Radiology is not unlike any other medical specialty – biases exist even in the face of the strongest efforts to eliminate them. And, some of the most difficult to handle are the ones providers do not even know they have – the unconscious biases that largely go unrecognized.

But, the industry is beginning to talk about this issue more and more, educating radiology professionals about how unconscious biases appear and how they can best combat it. During this year’s Radiological Society of North America (RSNA) annual meeting, Jorge Soto, M.D., chief of radiology for Boston Medical Center, and Maureen Kohi, M.D., associate professor of clinical radiology at the University of California at San Francisco, led a session focusing on these biases that fly under the radar.

 spoke with them about the importance of shining a light on this problem and what radiology can do.

Videos

RSNA

Jorge Soto, M.D., and Maureen Kohi, M.D., discuss how unconscious bias presents in radiology and the tools that providers can use to recognize and minimize the biases when they appear.

Unconscious Bias: Handling What You Don’t Know You Have

Work has been ongoing for years to decrease the presence and impact of disparities throughout healthcare, including in radiology. But, problems still exist, and there are new challenges on the horizon that could make it more difficult for radiologists to make it easier for patients to get the imaging services they need.

During this year's Radiological Society of North America (RSNA) annual meeting, Richard Heller, M.D., associate chief medical officer and national director of pediatric radiology with Radiology Partners, and Lucy Spalluto, M.D., associate professor of radiology and radiological sciences at Vanderbilt University School of Medicine, participated in a session dedicated to addressing the continued existence of health disparities. 

 about the underlying factors that result in healthcare disparities, as well as the challenges providers and facilities face to improving access to care.

Podcasts

Rich Heller, M.D., with Radiology Partners, and Lucy Spalluto, M.D., with Vanderbilt University School of Medicine, discuss the highlights of their RSNA 2020 session on health disparities, focusing on the underlying factors and challenges radiologists face to providing greater access to care.

RSNA 2020: Addressing Healthcare Disparities and Access to Care

Alzheimer’s disease can be diagnosed earlier with multi-modal imaging by using a novel machine learning model even when some modalities are missing, according to a new study.

Researchers from a multi-institutional team developed a model that outperforms others that are currently available, and they presented their work during this year’s Radiological Society of North America (RSNA) annual meeting.

Developing tools for early detection of Alzheimer’s disease is critical because early identification can help slow progression of the disease. So far, healthcare has looked to neuroimaging as an avenue to catch the signs of this condition. But, more data is needed for scan to be successful.

To meet this need, the team, led by Fleming Y. Lure, Ph.D., created a novel transfer learning-based machine learning model that can diagnose and prognose mild cognitive impairment due to Alzheimer’s with varying availability of imaging modalities, such as MRI, FDG-PET, and amyloid-PET.

“Our research provides a clinical tool to assist physicians in diagnosis and prognosis of Alzheimer’s disease when disease is still early for their patients, which has tremendous clinical benefits,” Lure said.

For the study, Lure’s team included 241 patients with mild cognitive impairment from the Alzheimer’s Disease Neuroimaging Initiative Database. Of the group, 97 had mild cognitive impairment due to Alzheimer’s disease. Within two years, 26 individuals with mild cognitive impairment had converted to Alzheimer’s disease, and another 46 had converted to Alzheimer’s within six years. The team divided the patients into four sub-cohorts based on imaging available: MRI only; MRI and FDG-PET; MRI and amyloid-PET; and all three modalities.

Based on their analysis, the team determined the machine learning model achieved much better accuracy than the competing model, using each cohort for prognosis and diagnosis.

Articles

Novel deep learning model can provide needed information from multi-modal imaging even when some modalities are absent.



For Early Alzheimer’s Disease Detection, Machine Learning Offers More Information

When it comes to neuroradiology, there has been a significant amount of trailblazing technology developed in recent years, and now many of these tools are coming into clinical practice.

During this year’s Radiological Society of North America (RSNA) annual meeting, sub-specialty experts laid out some of the more interesting and effective technologies within the modality that are in the neuroradiology pipeline.

Approved by the U.S. Food & Drug Administration in 2016, MRgFUS, which uses a water bath around the head to facilitate signal transmission, is showing great promise with several applications, said Levi Chasen, M.D., associate professor of radiology at Weill Cornell Medicine.

