by Anne Clement and Dr. Ilya Gipp, GE HealthCare
Imaging has long been a cornerstone of radiation therapy, providing essential information for precise treatment planning and dose delivery. From the early days of X-ray imaging to today's advanced modalities like CT, PET, and MRI, imaging has consistently played a transformative role in radiation oncology. It enables clinicians to better visualize tumors, accurately identify healthy tissues, and effectively navigate and monitor the treatment process. Many breakthroughs in radiation oncology, like adaptive therapy techniques, dose escalation, and hypo-fractionation largely relate to the ability to precisely locate and assess the tumor. Better targeting, enabled via using advanced imaging techniques, improves the accuracy of radiation dose delivery, minimizing harm to surrounding healthy tissues, and reducing treatment-related toxicities.
The introduction of computed tomography in the 1970s revolutionized radiation therapy planning by providing imaging data of, at that time, unprecedented value. CT's ability to accurately visualize internal anatomy and offer precise attenuation measurements for dose planning quickly became indispensable [1]. Although it took over two decades for dedicated CT virtual simulators to be developed, their adoption was rapid once available. Shortly after, CT scanners almost entirely replaced conventional 2D simulators, highlighting the significant impact of the added value that this new standard of advanced imaging data brought to the field of radiation oncology.
Similarly, the introduction of cone-beam computed tomography (CBCT) equipped dose delivery systems in 2005, enabled image-guided radiotherapy to quickly become the mainstream standard of care in radiation therapy [2].
Both examples demonstrate that radiation oncology has always been wide open to embracing innovative approaches in leveraging new data for advancing cancer therapy.
Now, radiology, as the primary consumer of imaging scanners, is itself undergoing a significant transformation. This is because the field of medical image analysis has grown, driven by larger, more complex data sets, as well as the development of more advanced pattern recognition tools. These advancements are enabling more efficient extraction of quantitative features, transforming images into mineable data inputs for decision support – an area known as radiomics. Imaging data can now be combined with other patient information, processed to create models, potentially improving not only diagnostic but also disease predictive and therapy prognostic accuracy [3].
The concept of leveraging data-driven insights for enhanced dose planning is not new to radiation oncology. The dose painting technique was first proposed in the early 2000s as a strategy to improve the therapeutic index of radiation therapy by targeting specific sub-regions within tumors that may require higher doses due to their biological characteristics. The original idea was to use advanced imaging techniques to identify these regions and then deliver tailored radiation doses to them, enhancing treatment effectiveness while minimizing damage to surrounding healthy tissues [4].
Those techniques have not yet become mainstream in radiation oncology due to several challenges and existing limitations. Imaging methods have not reliably provided the necessary precision for identifying specific tumor sub-regions that require dose escalation. Additionally, the inherent heterogeneity of tumors means that different regions within the same tumor or affected healthy tissues can respond differently to radiation, substantially complicating the determination of optimal dose distributions and treatment planning in general. Technological constraints also played a noticeable role, as implementing dose painting required more advanced and precise dose planning and delivery systems that were not widely available. Furthermore, there has been a shortage of evidence and a need for more clinical validation to establish the effectiveness of dose painting in improving patient outcomes compared to conventional methods. Lastly, physiological variations during treatment can affect the accuracy of dose painting, making it challenging to maintain consistent treatment plans.
Despite these obstacles, ongoing research and technological advancements continue to explore ways to integrate data-driven radiation therapy guidance into routine clinical practice [5]. Recent advances in radiation oncology have significantly enhanced the field's capabilities. We can now more easily integrate functional quantitative data into radiation therapy planning and delivery, allowing for more personalized and effective treatments. Robust treatment planning systems, capable of performing more powerful calculations for optimizing dose distributions, have emerged. Additionally, the precision of dose delivery toolsets has substantially increased. Given these advancements, it's a fitting moment to reflect on where we stand today in radiation oncology's ability to better harness data insights – and, once again, on the pressing opportunity for imaging to further enhance the precision and effectiveness of radiation therapy, thus improving patient outcomes.
With that in mind, GE HealthCare organized a dedicated symposium entitled “Data-enhanced Strategies in Improving Radiation Therapy Outcomes,” gathering top leaders in the field. The initial objectives of the meeting were:
-
To examine the expanding role of data in enhancing precision and personalization.
Delving into how data-driven insights can predict patient responses, refine planning accuracy, implement adaptive delivery strategies, and optimize response assessment techniques. -
To address barriers and propose solutions for improved adoption.
