Can You Trust Your Deformable Image Registration (DIR) Solution?

The key to trusting a DIR is assessing its accuracy—but not all image registration systems give you the tools to do this confidently. A guide by leading Radiation Oncology practitioners shows how image registration vendors compare.

Comparing Image Registration Software

To enable trust in a DIR, your image registration software must:

• Make it possible to perform quantitative quality assurance (QA) to answer the critical question: Did the registration map the anatomy of interest on my previous scan to the exact same anatomy on my current scan?
• Give you adequate tools to identify and address an inaccurate DIR.

How can you objectively compare the available image registration systems to see which one best meets the above two requirements?

A good place to start is this guide by a high-profile, multi-center NRG Oncology working group. The group graded four of the most popular image registration systems on their compliance with AAPM TG-132 guidelines for evaluating the accuracy of image registrations.

Which Image Registration Systems Comply with TG-132 Recommendations?

Availability of Quantitative Metrics for QA

The NRG group gave each software grades of 0 to 3 on its ability to provide each of the following five quantitative metrics for evaluating image registrations recommended by TG-132:

• Target registration error (TRE) — A point-based accuracy metric using implanted or naturally occurring landmarks visualized on a pair of images.
• Mean distance to agreement (MDA) — Mean surface distance between two contours on registered images.
• Dice similarity coefficient (DSC) — Volumetric overlap of two contours on registered images.
• Jacobian determinant — Volume expansion or contraction resulting from a deformable image registration.
• Consistency — Independence of an algorithm to the direction of the registration (image A to image B or image B to image A).

A grade of 0 for a metric indicates non-compliance, and a 3 indicates full compliance.

MIM^® scored a 3 for each of the five metrics. MIM was the only software to get the maximum grade in all categories. In fact, it was the only software to get the top grade in any category. This means that MIM is the only vendor tested that offers a compliant solution for all TG-132 recommended QA evaluation metrics with its base software.

Availability of Tools to Identify and Address an Inaccurate DIR

The NRG group also evaluated whether the image registration systems met the following TG-132 vendor recommendations with base software:

• Disclose basic components of their registration algorithm to ensure its proper use.
• Provide the ability to export the registration matrix or deformation vector field for validation.
• Provide tools to qualitatively evaluate the image registration.
• Provide the ability to identify landmarks on two images and calculate the TRE from the registration.
• Provide the ability to calculate the DSC and MDA between the contours defined on an image and the contours mapped to the image via image registration.
• Support the integration of a request and report system for image registration.

Just as MIM is the only vendor with top grades for all recommended quantitative metrics, MIM is also the only software that meets all of the above TG-132 vendor recommendations.

If you are a current MIM customer, get more information about all the tools in MIM that help you comply with AAPM TG-132 recommendations: Visit TG-132 Resources on the MIM Knowledge Center.