The emerging field of radiomics—the extraction and analysis of “hidden” data from large amounts of medical images—could help doctors more accurately choose the precise radiation therapy doses necessary to kill tumors, according to a study presented today at the 56th Annual Meeting of the American Association of Physicists in Medicine in Austin, Texas.
Currently, there is a one-size-fits-all process for selecting radiation therapy doses, which might be too much for some patients and not enough for others. Radiomics will help us know when we can turn down the treatment intensity with confidence, knowing we can still control the disease.
The researchers analyzed PET scans of 163 non–small cell lung cancer patients and 174 head and neck cancer patients taken before and after treatment. From each scan, they extracted a variety of information about the tumor, including the intensity value of the PET image, the roughness of the image, and the roundness of the tumor. They matched the information in the “before” and “after” scans with information on how the patients fared, including whether the tumor shrank and how long the patient survived. That gave them models to direct future therapy. For example, the researchers found that lung tumors that have a higher uptake of radiotracer need a higher than usual dose of radiation.
“Standard protocol today is to only use PET imaging to define the extent of a tumor to be treated,” Dr. Deasy said. “Based on the information from this study, the data would be extracted from those images and put into models that would tell the physician what dose was required to kill the tumor with a high probability.”
We suspect that we’ll get a flood of new research in the next few years as Dr. Deasy, his colleagues, and others explore the intriguing possibilities of radiomics.
* * *
Cutting physician payments is a popular reaction to increasing health care costs. But one physician argues that such a strategy actually increases costs. For his reasoning, see our Facebook page.
Related CME seminar (up to 20 AMA PRA Category 1 credits™): UCSF Practical Body Imaging