Unencrypted login | home

Program Information

A Multi-Modality Image Atlas for Transforming An MR Image Into a Pseudo CT Image for MRI-Based IGRT Application


L Zhang

L Zhang*, K Michel, J Yang, A Rubinstein, C Kingsley, J Delacerda, P Balter, L Court, MD Anderson Cancer Center, Houston, TX

SU-E-J-34 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: IGRT tools are most useful when registering similar datasets. In particular it is difficult to automatically register reference MR images with treatment CT or CBCT images. We proposed to create a pseudo CT from the reference MR images and to register the pseudo CT to the treatment CT.

Methods: Two mice were sacrificed and immobilized on a rigid platform, and then imaged with small animal T1 and T2 weighted MRI and CBCT. Data from the first mouse was to develop an atlas and data from the 2nd mouse was to validate the system. To create a pseudo CT from an MRI, we first established the intensity conversion curve between the atlas CBCT and mouse MRI. This intensity conversion curve maps the intensity of fat tissue and soft tissue of the MRI onto the atlas CBCT intensity of the same tissue type. We used this conversion curve to create an intensity-corrected MRI. Next, using global affine registration, local (bone-by-bone) affine registration, and deformable registration, we calculated the deformation field linking bony regions in the atlas MRI and the validation MRI. This deformation field was used to deform the CBCT image inside the atlas bone contour and to replace that section of the image in the validation mouse MRI.

Results: The pseudo CT and validation CBCT are visually similar. Soft tissue in the pseudo CT is more recognizable and differentiable than in the validation CBCT. Contouring the bone structure on both images, the Dice index of the two bone structures was found to be 75%. The cross correlation of the image contents inside the fat-soft tissue region between the pseudo CT and validation CBCT was 0.85.

Conclusion: A novel approach was developed to create pseudo CT images from MR images that has potential applications in both small animal and human IGRT.


Contact Email: