Fulltext (PDF)
DOI: 10.48087/BJMS.2026.130306
Authors: Lilia NAOUN1,2, Sanah NACERDDINE2, Karima MELAIS2, Soumeya MANSOURI2, Amina AMIRECHE2, Abdesselem BRIHMAT2
Affiliations: 1- Université Badji-Mokhtar Annaba. 2- Service de radiothérapie – CHU Annaba – Algérie
Abstract
Introduction: Delineation of target volumes and organs at risk (OARs) is a crucial but time-consuming step in head and neck (H&N) radiotherapy planning and is associated with significant interobserver variability. Atlas-based automatic segmentation systems, such as ABAS (Atlas-Based Automatic Segmentation), have been proposed as a promising approach to improve efficiency and standardization in clinical practice. Objective: To evaluate the geometric and dosimetric agreement between manual contouring and atlas-based automatic segmentation using ABAS in patients treated for nasopharyngeal carcinoma, and to assess the impact of the number of atlases on system performance. Methods: A prospective comparative study was conducted in 10 patients with nasopharyngeal carcinoma. Six OARs (spinal cord, brainstem, right and left parotid glands, optic chiasm, and mandible) were manually contoured by an experienced radiation oncologist and automatically segmented using ABAS. Volumes, maximum dose (Dmax), and mean dose (Dmean) were compared. The influence of the number of atlases (1, 10, and 20) and contouring time was analyzed. Correlations were assessed using Pearson’s correlation test, with statistical significance set at p < 0.05. Results: Using 10 atlases, strong correlations were observed between manual and automatic contours for most OARs (R > 0.9, p < 0.05), both volumetrically and dosimetrically. Larger discrepancies were noted for small-volume structures with complex anatomical boundaries, particularly the optic chiasm. The average contouring time was markedly reduced from approximately 4–5 hours for manual contouring to about 21 minutes using ABAS. Conclusion: Atlas-based automatic segmentation using ABAS demonstrates satisfactory agreement with manual contouring in head and neck radiotherapy while providing a substantial reduction in contouring time. This technology represents a reliable clinical decision-support tool, provided that systematic expert validation and manual correction are performed.
Keywords: Radiotherapy, Head and neck cancer, Atlas-based segmentation, Automatic contouring, Organs at risk.