2020 Vision — A Look at the Latest Trends and Best Practices in Dose Safety and Image Acquisition
By Keith Loria
Vol. 21 No. 2 P. 10
Dose-related discussions traditionally revolve around the realm of CT and X-ray with continuing efforts made to reduce the amount of ionizing radiation, such as by lowering peak kilovoltage while preserving image quality. Mahadevappa Mahesh, MS, PhD, FAAPM, FACR, FACMP, FSCCT, FIOMP, chair of the ACR Commission on Medical Physics, says much has changed in the past 10 years, driven by regulatory requirements, social media campaigns, and patient demand.
“There is a dose consciousness in all hospitals and image facilities, which is having a big impact,” Mahesh says. “What is helping is technology, how the technology is used, and the awareness of dose safety.”
For example, in CT, which was once responsible for almost one-half of all medical radiation exposure to the population, dose modulation is now on most equipment, allowing scanners to automatically lower or increase dose based on the patient. There has also been a growing trend of MRI utilization, due to its lack of ionizing radiation and much-improved soft tissue contrast resolution over CT.
In CT and fluoroscopy, Mahesh says scanners must exhibit a dose display so dose information is available when a patient undergoes a scan. Dose calculations—with regard to radiation dose and contrast dose—are based on patient weight, so there are differences between calculating pediatric dose and adult dose. He says the process is the same for pediatric or adult patients, but dose display information must be corrected for patient size and, by doing so, it can accurately estimate the dose for pediatric patients to ensure it is lower.
“You must track patient dose; it’s a requirement,” Mahesh says. “That has opened up the whole field of dose management software, which is a new trend we are seeing throughout the industry.”
For most imaging facilities, specific protocols are in place to always minimize dose while preserving image quality, particularly in the pediatric population. It’s important for each imaging facility to know how to structure and automate low-dose protocols. This can best be accomplished by following ACR accreditation guidelines for all CT scanners.
“The ACR supports the ‘as low as reasonably achievable’ concept, which urges providers to use the minimum level of radiation needed to achieve the necessary results for imaging exams,” says Suzie Bash, MD, medical director of neuroradiology at San Fernando Valley Interventional Radiology at RADNET. “ACR is also a participant in the Image Gently campaign for dose reduction in pediatric imaging, as well as Image Wisely, which is an adult radiation dose–reduction effort.”
Additionally, the ACR has established a dose registry and encourages all providers to take part in this registry, an effort which the FDA also supports.
“Here at RADNET, we standardize our CT protocols across all of our 340 imaging centers to ensure optimized low-dose protocols for all of our exams,” Bash says. “All of our CT reports list the total dose length product in mGy-cm and the CT dose index in mGY.”
RADNET utilizes multiple dose-reduction techniques, including automated exposure control, adjustment of the milliamperes and/or kilovolts according to patient size, and use of iterative reconstruction technique.
“All of our technologists are trained accordingly in the application of these standardized low-dose protocols,” Bash says. “These measures help us guarantee that our patients get the highest-quality exam with the least amount of radiation.”
A Different Type of Dose
In recent years, “dose considerations” also have become a hot topic with regard to MRI, and they have nothing to do with radiation; they’re focused on the dose of gadolinium-based contrast agents (GBCAs). Bash says each year awareness of the importance of dose safety, protocols, and technology improves, culminating in the lowest possible dose for patients while preserving excellent image quality.
“While low-dose protocols are now standard of care across the country, in special cases, ‘ultralow-dose protocols’ are also encouraged, such as for patients in the hospital setting who require daily CTs to monitor changes in an acute process, such as an intracranial hemorrhage,” Bash says.
Bash notes that in the past five years, multiple publications have described the deposition of GBCAs in the brain and other organs such as the skin, bone, and liver; GBCAs can also cause nephrogenic systemic fibrosis in patients with poor kidney function. Gadolinium deposits in the brain can be directly visualized with MRI as areas of T1 hyperintensity in structures such as the globus pallidus and dentate nuclei for years following dose administration.
“The amount of gadolinium deposited in the brain following contrast-enhanced MRI is dependent on dose concentration, repetitive contrast administration over time, and the stability of the gadolinium agent used,” Bash says. “Nonionic linear chelates are the least stable, and the ionic macrocyclic chelates are the most stable. Gadolinium deposition in the brain is observed in both adult and pediatric patients, even in the setting of normal renal function.”
Although some adverse effects of gadolinium deposition have been postulated, at the present time the limited studies available do not show strong evidence for definitive adverse clinical effects. However, Bash says the residual presence of gadolinium in the brain and other tissues has prompted an active movement to minimize gadolinium deposition by using the most stable GBCAs available and administering contrast only when clinically necessary.