Jointly with participants identifying challenges in data integration within both academic and community oncology settings and coming up with effective strategies to enhance adoption rates and practical implementation. -
To explore future opportunities for data-driven advancements in radiation therapy.
Highlighting emerging opportunities for leveraging data to improve radiation oncology outcomes, with a focus on innovative applications and long-term benefits.
Dr. Alejandro Berlin, clinician-scientist and director of research at the Radiation Medicine Program – Princess Margaret Cancer Centre, associate professor at the Department of Radiation Oncology at the University of Toronto, Canada
WHY this is the right time for radiation oncology to rethink and innovate
Dr. Alejandro Berlin, opened the meeting with a thought-provoking talk titled “Radiation Oncologists – Time to (Re)think How We Work?” He began by emphasizing the pervasive nature of data, stating, "everything can be turned, or is data." Dr. Berlin provided a high-level perspective on RT department workflows, aiming to challenge conventional thinking in radiotherapy and its integration into the patient journey, while urging better utilize the value of imaging data along the way.
Reflecting on his early career, he recalled skepticism about the future of radiation oncology, which underscored the need for the field to evolve and think 'outside of the bunker.' He highlighted the false dichotomy between radiology and radiation therapy, advocating for a more integrated approach founded on data-driven insights to improve patient outcomes while reducing inefficiencies. For such change to occur, Dr. Berlin argued, it is essential to focus outward. Rather than pursuing small, incremental improvements – such as minor reductions in planning target volume (PTV) or slight decreases in treatment planning time – he advocated for more transformative changes that, in his view, offer significantly higher returns on investment.
Dr. Alejandro Berlin continued his talk with an example of their study involving the use of diagnostic MR imaging data, naturally acquired in random positions, for the creation of 3D printed masks used for patient positioning and immobilization during radiation therapy. Masks’ accuracy (based on CT simulation and CBCT imaging) was then compared to the traditional thermoplastic masks. The study found that the 3D printed masks were reliable, with no fallback cases, and in some instances outperforming traditional masks. Additionally, patients reported better tolerance, particularly in the face and neck areas, due to the open-mouth design, which reduced claustrophobia.
Dr. Berlin emphasized that this approach created a separate pathway for using diagnostic imaging data not only for radiation therapy planning but also for the creation of patient-specific immobilization devices. This streamlined process could potentially reduce the number of required patient visits, avoid unnecessary CT and MR simulations, and enable an earlier start of treatment. Dr. Berlin mentioned the importance of diagnostic MRI quality assurance and setting specific imaging standards based on radiation oncology requirements. He also highlighted substantial work being done for other body parts (e.g. head and neck, thorax, pelvis) to utilize diagnostic CT imaging data, where trained models warp these scans to emulate simulation CT imaging. Achieving high similarity rates, at least for the first couple of fractions, after which online imaging could be used further for adaptive planning techniques.
He proposed that synthetic images from diagnostic scans could be used for anticipatory therapy planning prior to patient consultation. This would enable enhanced personalized patient care through same-day decision-making with individualized predictions of treatment outcomes. The existence of such framework and pipeline, Dr. Berlin anticipated, would unveil opportunities for seamless bi-directional learning through multi-modal data, translating not only from diagnostics to radiation therapy but also back to radiology, further elevating the use and value of data insights. Dr. Berlin stressed the importance of integrating diagnostic and therapeutic processes, arguing that the traditional separation is becoming a barrier for significant progress and a potential vulnerability for the fields. He called for a shift in thinking, emphasizing the need to operationalize new tools and bring in expertise from healthcare design, human factors, engineering, and behavioral science to truly transform workflows and meaningfully improve the patient’s journey.
Dr. Berlin's talk underscored the potential for innovation by embracing interdisciplinary collaboration between diagnostic imaging and radiation oncology, ultimately aiming to improve patient outcomes and streamline treatment processes.
Dr. Dave Fuller is a distinguished professor in the Division of Radiation Oncology at The University of Texas MD Anderson Cancer Center, Houston, TX
HOW can advanced imaging data insights improve radiation oncology strategies
Dr. Dave Fuller delivered an enlightening presentation titled: "Uncertainty Estimation: The Head and Neck Cancer - Use Case for AI-enabled Image-Aware Decisional Support" on the transformative potential of integrating imaging data into intelligent deep-learning systems for radiation therapy planning in head and neck cancers.
