Fast and efficient method for quality assurance of deformable registration of images from PET or CT scans, particularly for adaptive radiotherapy.
- Quantitative metrics of registration quality designed for routine use that are algorithm-independent, labor-efficient, and accurate.
- Increased usability and accuracy in detecting unrealistic warping over classical registration assessment methods.
Image guided radiation therapy (IGRT) is a relatively new approach in imaging before and after treatment visualizes anatomical changes. This image-guided therapy improves treatment accuracy and outcome because it takes into account anatomical changes that occur during the course of radiation therapy. It is important that the images input into the algorithms are of high quality because an algorithm’s accuracy is strongly dependent on the quality of input images. The posture of the patient, for example, can change between scans which can alter the results of analyses. These changes are corrected by deformable registration, but artifacts and inconsistencies can occur during these corrections but currently there is no standard method to correct these inaccuracies.
Deformable registration is an essential tool for IGRT that is used for segmentation, dose tracking, and response assessment. Emory investigators have developed a fast and practical method for quality assurance of images that easily identifies regions where the registration is unrealistic or inaccurate. When this method was tested on clinical cases, it quantitatively detected errors in the displacement field and identified problematic regions that were previously invisible to classical voxel-based evaluation methods.
Method was tested on a number of inter- and intra-patient cases using both single and multi-modality registration algorithms.