How High-quality Image Data Enables Adaptive Radiotherapy

Author: Ottar Schmitz, Senior Director of External Innovation at RefleXion

I will never forget my first factory tour at RefleXion. It was March 15, 2020, just as the pandemic was setting in, and I fell instantly in love with the X1 with its gantry exposed. When I first learned of RefleXion, my reaction was, “They are using a PET/CT to guide treatment, not that revolutionary to incorporate a PET image during treatment planning for contouring, but I’ll check it out.”  What I did not understand at the time was that RefleXion uses PET data in real time and that the photon emission generated was used to trigger radiation delivery from the LINAC in less than a second via a gantry spinning at 60 RPM. Seeing this marvelous technology, fully exposed, I immediately recognized that real-time PET addresses many of the day-of-treatment issues of patient motion and minor anatomical tumor shifts. The fundamental concept is that the machine calculates a definitive location of where it needs to deliver a radiation beam and does it in near-real time.

Another epiphany is that by spinning at 60 RPM the X1 also enables the onboard 16-slice, fan-beam CT to generate amazing images not commonly associated with a radiotherapy system.

Another epiphany is that by spinning at 60 RPM the X1 also enables the onboard 16-slice, fan-beam CT to generate amazing images not commonly associated with a radiotherapy system. Much of my career in medical imaging has been centered on maximizing image quality while minimizing the effort to create or interpret medical images, through software techniques. However, all software approaches benefit from the best possible image acquisition. The longstanding challenge in radiation therapy has always been understanding the target location at any given time. Daily imaging brings so many advantages to true up information between the planning CT and the image of the day.

A 16-slice fan-beam CT, such as the one installed on the X1, is perfect for generating images just prior to treatment. The superb image quality provides extra confidence to the physician that patient position matches the treatment plan, without the need for excessive post-processing to enhance images. That said, patient positioning is just one of the many benefits of having high quality CT imaging available on the day of treatment.

Medical images are data, with each pixel of an image representing a bit of information. Like all things in life, radiotherapy benefits from the most up-to-date information available assuming it is as accurate as possible.  Artificial intelligence (AI) has made tremendous strides in filling in missing data to make images appear cleaner and less fragmented.  However, there are concerns with this approach in a clinical setting. Indeed, a physician once asked when reviewing a heavily manipulated image, “How do I know that what I’m looking at isn’t a cartoon?”  If too much information is missing, confidence in AI naturally starts to falter, leaving clinicians to wonder, “Is what I see on the image actually there?” Not a great basis from which to operate in the setting of applying ablative radiation to a cancer patient.

Quantitative analysis, clinical measurements/evaluations, delta radiomics – these all benefit from high-quality daily imaging. In clinical comparison, the X1 images demonstrated excellent image stability and reproducibility^[1], opening the possibility of inter-fraction analysis of the tumor’s response at a radiomic level. Furthermore, autocontouring works better on a fan-beam CT than a cone-beam (CBCT)^[2] given the high-quality image data. Fundamentally, the better the underlying quality of the image data, the faster and more accurate results the algorithms will provide.^[3]

Where does this all lead?  Speed, comfort, and the potential to one day adapt treatment plans based on the most up-to-date information. For adaptive therapy to be clinically reasonable, speed and accuracy are paramount. Cancer patients can only be expected to endure lying on a flat table for so long, so speed is of the essence. Like the adage of Occam’s Razor, when it comes to image quality in radiotherapy, the best technology is often the simplest: if the underlying image is good, all things flow more smoothly.

[1] Ketcherside T, Shi C, Chen Q, et al. Evaluation of repeatability and reproducibility of radiomic features produced by the fan-beam kV-CT on a novel ring gantry-based PET/CT linear accelerator. Med Phys. 2023; 50: 3719–3725. https://doi.org/10.1002/mp.16399

Han, C. et al. Reproducibility and Repeatability of Pelvic Radiomics Features with Daily Imaging on a Novel Biology-Guided Radiotherapy Machine Compared to Daily Imaging on Other Radiotherapy Delivery Systems

International Journal of Radiation Oncology, Biology, Physics, Volume 117, Issue 2, e670

[2] Han C, et al. Comparison of AI-Based Auto-Segmentation Quality with Different Daily IGRT Imaging Modalities for Adaptive Radiotherapy Treatment Planning International Journal of Radiation Oncology, Biology, Physics, Volume 117, Issue 2, e670 https://doi.org/10.1016/j.ijrobp.2023.06.2116

[3]ibid