 This condition affects more than 5 million Americans, and in recent history, it has been treated with stereotactic radiosurgery, gamma knife, and deep brain stimulation. But, MRgFUS is growing in popularity as a way to target the tremor more precisely via ventral intermediate nucleus thalamotomy. It can be used to intricately perform ablative techniques without damaging surrounding tissue.

Because there is currently no head coil for the 3T MRI system, MRgFUS is being used to provide a better peak inside the skull during surgical procedures. Although further advancements are needed to refine this use, images of diagnostic quality are possible, Chasen said.

“We’ve been able to come up with some very nice images of diagnostic quality,” he said. “We can see, while the patient is still on the table with the water bath in place, the kind of lesion we’re creating.”

It is possible, he said, to use a lower-frequency helmet to deliver cavitations. To do so, patients are injected with micro-bubbles that can target different areas of the brain to induce cavitations and transiently open the blood-brain barriers, which subsequently re-seals. Being able to do this successfully opens the door for drug delivery applications, he said.

Although 7T MRI received FDA approval for clinical use in 2017, there have been several obstacles to its use, including the need for a larger scanner room and a lack of a body coil. But, the technology also presents several advantages, said Kirk Welker, M.D., associate professor of radiology at the Mayo Clinic. And, with those advantages have come several successes, he said.

7T does a better job of other MRI strengths in detecting focal central dysplasia, mesial temporal sclerosis, cavernous malformations, and cerebellar hemorrhage. For example, in recent research with focal central dysplasia, 7T MRI was able to change treatment for 19 percent of patients for whom these findings were occult on 3T and 1.5T MRI scans.

7T MRI also performs better in detecting ACTH secreting microadenoma, Welker said. According to recent clinical findings he presented during the session, 7T MRI detected 90 percent of these findings that were present in patients who had other previous negative MRI scans.

This magnet strength is effective in identifying the central veins in the multiple sclerosis white matter lesions, he said, but its capabilities on follow-up are limited because no spine RF coil exists to image the spinal cord.

But, 7T MRI is not without its challenges, Welker explained. It still faces some stumbling blocks to wider use. Overall, six things stand in its way to getting more patients into the system:

There are few metallic devices that are approved by the manufacturers to be used in 7T systems.

Due to SAR limitations, 7T is not approved for use in children under 30 kg in weight.

SAR limitations also make already long scans longer.

The magnetic field can cause dizziness in patients.

The long narrow bore can create claustrophobia in some patients.

The small head coil might be not able to accommodate larger patients.

Overall, Welker said, it is important to remember that a 7T MRI does have unique factors.

“7T is not just 3T ‘on steroids,’” Welker explained.

Consequently, further comparative effectiveness research is necessary to address whether its benefits warrant the additional cost attached to the extra field strength, he said.

7T MRI and MR-Guided Focused Ultrasound present new clinical opportunities in the pipeline.

What’s on the Horizon with MRI for Neuroradiology

D-dimer values can help providers determine which COVID-19-positive patients should undergo CT pulmonary angiography (CTPA) to rule out the presence of a pulmonary embolism (PE), potentially alleviating some of the current workload for emergency radiology. 

In research presented during this year’s Radiological Society of North America (RSNA) annual meeting, investigators from the Hospital Universitario Ramón y Cajal in Madrid, share that there is a statistically significant difference in D-dimer levels between COVID-19-positive patients who have a PE and those who do not. Including those values in decision-making around imaging could reduce the number of patients sent for scanning, the team said.

“As a consequence of the inconclusive clinical and analytical criteria, selection of COVID-19 patients for CTPA may be a daunting task,” they said. “This has resulted in a noticeable increase in CTPA requests in the emergency radiology setting and has led us to ask ourselves if a better CTPA selection of COVID-19 patients is possible.”

PE is one of the more common complications that occurs in COVID-19-positive patients, and knowing whether a patient is PE-positive can potentially help providers avoid a fatality. Based on the results of their analysis, the team said, D-dimer values could be useful in pre-test probability assessments of PE in COVID-19 positive patients.

To make this determination, they evaluated CTPA scans from 187 COVID-19 positive patients – 119 men and 68 women – imaged in their institution this year from March 1 to April 20. The scan was performed with a 320-detection CT scanner after a 5 mL/s contrast agent injection. The team assessed both the D-dimer rate within 48 hours of the CTPA, as well as whether a PE were present,  what the PE characteristics were, such as lateral, unilateral, or bilateral, and the area affected.