He began by addressing the critical issue of trust in AI systems within the radiation oncology community. He emphasized the need for decision support tools that leverage imaging data inputs to build trust and ensure scalability for easier and broader adoption. He then delved into the complexities of decision-making in head and neck cancer treatment, highlighting the availability of multiple advanced treatment techniques. Dr. Fuller pointed out the risks associated with incorrect therapy modality selection, which can lead to unnecessary harm to the patient.
Dr. Fuller also discussed the benefits of additional data insights, such as multi-parametric MRI and multi-modal imaging information, in improving AI-based segmentation capabilities. While AI can match or exceed human performance in these tasks, adding significant value to the characterization of volumes and structures, there is still a need for more clinical evidence in converting this additional information into treatment decision-making and radiation therapy planning.
Following that Dr. Fuller discussed the development of risk estimation and prediction tools designed to aid clinicians in deciding between therapy types based on imaging data. Depending on whether the goal is to optimize short-term subjective patient-reported outcomes, or functional more objective outcomes, clinicians can select a risk threshold and receive actionable recommendations. This provides evidence for the selection between radiation or surgery, which is already possible with existing models today and should lead to more accurate risk predictions through better utilization of imaging data.
Dr. Fuller shared results from testing the tool with surgeons, neuroradiologists, and head and neck radiation oncologists who were asked to estimate the probability of a patient developing extracapsular extension based on provided images. The results showed that human estimations were almost as random as flipping a coin, highlighting the potential harm to patients due to the lack of certitude. On the other hand, the model from the ECOG 3311 clinical trial and other studies, using deep learning systems with high-quality CT data, can enable high accuracy prediction, significantly outperforming human observers.
The concept of uncertainty quantification was a major highlight of Dr. Fuller's presentation. He explained how this approach allows for more informed and confident decision-making by quantifying uncertainties in various components of the treatment planning process. He provided examples from several studies, demonstrating how this and similar approaches can predict treatment failures and guide therapeutic decisions with greater confidence.
While radiation oncologists today are forced to make binary decisions, such as determining whether a voxel in an image represents a tumor or not, the shift from binary decision-making to a more nuanced, quantized approach has significant potential to improve clinical outcomes. Dr. Fuller underscored the immense potential of AI-enabled, image-aware decisional support systems, which are capable of handling vast amounts of image feature information, in revolutionizing radiation therapy. By using imaging data insights to quantify uncertainties, these systems open possibilities for more accurate and actionable integration, ultimately improving patient outcomes and reducing the risks associated with traditional binary decision-making.
Dr. Arash Naghavi, Associate Professor at the Department of Radiation Oncology at H. Lee Moffitt Cancer Center, Tampa, FL
WHAT can the integration of data-driven insights do for radiation oncology outcomes
Dr. Arash Naghavi, gave a remarkable talk titled “Personalized Radiotherapy in Soft Tissue Sarcoma.” He began by emphasizing the unique heterogeneity of characteristics within tumors, particularly sarcomas, which makes them “ideal” for developing personalized medicine approaches.
Dr. Naghavi discussed their HEAT (Habitat Escalated Adaptive Therapy) phase II trial, which uses radiomic habitat features to identify aggressive areas within tumors. The process that involves advanced computer based quantitative imaging analysis, providing a radiographic representation of tumor biology. This allows for more precise targeting of radiation treatment.
Dr. Naghavi referenced their previously published data on Genomic Adjusted Radiation Dose (GARD), which evaluated a gene expression signature for predicting radiation sensitivity in sarcomas. This approach allows for tailoring radiation doses based on the tumor's genomic profile, thereby improving treatment efficacy. Dr. Naghavi highlighted the success of dose escalation using the Simultaneous Integrated Boost (SIB) technique, which aims to improve local control in sarcomas. Achieving a favorable pathologic response, defined as 95% or more necrosis before surgery, was associated with improved outcomes, including better R0 resection rates, local control, distant control, and potentially progression-free survival and overall survival.
Dr. Naghavi also discussed the NBTXR3 trial, which showed that injecting hafnium particles doubled the rate of achieving 95% necrosis, significantly improving R0 (no residual tumor) resection rates. This trial demonstrated the combined potential of novel therapy agents and radiation to enhance treatment outcomes. Estimating the potential for personalized radiation therapy planning using genomic-radiomic correlation based on sarcoma habitat information, Dr. Naghavi used NBTXR3 results of 8% pathologic response of 95% necrosis as a baseline and set up a HEAT trial with an ambitious goal of tripling that.
