Diagnostic Imaging brings you daily coverage of the RSNA 2020 virtual annual meeting.

RSNA 2020: Nov. 30-Dec. 5

A deep learning algorithm that pulls biomarkers from screening mammograms can outperform traditional risk assessment models when predicting a patient’s individual breast cancer risk.

A team from Massachusetts General Hospital, led by breast radiologist Leslie Lamb, M.D., M.Sc, will present their findings during the Radiological Society of North America (RSNA) annual meeting this week.

“Traditional risk assessment models do not leverage the level of detail that is contained within a mammogram,” Lamb said. “Even the best existing traditional risk models may separate sub-groups of patients but are not as precise on the individual level.”

In fact, she said, current traditional risk assessment models rely on a sliver of patient data, including family history, prior breast biopsies, and hormonal and reproductive history. Breast density is the only bit of information pulled from screening mammograms that is incorporated into these models.

Expanded Risk Model Identifies Women Who Need Additional Breast Cancer Screening

Making use of only that one piece of information undermines the potential ability to give women individualized assessments of their personal risk, the team asserted.

“Why should we limit ourselves to only breast density when there is such rich digital data embedded in every woman’s mammogram,” said Constance D. Lehman, M.D., Ph.D., senior study author and division chief of breast imaging at MGH. “Every woman’s mammogram is unique to her just like her thumbprint. It contains imaging biomarkers that are highly predictive of future cancer risk, but until we had the tools of deep learning, we were not able to extract this information to improve patient care.”

To develop this deep learning algorithm, which has been externally validated in Sweden and Taiwan, Lamb’s team gathered data from women with a personal history of breast cancer, implants, or prior biopsies who were seen at one of five MGH breast screening locations. Overall, they included 245,753 consecutive 2D digital bilateral screening mammograms performed in 81,818 patients between 2009 and 2016. Of these exams, 210,819 studies from 56,831 patients were used for training, 25,644 from 7,021 women were used for testing, and 9,290 exams from 3,961 patients were used for validation.

AI Risk Score Predicts Breast Cancer Risk Better Than Mammographic Density

The team, then, used statistical analysis to figure out how well the deep learning model performed in predicting future breast cancer within five years of the index mammogram when compared to an existing commercially available risk assessment model, particularly the Tyrer-Cuzick version 8. The outperformance was significant, the team said – the deep learning model had a predictive rate of 0.71, but the traditional model only reached a 0.61 rate.

“Our deep learning model is able to translate the full diversity of subtle imaging biomarkers in the mammogram that can predict a woman’s future risk for breast cancer,” Lamb said, noting that future studies are planned with larger African American and minority populations.

Ultimately, she said, using a deep learning model can save both time and money.

“Traditional risk models can be time-consuming to acquire and rely on inconsistent or missing data,” she said. “A deep learning image-only risk model can provide increased access to more accurate, less costly risk assessment and help deliver on the promise of precision medicine.”

For more coverage based on industry expert insights and research, subscribe to the Diagnostic Imaging e-Newsletter 

Articles

A model that uses biomarkers pulled from a woman’s mammogram does produce a more accurate breast cancer risk assessment.

Deep Learning Model Outperforms Traditional Breast Cancer Risk Assessment

As the need for faster, accurate COVID-19 diagnostic tools continues to grow, another group of researchers has announced a tool that can help clinicians treat patients based on their specific needs and history.

Specifically, a team from Rensselaer Polytechnic Institute, with support from a National Institutes of Health grant, has developed a new algorithm that combines CT images of the patient’s lungs with non-imaging data, such as demographics, vital signs, and laboratory test results, to help providers pinpoint which COVID-19-positive patients will need intervention in the intensive care unit.

The team published the details of their algorithm recently in 

To determine the algorithm’s efficacy, the team, led by Pingkun Yan, Ph.D., associate professor of biomedical engineering, tested the tool on datasets collected from 295 patients from three hospitals, one each in the United States, Iran, and Italy, comparing the algorithm’s treatment predictions to the treatment each patient actually needed and received.

AI Algorithm Faster, More Accurate in Detecting COVID-19 on Chest X-rays

Based on their analysis, they determined that adding non-imaging features to the algorithm significantly improves prediction performance, enabling the algorithm to reach an AUC up to 0.884 and sensitivity as high as 96.1 percent. Consequently, Yan said, it can be a valuable clinical decision support tool for the management of COVID-19-positive patients.

“As a practitioner of AI, I do believe in its power,” he said. “It really enables us to analyze a large quantity of data and also extract the features that may not be that obvious to the human eye.”

In collaboration with researchers from Massachusetts General Hospital, Yan’s team is working to integrate this algorithm with another he previously developed that can use CT scans to assess cardiovascular disease risk. Underlying conditions, such as heart disease play a considerable role in COVID-19 mortality, he said, so being able to quantify a patient’s heart condition can help determine how the team can factor it into the AI prediction.

Ultimately, Yan said, the goal is to translate the algorithm into a method that doctors at Massachusetts General can use in clinical assessments – both with COVID-19 and beyond.

“We actually are seeing that the impact could go well beyond COVID diseases. For example, patients with other lung diseases,” he said. “Assessing their heart disease condition, together with their lung condition, could better predict their mortality risk so that we can help them to manage their condition.”

By incorporating non-imaging data, the algorithm can effectively pinpoint which patients will need ICU intervention. 


CT-Based AI Tool Offers Accurate Outcomes Predictions For COVID-19 Patients

Advanced MRI imaging – in combination with robotics – offers a look into the activity inside a baby’s brain, opening the door to the possibility of rehabilitating babies who have brain injuries.

investigators from King’s College London visualized how brain activity can be changed through associations. These findings could have implications on promoting the development of speech, language, and movement, they said.