Among the group, 59 patients were PE-positive, and 128 were PE-negative. The most frequently affected lobe was the right lower lobe, and the most proximal affected branch was on the segmental level. Based on their evaluation, the median D-dimer values for PE-positive and PE-negative patients were 9,668 ng/mL and 2,233 ng/mL, respectively, revealing a statistically significant difference (p<0.01), the team said.

“This might allow a better management of CTPA requests in COVID-19 patients,” they said, “and, thus, help relieve the burden on emergency radiology departments.”

Still, they said, special attention to be paid to the value that is actually used to determine who is sent for CTPA scanning. Among COVID-19 positive patients, the cut off value for PE probability prediction should be higher than what is used in non-COVID-19 positive patients because the D-dimer values in patients who have the virus are typically above what is considered normal for both PE-positive and PE-negative groups anyway.

To identify an acceptable and accurate value, the team suggested, further research is needed.

Whether a patient has a pulmonary embolism, a higher-than-normal D-dimer value can help identify patients for whom CTPA is most appropriate, alleviating some workload from emergency radiology.

Elevated D-Dimer Values Can Help Select Better COVID-19 Patients for CT Pulmonary Angiography

Hologic received 510(k) clearance from the U.S. Food & Drug Administration for its Genius AI Detection software this week during the Radiology Society of North America (RSNA) annual meeting. This move is one more development that is feeding growing trends in breast imaging, said Jennifer Meade, division president of breast and skeletal care for women’s imaging vendor Hologic.

 not only about some of Hologic’s more recent product offerings, but she also discussed feedback from the radiology community and some growing trends in breast imaging. In the coming years, she said, the industry can expect to see a greater focus on improved patient outcomes, workflow efficiency, and patient experience. In addition, she said developments and investments will continue around artificial intelligence, as well as personalized medicine.

With this latest FDA clearance, the Genius AI Detection software is set to follow that path, according to company information. It is designed to optimize workflow in four specific ways – by providing a case score that reflects confidence of malignancy, categorizing cases based on the number of lesions present, flagging the most concerning cases in less than a minute, and providing an estimate of how much time a radiologist can expect the image review to take.

Hologic’s Jennifer Meade, division president of breast and skeletal care, discusses the focus for breast imaging in the near future and some latest technologies.



Trends in Breast Imaging Care: Improving Outcomes, Efficiency, and Experience

Body mass index (BMI) is conventionally used to determine a person’s body-to-fat ratio, but analyzing abdominal CT images with an automated deep learning tool provides a more precise body composition measurement – and it can predict major cardiovascular events, according to new research.

In a presentation given during the Dec. 2 session of the Radiological Society of North America (RSNA), investigators from the University of California San Francisco (UCSF) discussed the superiority of using artificial intelligence with CT scans for assessing abdominal fat and determining whether a patient might face a heart attack or stroke within five years.

This research received the RSNA 2020 Trainee Research Prize, and the work was completed at Brigham and Women's Hospital (BWH) and the Dana Farber Cancer Center.

At a fundamental level, BMI as a measurement is flawed, said Kirti Magudia, M.D., Ph.D., an abdominal imaging and ultrasound fellow at UCSF and former BWH radiology resident, because it factors in only weight and height. Consequently, two people can have vastly different muscle-to-fat proportions and vastly different health outcomes, but still have the same BMI.

With a single CT slice of the abdomen, though, providers can get a much more accurate picture of a patient’s body composition. The image shows the volume of subcutaneous fat, visceral fat area, and skeletal muscle. The problem, though, is that taking these measurements manually is time consuming and expensive.

“Abdominal CT scans that are routinely performed provide a more granular way of looking at body composition, but we’re not currently taking advantage of it,” Magudia said.

To make it easier and more financially expedient to use these CT scans, Magudia and a multidisciplinary team created a fully automated method using deep learning that could accurately pinpoint body composition metrics from those abdominal images.

Using 33,182 abdominal CT scans performed on 23,136 outpatients at Partners Healthcare in Boston in 2012, the team identified 12,128 patients who had no major cardiovascular or cancer diagnoses at the time of imaging. With an L3 CT slice from the third lumbar spine vertebra, they calculated the body mass composition for each patient – 57 percent of whom were women – and divided the group into four quartiles based on normalized values of subcutaneous fat area, visceral fat area, and skeletal muscle area.

