For medical physicists considering surface-guided radiation therapy… here’s a thought on the value of a deformable registration algorithm

Surface-Guided Radiation Therapy (SGRT) systems work by continuously collecting 3D images of a patient’s surface, typically the patient’s skin, and using this information to accurately identify a tumor position within the patient. Using the current 3D surface image, the system calculates the current patient tumor position relative to the tumor position from a reference 3D surface image. During the delivery of a radiation therapy treatment, the SGRT system is monitoring and verifying that the tumor is in the correct treatment position several times per second using data collected hundreds of times per second.

The simplest method to calculate changes in the patient’s position would be to assume that the patient is a “rigid body” and use a rigid, non-deformable registration algorithm. If the patient were a rigid body, you could shift and rotate the patient and the patient’s 3D image would exactly match.

Unfortunately, patients are not rigid. Changes in weight, changes in posture, breathing, contracting muscles, and relaxing muscles will cause the surface of a patient to change. Because the patient’s skin stretches and distorts (referred to as being elastic), a non-deformable algorithm may not deliver the best results for clinical use. And if the algorithm cannot calculate an accurate patient position, a new patient reference 3D image will need to be collected.

The most common technique used to adapt a non-deformable registration algorithm for use in the clinic is to only monitor a small subset of a patient’s position. Called a region of interest, a small subset of the patient’s surface cannot be relied upon to accurately calculate a patient’s change in position when the patient’s posture is also changing. Often clinicians will need to adjust the region of interest to allow a treatment to continue or be forced to create a new surface image to account for patient changes.

An effective solution for registering two, 3D images is to use a non-rigid, deformable registration algorithm. Deformable registration algorithms consider changes to elastic surfaces, such as a patient’s skin, and very reliably calculate the relative position and orientation of a patient’s current 3D image compared to the reference 3D image. One of the biggest advantages of using deformable registration is that clinicians do not need to take the time to create and repeatedly correct the region of interest used during the patient’s multiple treatment sessions.

One of the key advantages of the deformable algorithm unique to C-RAD Catalyst is that the system can identify specific areas of the patient that are out of position and are relevant to the position of the tumor being treated. For example, if a patient’s shoulder is slouched, the algorithm may calculate that the tumor is not in the correct treatment position. The SGRT system will identify the local area out of position and use color projections to guide a therapist to correct the patient’s posture.

Developing a technical solution to capture the patient’s surface image and calculate the relative tumor position in real-time, while being precise and reliable for use in radiation therapy is challenging. To overcome this hurdle, C-RAD has developed a surface imaging camera system that captures hundreds of images per second, which includes a deformable registration algorithm that calculates the tumor treatment position several times per second. The C-RAD solution both is fast and accurate and can be used clinically to reduce treatment times and expedite workflow.