Warning! OUTDATED BROWSER DETECTED!   Please update your browser immediately for a better experience on this website. Learn More
  • Precision Medicine and Cancer Treatment Response

    Researchers are actively exploring the promise of radiomics as a better way to predict response to cancer therapy. By Mike Bassett

    August 1, 2017

    One of the key components of precision medicine — particularly in oncology — is gauging a patient’s response to treatment.


    Increasingly, researchers are beginning to understand that genomic heterogeneity exists among — and within — tumors, and that these differences profoundly impact how a patient will respond to treatment with a particular therapy.

    Consequently, as pointed out in a recent study in Precision and Future Medicine, “the success of precision medicine requires a clear understanding of each patient’s tumoral heterogeneity and individual situation.”

    With this in mind, the evolving fields of radiomics and radiogenomics (also called imaging genomics) have garnered strong academic interest. Radiomics is the high-throughput extraction of quantifiable imaging characteristics and features from modalities such as CT, MRI and PET which are converted into high-dimensional data. The mining of this data to detect correlations with genomic patterns is known as radiogenomics.

    “Studies show that there is quantitative information inherent in all of these images that is very important,” said Evis Sala, MD, PhD, chief of Body Imaging Service at Memorial Sloan Kettering Cancer Center and a professor of radiology at Weill Cornell Medical College, both in New York City. “And we can use quantitative information to actually describe heterogeneity.

    “I don’t think imaging can replace genomics,” she added. “But it should be integrated with genomic and histologic information. Ultimately, synthesizing all of this information will lead to potentially high-powered precision medicine. If we are actually able to devise an integrated system using this genomic, imaging and clinical information, we will be able to treat patients in a more personalized way.”

    Getting to that point is the challenge researchers now face.

    “Radiomics is difficult,” said Anthony Shields, MD, PhD, professor of oncology and medicine, Wayne State University, and associate director for clinical sciences, Karmanos Cancer Institute, both in Detroit. “The platforms are not yet standard and the way the images are obtained are not necessarily standard, so that will affect the results.

    “It’s an exciting field, but I’m not making any clinical judgments based on radiomics at the moment,” he said. “It’s certainly an area of very active research, and we are hoping that it will lead to improved patient assessment and that it will provide early responses and better information. But it’s going to take us a while to get there.”

    Predicting Tumor Response

    Nevertheless, Dr. Sala pointed to important work that has shown the potential power of these developing fields. For example, she referred to a 2014 study in Nature Communications in which the researchers described the radiomics analysis of CT scans of more than 1,000 patients with lung and head and neck cancers.

    The researchers found that certain tumor characteristics formed a signature that was able to capture intra-tumor heterogeneity and could predict how the tumor would react in individual patients.

    “These data suggest that radiomics identifies a general prognostic phenotype existing in both lung and head and neck cancer,” the authors wrote. “This may have a clinical impact as imaging is routinely used in clinical practice, providing an unprecedented opportunity to improve decision support in cancer treatment at low cost.”

    Dr. Sala also noted research exploring the promise of radiomics as a better way to predict response to therapy — an area where she envisions radiomics and radiogenomics making a huge difference.

    “What is going to be more crucial than the distinction between a benign or a malignant tumor, or the aggressiveness of the cancer, is actually predicting resistance to treatment,” she said. “Because that is really the holy grail.”

    Currently, if a patient is resistant to treatment it usually means that the cancer has reached the metastatic stage. “And you can’t turn your patient into a pin cushion — you can’t biopsy every single lesion you find,” she said. “But you can image every single metastasis and mine the quantitative data and possibly predict treatment response, and predict resistance and pseudo-progression.

    “This is all extremely important, and that is actually where I see the added value in radiomics and radiogenomics,” she said.

    Dr. Sala also believes that radiomics and radiogenomics will help with immunotherapy and the concept of pseudo-progression, in which a tumor appears to be progressing after the start of immunotherapy, but is actually in a state preceding a prolonged response to the therapy.

    Even with pseudo-progression, an oncologist may continuetreatment for weeks to months to see whether a tumor is truly progressing, Dr. Sala said. “That’s a long time to keep someone on immunotherapy, because it’s expensive and potentially toxic,” she added.

    For that reason, researchers are seeking better ways to measure treatment resistance or pseudo-progression in the case of immunotherapy, she said.

    Tumors exhibit genomic and phenotypic heterogeneity, which has a prognostic significance and could influence response to treatment.

    There are differences within the tumor — with blood flow, metabolism and cell density — that can be probed with imaging, she said. “With imaging not only can we look at intra-tumor heterogeneity within one site, but we can also observe the differences between different sites. This is important, because during this resistance to treatment, various sites can become more heterogeneous, and those may be the ones that can drive resistance.”

    So the potential is there to non-invasively quantify this heterogeneity within and between tumors in individual patients in order to select more effective therapies and inhibit treatment resistance, she said.

    Progress Depends on Cooperation among Specialties

    Whether radiomics and imaging genomics will replace the current cancer staging system — tumor, node, metastasis (TNM) — which codifies the anatomic extent of disease at diagnosis in order to assess the prognosis of a patient’s cancer, is still unknown.

    “While TNM staging is not yet a thing of the past, we can add more than TNM coding with radiomics and radiogenomics — and I think the data is emerging on that,” Dr. Sala said.

    Dr. Shields agrees that more work needs to be done before TNM coding becomes obsolete.

    “Our knowledge of the molecular biology of these tumors — even with gene panels and proteomics — is still not up to subdividing patients without knowing the TNM coding of the tumor,” Dr. Shields said. “Knowing that it’s a stage 1 tumor is still very different than knowing it’s a stage 4 tumor.”

    He recalled a presentation in which he listed a number of factors that helped predict the patient’s prognosis, including details such as histology, weight loss, whether the patient had fevers and performance status.

    “For the cost of one PET scan, you could get 20 different prognostic indicators that are really very helpful in defining what is going to happen with the patient,” he said. “So the molecular phenotype adds to that, as does the imaging. But the clinical element — just walking into a room and looking at a patient and noticing that he really looks sick —still has a role.”

    Dr. Sala said that further progress in radiomics and imaging genomics will depend upon a certain degree of cooperation between radiologists, oncologists, molecular biologists and computer scientists.

    “Unless we bring them all together we won’t be making any strides forward,” she said. “Oncologists have to be imaging friendly, while radiologists have to understand that they have to use artificial intelligence and deep learning. It all has to be integrated.”