To accomplish this, Dr. Naghavi explained the use of a dedicated MR-imaging protocol with DWI ADC data, then adding dynamic 4D contrast-enhanced imaging to measure the amount of contrast taken up by each voxel, its peak, and the speed of washing out. These data-driven quantitative and objective insights, together with using K-trans for identifying a surrogate of vascularity, allowed them to determine which areas of the tumor are cell-dense but not very vascular (presumed to be hypoxic), which areas are cell-dense and well-oxygenated, and which areas are mostly fluid and thus do not require dose escalation.
Dr. Naghavi emphasized the importance of precise image-guided, rather than random, biopsies for correlating imaging data with tissue characteristics. Precise mapping of the tumor habitats will enable the assessment of which parts are contributing to circulating tumor DNA (ctDNA). As the tumor responds to treatment, the expected drop in ctDNA can be correlated. For this purpose, liquid biopsies were also performed at the time of MR-imaging to evaluate ctDNA, providing insights into follow-up control benefits and helping to understand the mechanisms behind improved survival rates.
Looking to the future, Dr. Naghavi emphasized the potential of genomic-radiomic correlation, using imaging and genomic data to guide treatment planning decisions. He also highlighted the role of machine learning in enhancing personalized medicine approaches for not just sarcomas. By integrating genomic and radiomic data, clinicians can better assess the biology of tumors and tailor treatments accordingly. The HEAT trial exceeded expectations with the highest pathologic response rates reported in the literature, showcasing the promise of personalized treatment through advanced technology.
Dr. Naghavi's talk underscored the innovative approaches being taken to improve outcomes for patients through personalized radiotherapy based on precise evaluation of tumor habitat characterization. He emphasized the use of imaging, therapy planning, and treatment delivery technologies to push the field forward and further tailor treatments, ultimately aiming to generalize these approaches for other disease sites. The collaborative efforts and technological advancements presented in his talk highlight the ongoing progress in the field of radiation oncology and the potential for personalized medicine to revolutionize cancer treatment
In his talk, Dr. Naghavi referred to a more detailed published report on the HEAT trial [6].
In conclusion, the integration of advanced imaging data insights into radiation oncology holds immense potential for transforming cancer treatment. As demonstrated by the innovative approaches presented at the symposium and discussed in this paper, leveraging data-driven strategies can significantly enhance precision, personalization, and overall treatment efficacy. The HEAT trial and other studies underscore the importance of combining imaging and radiomic data to better understand tumor characteristics and tailor radiation therapy for improved outcomes.
The collaborative efforts and technological advancements presented, highlight the ongoing progress in the field of radiation oncology. By embracing interdisciplinary collaboration and integrating diagnostic and therapeutic processes, we can push the boundaries of personalized medicine and improve cancer patient outcomes. The future of radiation oncology seems to lie in harnessing the power of data-driven insights to further refine personalized planning accuracy, implement adaptive delivery strategies, and optimize response assessment techniques.
As we continue to explore and adopt these data-enhanced strategies, the potential of imaging to revolutionize cancer treatment becomes increasingly evident. The commitment to innovation and the pursuit of more effective, personalized therapies will undoubtedly shape the future of radiation oncology, offering hope and improved outcomes for patients worldwide.
Please watch the full Symposium Video recording - click HERE.
References:
1. Dobbs HJ, Parker RP. The respective roles of the simulator and computed tomography in radiotherapy planning: a review Clin Radiol 1984; 35:433–9
2. Bujold A, Craig T, Jaffray D, Dawson LA. Image-guided radiotherapy: has it influenced patient outcomes? Semin Radiat Oncol. 2012 Jan;22(1):50-61
3. Gillies RJ, Kinahan PE, Hricak H. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016 Feb;278(2):563-77
4. Hall, Eric J. Dose-painting by numbers: a feasible approach? The Lancet Oncology, Volume 6, Issue 2, 66
5. Peng H, Deng J, Jiang S and Timmerman R (2024) Rethinking the potential role of dose painting in personalized ultra-fractionated stereotactic adaptive radiotherapy. Front. Oncol. 14:1357790
6. Naghavi, A.O., Bryant, J.M., Kim, Y. et al. Habitat escalated adaptive therapy (HEAT): a phase 2 trial utilizing radiomic habitat-directed and genomic-adjusted radiation dose (GARD) optimization for high-grade soft tissue sarcoma. BMC Cancer 24, 437 (2024)
