By Aine Cryts

The U.S. FDA approved the use of immunotherapy to treat bladder cancer in 1990. In 2015, the federal agency approved an immunotherapy treatment for lung cancer, which is the second most common cancer among men and women in the United States, according to the American Cancer Society. In addition, the federal agency has approved the use of immunotherapy for the treatment of breast cancer, leukemia, melanoma, and other cancers.

One of the problems is that immunotherapy doesn’t work for every patient. “As promising as immunotherapy has proven thus far, only a minority of patients responded to checkpoint inhibitor drugs, and although some of the responses are long-lasting, the cancer often progresses again,” according to Boston’s Dana Farber Cancer Institute.

And then there’s the cost. Immunotherapy treatment can cost anywhere from $200,000 to $250,000 per patient, per year, Anant Madabhushi, PhD, professor of biomedical engineering at Cleveland’s Case Western Reserve University, tells AXIS Imaging News. Madabhushi also teaches in the university’s radiation oncology and radiology departments.

He co-authored a study in Cancer Immunotherapy Research, which revealed that artificial intelligence (AI) can be used by radiologists to validate the success of immunotherapy treatment on patients. AXIS recently discussed with Madabhushi the potential impact of this work.

AXIS Imaging News: Please discuss the challenges radiologists have when visually assessing a patient’s response to immunotherapy.

Anant Madabhushi: Immunotherapy engenders an immune response, which can be seen in the tumor. What looks like tumor expansion is essentially the body’s immune system kicking into high gear and attacking the tumor. But it looks on a CT scan like the tumor is increasing in size. That’s a major challenge for radiologists who use RECIST to assess changes in tumor size and assess treatment response.

AXIS: Why is this important?

Madabhushi: Radiologists and clinicians need to be able to distinguish this pseudoprogression versus true progression of the disease, where the disease is actually growing in size and not responding to the therapy. And, of course, if a clinician pulls the patient off immunotherapy when they have pseudoprogression, they’ve done the patient a disservice because they were on a therapy that was working.

AXIS: How can AI help?

Madabhushi: With our AI algorithm, we were able to detect a change in the texture, a change in the heterogeneity of the cancer nodule. We didn’t focus on the size of the tumor. For our study, we looked at a series of 140 patients from three different sites.

We could then determine which patients truly responded to the treatment versus those who didn’t respond. We monitored the changes in the texture and the heterogeneity of the tumors and surrounding milieu because we appreciated that a lot of the pseudoprogression is really a reflection of the body’s immune response.

AXIS: What’s next? When do you expect to have this tool ready for radiologists?

Madabhushi: We’re not at the point where this AI tool is clinically deployed. We’re looking at a three-year time window for that.

Before we get to clinical deployment, we’ll need to continue to further validate the AI algorithm. We may even do some prospective validation. All the data in our study was based on patients who previously had been treated, where we knew the outcomes. After further validation, we’ll need to figure out a way to overlay these measurements directly on top of the scans that radiologists are looking at in their PACS workstations.

Aine Cryts is a contributing writer for AXIS Imaging News.