Inter-Observer variability in organs at risk contouring among radiation therapy students and qualified radiation therapists.

INTRODUCTION

Contouring organ-at-risk structures (OARs) remains an important skill that must be developed through education and training. This study assessed variations in OAR contouring among undergraduate radiation therapy students and qualified radiation therapists (RTs).

METHODS

Four planning computed tomography datasets (Brain, Lung, Parotid, and Prostate), which included 21 clinician-validated OAR contours, were utilised as references in this contouring study. Participants included 2nd to 4th year students, as well as qualified RTs. A quantitative comparison of contours was performed using the Dice Similarity Coefficient (DICE) and Hausdorff distance (HD) metrics. Statistical analyses assessed variations across structures, tumour sites, experience levels, and contouring methods.

RESULTS

A total of 440 OAR contours were analysed. The bladder and lung OARs achieved the highest mean DICE scores (>0.9), while structures such as the brainstem, heart, and parotid gland exhibited mean DICE scores between 0.76 and 0.89. Smaller structures, such as the lens, demonstrated very low HD95 values. A moderate positive correlation (r = 0.591) was observed between OAR volume and DICE scores (<200 cc). Statistical analyses indicated significant differences across tumour sites (P < 0.001 for DICE, HDavg, HD95, and HDmax). The comparison of manual and guided contouring showed statistically significant differences only for DICE (P < 0.001) and HDmax (P = 0.004). There were no significant differences in median scores between students and qualified groups. However, students exhibited higher variance than qualified professionals.

CONCLUSION

This study highlighted OARs that are challenging to contour or edit, suggesting the need for a comprehensive educational framework.