Abstract
Deformations can be digitally applied to patient images to evaluate deformable image registration (DIR). However, this can deform image noise and artifacts, leaving a fingerprint of the underlying deformation that could skew accuracy determination. Image processing can erase this fingerprint and create a more realistic DIR test scenario. The importance of image processing to simulated deformations is tested here. These tests utilize a virtual pelvic phantom. Two image-filtering techniques are utilized: a spatial convolution with a Gaussian (SCG) and an edge-preserving filter (EPF) that preferentially removes the Fourier components associated with noise. Four different processing scenarios are evaluated. The first is no processing (NP). For the second, noise is added to the set of test images without filtering (NF). The third and fourth scenarios add noise after applying the SCG and EPF filtering methods. EPF provides the most realistic test scenario. These scenarios are evaluated for their effect on the spatial accuracy of the DIR algorithms from MIM Software and Velocity Medical Solution. For NP, NF, EPF, and SCG, the mean errors from MIM were 0.78, 1.11, 1.24, and 1.44 mm, respectively. Velocity was insensitive to these changes. An objective evaluation of DIR accuracy should include realistic noise scenarios.
Keywords: Deformable image registration, image artifacts, image processing, noise, quality assurance, simulated deformations.
5