had the worst performance (average DSC 0.5, MSD 4 mm), showing significantly lower metrics compared to other OARs (P< 0.001). ICC showed excellent inter-observer agreement for normal tissue ROIs (DSC: 0.93, Max HD: 0.97, MSD: 0.87) but moderate to poor agreement for GTV ROIs (DSC: 0.5, MSD: 0.71). Manual segmentation of eCBCT in HNC patients aligns closely with CTsim for most organs at risk, maintaining high observer agreement. However, GTV segmentation using eCBCT alone is unreliable. Further studies are needed to assess methods for augmenting eCBCT's capability in primary tumor delineation for future RT adaptation applications. These findings provide a foundation for potential implementation of eCBCT in adaptive radiotherapy strategies for HNC patients. Piyush Pathak, MD, MPH (Presenting Author) - Baylor College of Medicine; Zaid Siddiqui, MD (Co-Author) - Baylor College of Medicine; Cheryl Claunch, MD, PhD (Co-Author) - Baylor College of Medicine; Caleb Stewart, MD (Co-Author) - Baylor College of Medicine; Baozhou Sun, PhD (Co-Author) - Baylor College of Medicine; Shelly Sharma, MD (Co-Author) - Baylor College of Medicine; Daniel Hamstra, MD, PhD (Co-Author) - Baylor College of Medicine; Pavan Jhaveri, MD (Co-Author) - Baylor College of Medicine; Abdallah Mohamed, MD, PhD (Co-Author) - Baylor College of Medicine