The team examined the scans of those 12,128 patients retrospectively and found that, within five years of the abdominal CT scan, there were 1,560 heart attacks and 938 strokes in the group. Their statistical analysis revealed that visceral fat area was independently associated with both of these events, but BMI was not. Even when the team adjusted for known cardiovascular risk factors, patients with the highest visceral fat area were most likely to suffer a heart attack, Magudia said, and those with the lowest amount of visceral fat area were protected against stroke.

“These results demonstrate that precise measures of body muscle and fat compartments achieved through CT outperform traditional biomarkers for predicting risk for cardiovascular outcomes,” she said.

Based on the team’s work, Magudia said, providers can feel confident using fully automated and normalized body composition analysis on larger scales.

“This work shows the promise of AI systems to add value to clinical care by extracting new information from existing imaging data,” she said. “The deployment of AI systems would allow radiologists, cardiologists, and primary care doctors to provide better care to patients at minimal incremental cost of the healthcare system.”

Using deep learning analysis with abdominal CT scans provides a more accurate body composition measurement that can predict which patients will suffer major cardiovascular events within five years.

Fat Measures from AI Abdominal CT Predicts Heart Attack and Stroke

A targeted approach to screening mammography for men who have genetic risk factors for breast cancer can lead to earlier detection and better outcomes. But, the exam is infrequently – and inconsistently – recommended, according to industry experts.

Even though only 1 percent of all diagnosed breast cancers occur in men, the incidence rate in this group has risen by 20 percent to 25 percent over the past two decades, according to industry experts. Screening mammography is known to catch cancers earlier, but there is still no official guidance for implementing this exam with men. In fact, the direction itself is inconsistent – The American Society of Clinical Oncology supports mammography for any man with a personal breast cancer history regardless of genetic factors, but the National Comprehensive Cancer Network only endorses clinical examination beginning at age 35 for men with a genetic mutation.

According to researchers from the New York University School of Medicine, very little consistency exists for when a clinician recommends screening for a man with a genetic mutation, such as BRCA 1 or BRCA 2. The team, led by Sushma Gaddam, M.D., clinical radiology instructor, presented their findings during this year’s Radiological Society of North America (RSNA) annual meeting.

“Although the absolute lifetime risk of breast cancer in men is relatively low, multiple concurrent risk factors can compound the actual risk for certain individuals,” the team said. “Given often delayed diagnosis in men due to lack of screening and worse prognosis as compared to women, optimizing early detection via screening in high-risk men is desirable.”

To assess how many at-risk men are screened, Gaddam’s team reviewed patient data and imaging studies from 414 men who had undergone genetic counseling, as well as multi-genetic panel testing. All the men had either a personal history, family history, or a combination of factors that put them at greater risk for developing the disease. According to the team’s analysis, 27.1 percent of men had a breast cancer-related genetic mutation – more than half of which were either BRCA 1 or BRCA 2.

Despite knowing the genetic mutation status of these men, though, Gaddam’s team discovered that physicians recommended only a clinical exam in 81 percent of BRCA carriers and suggested screening in only 50 percent of men who had a personal history of breast cancer. Additionally, among men who had both a personal history of the disease and a genetic mutation, only 40 percent received a recommendation for either a screening mammogram or a clinical exam – 60 percent were given neither suggestion.

But, when given the opportunity to pursue screening mammography, the team said, the majority of men complied. The 71.4 percent of men who adhered to a schedule indicate that men who are affected by a genetic mutation or who have a personal history of breast cancer are willing to undergo screening for early detection. And, the results can have a positive impact on patient health, the team said. Within their study population, screening mammography pinpointed two node-negative cancers.

Based on these findings, the team said, additional research is needed to create greater consensus on and standardization of risk management and screening.

Conducting screening mammography on men with a genetic risk for breast cancer can increase early detection, but clinical guidance and recommendations remain inconsistent. 

Targeted Screening Approach Could Identify More Men with Breast Cancer

Radiomics could soon be used to solve mysteries or crimes, according to newly presented research.

It is possible, according to investigators from the Radiology Institute at the University of Padova, for radiomics to reliably evaluate changes in bone structure due to fire exposure. In fact, specific features of radiomics can determine intervals of combustion.

The team presented their findings during the Radiological Society of North America (RSNA) annual meeting this year.

Being able to better assess these intervals is critical, said lead study author Amalia Lupi, M.D., because fire is frequently used to hide bodies after murders.

“In forensic scenarios with corpse concealment by combustion, every detail can be crucial to investigation,” she said.

