In some cases, the results of deformable image registration (DIR) won't meet your expectations. Though frustrating, you understand that inaccurate deformation could negatively impact future planning and treatment. In most clinical cases—aside from extreme anatomical differences between scans—accurate deformation is possible.
In a previous blog, we talked about accurately assessing the quality of DIR through a singular critical question:
Did the registration map the anatomy of interest on my previous scan to the exact same anatomy on my current scan?
If you’re uncertain when answering this question, your QA method for assessing DIR has failed. If you answer this question with a clear no, then you’re left asking another question.
What do I do now?
Physicists and dosimetrists must decide how to respond to and resolve erroneous DIR. The issue can be addressed in several ways, impacting accuracy, efficiency, and confidence. Let’s explore the four most common responses to failed deformation.
Clinicians may choose not to correct the failed deformation, either because of time limitations or a lack of confidence in their ability to improve it. Instead—not knowing if the DIR can be trusted—they will warn the treating physician to use caution with the deformed images.
This choice abandons deformable registration altogether. Since registration is still desired, rigid is chosen. And because clinicians typically perform deformable registration to align multiple areas of interest, they must now perform multiple rigid registrations.
Sometimes clinicians will choose to do both A and B. The combination of these options demands that the treating physician can keep track of which registrations their colleagues feel confident in. It also requires extra time to perform and QA multiple registrations.
Providing additional information to improve a failed deformation is a potential alternative. This requires the DIR system to have the capabilities to perform this process efficiently. The process must also improve the DIR reliably.
So, which option is best?
Which option creates the most accurate registration in the least amount of time? Looking closer, options A, B, and C are not ideal when time is precious and accuracy is essential. There are a number of safety and efficiency concerns when accepting questionable DIR, reverting to multiple rigid registrations, or a combination of the two.
Working with a questionable deformation creates unnecessary confusion and increases patient safety concerns. It may also cause clinicians to stop using or trusting deformation altogether.
Accepting questionable DIR or supporting it with rigid registrations (Options A and C) necessitates communication that fails to capture when and where the deformation shouldn’t be trusted. Even with a verbal or written disclaimer, the physician may not remember or accurately interpret the message, increasing the risk of mistreatment.
While it might seem more convenient to revert to rigid registrations, Option B does not adequately provide the treating physician with what they were looking for—yesterday’s anatomy mapped accurately to today’s anatomy on the planning image. In many clinical scenarios, rigid registration can lead to an inaccurate target volume.
When a DIR system is unable to fix errors, additional communication between a clinician and treating physician is required, and the possibility for misuse and miscommunication abounds. Patient safety concerns would be reduced by ensuring that only accurate deformable registrations reach the treating physician's desk.
Unclear communication and the need to perform multiple rigid registrations creates inefficiency.
When physicians are unsure whether or not they trust a deformable image registration, it adds unnecessary back-and-forth communication to clarify. Multiple rigid registrations create additional work for the dosimetrist and physicist in preparation, and they increase the review time for the physician. All of these scenarios reduce productivity.
If done correctly, DIR should be a more precise and efficient pathway to treatment planning than multiple rigid registrations. Providing only accurate deformable registrations to the treating physician ensures efficiency.
The optimal choice for dealing with a failure in DIR is to efficiently fix the deformation and confirm it correctly maps all anatomy of interest from one image to another (Option D).
Beware of any DIR QA system that suggests the ability to fix deformation but forces inefficiency and risk.
For example, those that require restarting the deformable registration process with a more satisfactory initial rigid registration. This wastes the time you’ve already spent, costs significant time with multiple attempts, and may not improve your outcome. Also, beware of a solution that suggests the correct response is to have the software exclude peripheral areas of the scan to try and improve the deformable registration in the areas you care about. This response may not improve the results where it matters. The right DIR QA system would enable you to fix those inaccurate areas reliably and efficiently.
The capability to fix a failure is integral to equipping your clinic for successful DIR. Finding the right assessment and correction tools allows DIR to fulfill its critical role in the treatment planning process.
Want to explore effective tools to efficiently and reliably fix failed DIR? Sign up for our upcoming webinar, “DIR QA in Clinical Practice.”
Jeff Kuhn is a Product Manager at MIM Software. Jeff works closely with clinics across the country to improve their pre-treatment processes such as image registration, contouring, 4D motion management, and plan assessment.
You can visit them online at
The use of AI technology has transformed the field of Radiation Oncology treatment planning. This...