Exhibit Hall | Forum 1
Purpose: Respiratory motion may substantially change the volume and SUV of small lesions in PET images for thoracic and abdominal patients. The hypothesis of the study is that PET images blurred by respiratory motion can be corrected with 4DCT data. In this study we will use an experimental moving phantom to benchmark a blurry image decomposition (BID) algorithm to reduce PET motion artifacts for stereotactic body radiotherapy (SBRT).
Methods: The BID algorithm was developed that uses a motion model to decompose a blurry PET image into a set of motion-freeze images. To verify this algorithm, a capillary tube containing ~0.2 μCi of 18F-FDG was attached on a 4DCT moving phantom. The phantom was set in the stationary and moving mode, respectively, and PET images were acquired with a clinical PET/CT scanner (GE Discovery 710). 4DCT was acquired from the moving phantom and time-equally sorted into 20 phases. The moving distance of the phantom in each phase was measured from the 4DCT and substituted into the BID algorithm to correct PET motion artifact.
Results: From the moving, static and BID-generated images, measured maximal activity concentrations of the capillary tube are 1.3391×104, 4.1722×104, 4.9651×104 Bq/ml, and measured profiles in the moving direction have the full width half maximum (FWHM) of 23.47±0.49, 5.61±1.14 and 6.1±0.91 mm, respectively. The BID correction reduced the motion-induced error from 317.5% to 8.7% in FWHM and 67.9% to 19.0% in activity concentration. With 1000 Bq/ml taken as a threshold, the contoured source has a volume of 19.47, 11.47 and 8.57 cc in the moving, static and BID-reconstructed images. The BID correction reduced the motion-induced volumetric error from 69.7% to 25.2%.
Conclusion: The phantom test has verified that the BID algorithm can help correct motion-induced error in activity concentration and reduce target volume in PET scans for SBRT patients.