MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Evaluation of Deformable Image Registration Accuracy and Its Impact On Dual-Energy CT Images

J Huang*, C Matrosic , M Lawless , L Di Maso , J Miller , University of Wisconsin, Madison, WI

Presentations

(Sunday, 7/14/2019) 5:00 PM - 6:00 PM

Room: 221CD

Purpose: Several DECT techniques require a deformable registration of the high- and low-energy images. This work evaluates the accuracy of a deformable registration algorithm used in dual-energy CT image processing and quantifies the impact of registration errors on iodine concentration accuracy and monoenergetic HU values.

Methods: The deformable registration algorithm implemented in Syngo.via (Siemens) was evaluated using two phantoms designed to mimic distinct types of motion encountered clinically: a deformable PVC plastisol-based phantom to mimic abdominal motion due to respiration and a rigid phantom to mimic sliding motion in the thorax. Both phantoms were imaged using sequential 80kVp and 140kVp scans with motion applied to the phantoms between scans. Syngo.via does not allow the export of the deformed images, so this study focused on quantifying the accuracy of various dual-energy images resulting from processing of the deformed low-energy and high-energy images (mixed 120kVp, monoenergetic, and iodine images).

Results: The algorithm performed well for deformation in the abdomen phantom with a mean registration error of 0.4mm for landmark analysis of mixed 120kVp images, DICE similarity coefficients > 0.90 for five organs contoured, and mean iodine concentrations within 0.2mg/mL of static images. For rigid sliding motion, the algorithm performed poorer and resulted in noticeable registration errors towards the superior and inferior scan extents due to the regularization constraint in the algorithm. The iodine concentration accuracy in the center of iodine rods was not substantially impacted by the registration errors for less severe motion, but local errors in areas of misregistration exceeded 100HU for 50keV monoenergetic images and 3mg/mL for iodine images.

Conclusion: This data supports the clinical use of the deformable registration algorithm for abdominal sites with limited motion (i.e., pancreas and liver). However, caution should be used for quantitative measurements in areas with sliding motion (i.e., thoracic spine near the diaphragm).

Keywords

Deformation, Registration, Dual-energy Imaging

Taxonomy

IM- CT: Dual Energy and Spectral

Contact Email