Mayo Clinic: Perfusion CT radiation dose may be halved without loss of image quality

A new protocol developed by the Mayo Clinic for reducing radiation exposure during diagnostic perfusion computed tomography (CT) brain scans could have a huge impact on clinical practice.

The research comes at a time when several hospitals and a CT scan manufacturer are pointing fingers at each other following reports that hundreds of  patients were exposed to radiation overdoses during CT brain perfusion scans. The Food and Drug Administration has launched an  investigation to determine where things went wrong. The FDA has also called on hospitals with CT scanners to check their settings. However, in many cases, radiation overdoses have been attributed to poorly trained technicians and gaps in hospital safety protocol.

Researchers at the Mayo Clinic, who stand by the technology's clinical value in diagnosing stroke and cancer and its potential for heart disease diagnosis, say it’s possible to reduce the amount of radiation used in CT brain perfusion scans without compromising image quality. “We believe in the clinical value of perfusion CT, and though there is no documented risk of injury at the currently prescribed radiation levels, we are trying to lower the dose for the benefit of patients,” said researcher Cynthia McCollough, professor of radiologic physics at the Mayo Clinic.

McCollough and colleagues at the Mayo created an image-processing algorithm that gives radiologists the information they need for diagnosis using up to 20 times less radiation, depending on the diagnostic application. The research was presented at the AAPM annual meeting in Philadelphia July 20.

Depending on the diagnostic application, a perfusion CT exam is preceded by injection of a contrast agent such as iodine. It takes about 30 seconds to scan the tissue 20 or 30 times, each scan at a low dose. The images detect changes in blood volume and flow that may reveal injuries to vessels or tumor response to treatment. Information from each consecutive scan is then digitally cross-referenced with other images taken during the exam to improve image quality and reduce distortions.

The new adaptive algorithm compares the scans and can differentiate between anatomical regions that do not change from moment to moment and those regions that carry the contrast agent -- effectively reducing image noise while preserving iodine signal. The quality of each scan improves through non-linear comparisons with scans acquired earlier and later in the exam.

"When we use very low doses, the noise gets so high that it's hard to tell what you are seeing," said lead study author Juan Carlos Ramirez Giraldo, Mayo Clinic. "With this algorithm, we're trying to maintain both the image quality, so that a doctor can recognize the anatomic structures, and the functional information, which is conveyed by analyzing the flow of the contrast agent over the many low dose scans."

The new perfusion CT algorithms have proved effective in animal models, and the researchers reported in Philadelphia that they have begun to process data from clinical brain perfusion CT exams in patients. "We're up to 15 or 20 cases that we've shown to the docs, and they're all giving us the thumbs up," said McCollough.

According to neuroradiologist David DeLone, Mayo Clinic, “Patients aren’t aware that anything has changed, and as radiologists looking at a study, we don’t know anything has changed. Yet, we are obtaining high-quality images, more consistently and in shorter times, while exposing patients to about half of the radiation dose. It’s a win-win for everyone.”

The Mayo Clinic has recently adopted another significant change regarding routine CT head scans. The protocol, developed by radiologists and medical physicists, calls for reducing head CT radiation doses to 38 mGy, almost 50 percent less than the 75 mGy safety limit set by the American College of Radiology.

Research abstract:

20-Fold Dose Reduction Using a Gradient Adaptive Bilateral Filter: Demonstration Using in Vivo Animal Perfusion CT (pdf) by J Ramirez Giraldo et al.