By using radiomics, investigators can establish a timeline for the fire exposure, offering key details for a forensic investigation. In this study, Lupi’s team looked at the role radiomics can play in characterizing these time intervals. Based on their analysis, they proposed a method that can be used to gather more information that can further a crime investigation.

The team used high-resolution CT to examine 15 samples of human fibulae both before and after fire exposure for three combustion intervals, and they removed any samples that fractured after the second exposure. For the remaining bones, they collected information on bone density and volume, as well as 55 radiomics features.

According to that data, Lupi said, radiomics is able to detect changes in bone dependent on time and exposure.

Using high-resolution micro-CT with bones exposed to fire could help forensic investigators determine the timing of murders.



Radiomics Could Be the Next Crime Scene Investigator

Since the beginning of the pandemic, investigators have been trying to learn more about the connection between the COVID-19 virus and stroke. While the story is still unraveling, radiologists do know now that the link exists, and understanding more about the prevalence and appearance of stroke can impact patient care.

In a Nov. 30 session during this year’s Radiological Society of North America (RSNA) annual meeting, Ajay Gupta, M.D., professor of radiology and neuroscience at Weill Cornell Medicine, outlined what he dubbed the “4 Ms” of COVID-19-related stroke: mechanisms, magnitude, manifestations, and management.

Increasing, Gupta said, COVID-19 is being viewed as an endothelial cell disease in addition to being involved with the lungs. It causes significant deregulation in the immune system, leading to cytokine release syndrome.

“Within the blood vessels, this cytokine release, in the form of IL-6 and TNF-alpha, resulted in a vicious cycle of inflammation and thrombosis,” he said. “It’s this increase in inflammation that really can increase the risk of vascular events.”

A bit of uncertainty exists around the epidemiology of COVID-19-related stroke, he said, because in the initial phases of the pandemic, medical data and literature show a significant dip in the number of stroke cases reported. This was likely prompted by patients’ fear of coming to the hospital at that time.

“Without the true estimate, it’s hard to know how much excess stroke we have seen and what the risk is with COVID-19,” he explained. “But, we know we saw a precipitous drop in stroke cases reported during March and April – close to 50 percent. So, you would expect the number of strokes to increase.”

In addition, research in the past few months has shown a spotlight on how COVID-19 should not be compared to the flu, particularly in context of the risk of stroke. By comparing patients who were hospitalized for the flu between January 2016 to March 2018 to patients who were hospitalized for COVID-19 earlier this year, Gupta said, investigators determined that patients hospitalized for COVID-19 have an 8-fold increased risk of suffering a stroke while hospitalized.

“This underscores the argument that healthcare providers have been making,” he said. “COVID-19 really isn’t the flu. It is a much more serious disease.”

Do not expect a COVID-19-related stroke to appear differently on imaging than the strokes that typically show up in reading rooms, however. When it comes to presentation, these strokes are unremarkable, making them roughly indistinguishable from ischemic or other forms of stroke that occur in patients who are not positive for the virus.

Overall, patients mostly present with cryptogenic stroke without any known etiology, as well as some cardioembolic and large-vessel occlusion strokes.

But, when it comes to patients who are hospitalized for COVID-19, there is a distinct difference in the incidence and effect of stroke. In a paper published recently in 

 Gupta said, between 1.5 percent of patients in this group experience an ischemic stroke, and 1.8 percent of patients have some type of stroke event. This leads to a roughly 35-percent to 45-percent mortality rate. These findings can be particularly disturbing for younger patients who test positive because nearly 40 percent who suffer a stroke have no prior risk factors, and approximately 50 percent are largely asymptomatic for the disease.

Additionally, he said, pre-existing stroke history has been identified as an independent risk factor for a COVID-19-related stroke, as well.

While there has been very little official guidance for managing these patients, Gupta did offer three recommendations:

Continue to adhere to guidelines, following best practices for treatment, such as providing mechanical thrombectomy and endovascular therapy.

Use telemedicine to assess stroke severity, not only to maximize social distancing, but also to provide evaluation for a patient who may be unable to come to the hospital.

Rely on teamwork. Reach out to other neuroradiologists and interventional radiologists who have worked with this patient population.

“For those who went through those harrowing times in March and April,” he said, “we know that working closely and collegially with all providers is key to being able to take care of patients during this difficult time.”

The mechanisms, magnitude, manifestations, and management of stroke during the pandemic.

Latest Conference Articles