“One exciting new byproduct of this discussion is that it has prompted the radiology community to explore AI products utilizing deep learning algorithms to dramatically reduce the dose of gadolinium required for an exam,” Bash says.
For example, Subtle Medical has a new product called SubtleGAD, which can decrease gadolinium dose to only 10% of what was originally required. This product is still in development and not yet FDA approved, but Bash notes the preliminary results are promising.
AI is a topic being bandied about throughout the medical industry, and it has already had tremendous impact on radiology, one that will help with dose safety.
Mahesh says AI in CT is being used mostly in image reconstruction to reduce noise, improve image quality, and reduce dose.
“The other place we are beginning to see AI is in dose management, and I am sure we are going to see a lot more products trying to help in radiation dose management, especially from the informatics side,” Mahesh says.
Lawrence N. Tanenbaum, MD, FACR, director of CT, MR, and advanced imaging for Lenox Hill Radiology and Medical Imaging Associates in New York, says dose reconstruction took the industry by storm a decade ago and is now universally employed throughout the industry.
“It has contributed to a substantial reduction in dose, and overall it’s down to about 20% compared to just a few years back,” Tanenbaum says. “Recently however, it has become clearer through some of the literature that there are some shortcomings to reducing dose and using it with reconstruction, particularly when it comes to some low-contrast detectability issues and small lesions.”
That has led to an increased use of AI in CT, which provides the same dose reduction but doesn’t produce any of the image corruption associated with reconstruction.
“AI should take us at least as far as traditional dose-reducing techniques but preserve the image quality we are accustomed to seeing,” Tanenbaum says. “That will take the world by storm as it rolls out to wider availability.”
Tanenbaum says the vital impact of AI is not something to be feared but, rather, embraced.
“It was the future … but now the future has arrived,” he says. “AI tools can be applied either prior to, during, or following image acquisition.”
During image acquisition, AI tools such as FDA-approved SubtleMR and SubtlePET use deep learning algorithms to accelerate image acquisition and reduce scan time by 65% to 75%, while at the same time improving image quality. GE has similar deep learning acceleration software, which was presented at RSNA 2019 but is still FDA pending. Medic Vision’s iQMR can also accelerate image acquisition. Faster scans help limit dose.
Faster Scans, Added Value
Following image acquisition, triage tools can be applied, such as Aidoc’s FDA-approved software, which identifies intracranial hemorrhage and cervical spine fractures, then flags these findings and prioritizes the studies to the top of radiologists’ worklists, decreasing reader report turnaround time. MaxQ AI and Zebra Medical can also identify bleeds. Viz.ai’s Viz LVO can identify large vessel occlusions in acute stroke patients and not only flag the finding, but also alert the entire stroke team via transmission of pertinent images to an individual team member’s cell phone. Synaptive can apply real-time tractography to guide neurosurgeons during brain surgeries.
Bash has been actively using AI tools for neuroimaging studies for the past 15 years. Over that time, she has witnessed numerous tools emerge on the scene, especially since the onset of deep learning algorithms.
“However, the AI tool that I use every day in my clinical practice is quantitative volumetric postprocessing in the evaluation of patients with dementia, multiple sclerosis, seizures, traumatic brain injury (TBI), and pediatric imaging,” she says. “The two big players in this realm are Cortechs, with their NeuroQuant (NQ) and LesionQuant (LQ) products, and icometrix with icobrain.”
In her opinion, this volumetric software is an invaluable tool because it eliminates reader subjectivity by calculating the volume of substructures of the brain and comparing them with a large normative database to determine statistical significance for patient age. For example, in patients with dementia, NQ and icobrain help determine whether the degree of hippocampal atrophy is outside of the range of normal for patient age. In patients with multiple sclerosis, LQ and icobrain will calculate the overall volume of plaque burden and also drop heat color maps over individual plaques that are shrinking or growing. They then compare these volumes with a prior study to convey the rate and degree of disease progression over time.
“The NQ seizure protocol can help us to identify subtle cases of mesiotemporal sclerosis by assessing for hippocampal volume asymmetry,” Bash says. “The icobrain CT TBI software will calculate the volume of acute blood in various spaces of the intracranial vault.”
Bash also routinely uses MIM Neuro AI software on every FDG brain PET/CT and every amyloid PET study she interprets. The software uses Z-score analysis to calculate the statistical significance of metabolic activity in various substructures of the brain by comparing the standard uptake values with a normative database to help determine whether a true neurodementia syndrome is present.
“In my experience, these AI tools make us better radiologists,” she says. “They also make our referrers better clinicians by impacting management and adding value. These AI tools elevate our capabilities in many ways, including accelerated acquisition speed, improved image quality, and greater diagnostic benefit for our patients.”
— Keith Loria is a freelance writer based in Oakton, Virginia.