WE‐E‐141‐10: Automated Patient Identification and Localization Error Detection Using X‐Ray Setup Images

Abstract

Purpose: To show that planar x‐ray radiotherapy setup imaging can be used to perform computer‐based automated patient identification and localization error detection. Methods: For this proof‐of‐concept study, setup imaging from a sample of 100 cranial radiotherapy patients, 100 prostate patients, and a group of 83 patients treated for spinal lesions was analyzed. The patients were treated using a commercially‐available image‐guided radiotherapy system that derives setup corrections by performing 2D‐3D registration. The 2D‐3D registration is performed by dynamically generating digitally‐reconstructed radiographs (DRRs) with varying translational and rotational shifts until a measure of the similarity between the x‐ray projections and DRRs is optimized. Using the cranial patients, and separately the prostate patients, we investigated whether the value of the optimized image similarity measure could be used to discriminate between cases where x‐ray projections and DRRs from the same patient (correct match) and from different patients (incorrect match). Using the spinal patients we investigated whether the optimized similarity measure could be used to discriminate between localizations performed to the correct vertebral body and to vertebral bodies off by one from the correct body. Results: A threshold value of the similarity measure could be chosen to separate correct and incorrect patient matches and correct and incorrect vertebral body localizations with excellent accuracy for the patient cohorts studied. A 10‐fold cross‐validation using linear discriminant analysis yielded misclassification probabilities of 0.000, 0.0045, and 0.014 for the cranial patient identification, prostate patient identification, and spinal localization cases respectively. Conclusion: Automated patient identification and localization error detection can be performed with excellent accuracy using an IGRT system with 2D‐3D matching. This technique can be used to reduce errors in radiotherapy with very little procedural overhead. Our group is currently extending this methodology to the more common case of image guided radiotherapy systems which use 2D‐2D matching.