“A baby’s brain is constantly learning associations and changing its activity all the time so that it can respond to the new experiences that are around it,” said Tomoki Arichi, Ph.D., clinician scientist and clinical senior lecturer in the Centre for the Developing Brain. “In terms of influencing patients and interpreting it in a wider context, what it means is that we should be thinking about how we could help with disorders of brain development from a very early stage in life because we know that experience is constantly shaping the newborn brain’s activity.”

This study, he said, is the first to show that it is possible to alter a baby’s brain activity through associative learning with different types of sensory experiences, particularly sound.

“We…found that when a baby is learning, it actually is activating lots of different parts of the brain, so it is starting to incorporate the ‘wider network’ inside the brain which is important for processing activity,” Arichi explained.

To assess brain activity in babies, the team played the sound of a jingling bell for six seconds for 24 infants. They paired the sound with the gentle movement prompted by a custom-made 3D printed robot that they strapped to each baby’s right hand. Throughout the experiment, investigators used a functional MRI to measure brain activity.

At the end of 20 minutes of repeating the association between the sound and the movement, the team played the sound alone and compared the babies’ brain activity to that captured during the learning period. The results, Arichi said, not only shed light on what happened inside a normal baby brain, but it can also help guide clinicians who are working with infants who have injured brains.

For example, he said, if a baby cannot process movement or movement is not associated with normal activity inside his or her brain, such as in the case of cerebral palsy, clinicians could try using the association with sound to help induce learning, potentially amplifying or rehabilitating their movements.

“With our findings, it raises the possibility of trying to do something to help with that through targeted stimulation and learning associations,” Arichi said. “It is possible to induce activity inside the part of the brain that normally processes movement, for instance, just by using a single sound. This could be used in conjunction with rehabilitation or to try to help guide brain development early in life.”

fMRI scans show that babies’ brain activity can be influenced through associative learning, presenting a potential strategy for promoting the development of life-long skills in infants who have injured brains. 

Functional MRI Reveals Secrets to Baby-Brain Activity, Opening Door to Rehabilitating Injured Brains

It’s Lung Cancer Awareness month, and one of the biggest advancements in the detection of the disease in recent years has been low-dose CT for screening. Not only does this technique offer a way to pinpoint whether a patient has lung cancer without exposing him or her to high levels of radiation, but it also provides a safe way to conduct multiple follow-up scans, as needed. 

However, to date, utilization of this low-dose method has fallen well below expectations and hopes. With only 4.5 percent to 5 percent of eligible patients actually getting scanned, there’s a lot of room for improvement. This week, 

spoke with Dr. Debra Dyer, chair of the American College of Radiology’s Lung Cancer Screening 2.0 Committee about the benefits of low-dose CT, the challenges that exist to wider adoption and implementation, and what providers and facilities can do to increase screening rates. 

Videos

Low-dose CT for lung cancer screening offers significant benefits for detection and follow-up, but utilization, to date, has been low. Debra S. Dyer, M.D., chair of the American College of Radiology's Lung Cancer Screening 2.0 Committee discusses what the screening provides and the challenges to more widespread implementation.

Diagnostic Imaging's Weekly Scan: Nov. 27, 2020

A newly developed sensor can measure weak magnetic brain signals that are easier to pinpoint than the electrical signals picked up by electroencephalogram.

These magnetic signals, measured by magnetoencephalography (MEG), are useful for earlier and more accurate diagnosis, potentially improving how providers understand connectivity in the brain and detecting signs of traumatic brain injury, dementia, and schizophrenia.

A team from the University of Birmingham in the United Kingdom published their work recently in 

“We know that early diagnosis improves outcomes and this technology could provide the sensitivity to detect the earliest changes in brain activity in conditions like schizophrenia, dementia, and ADHD,” said Ole Jensen, neuroscience professor and co-director of the University of Birmingham’s Centre for Human Brain Health.

To reach that goal, a team of scientists led by Anna Kowalczyk, a physicist with the School of Physics and Astronomy and the School of Psychology, designed a new optically pumped magnetometer (OPM) sensor that uses polarized light to detect changes in orientation of the way atoms spin when they are exposed to a magnetic field. These sensors are used in the MEG laboratories.

According to their results, not only do these sensors do a better job than commercially available sensors of detecting brain signals, but they also can distinguish them from background magnetic noise. Being able to make that differentiation means there is a possibility of MEG testing outside a specialized unit or hospital ward.

In addition, Kowalczyk said, the OPM sensor does not require a helium-cooling system to keep it constantly cool, nor does it need a zero-magnetic field environment to pick up brain signals. As a result, she said, the expectation is that these new sensors will extend the use of MEG for diagnosis and treatment.

The team is currently looking for research partners, they said, to explore how this sensor can be used for better diagnostics in neurological injury, neurological disorders, such as dementia, and psychiatric disorders, such as schizophrenia.

Optically pumped magnetometer sensor detects magnetic signals that could augment the detection of traumatic brain injury and disease.



New Magnetic Sensor Picks Up Weaker Brain Waves, Potentially Improving Diagnosis

The structure of the heart can make a difference in how stroke symptoms present in both African American and white patients, a new study has found. These results point to a need for updated diagnostic standards, investigators said.

 researchers from Weill Cornell Medicine outlined differences in the left atrium of the heart that could play a role in a patient’s risk of stroke.

“Our findings should not be taken to imply that these racial differences in people with stroke reflect innate biological differences,” said study author Hooman Kamel, M.D., director of the Weill Cornell Medicine Clinical and Translational Neuroscience Unit t the Brain and Mind Research Institute. “Our findings suggest that the current standards for diagnosing left atrial disease that can lead to stroke may need to be updated. These standards were established in groups of exclusively white people and may not fully reflect the variety and complexity of how these acquired cardiac issues are manifested in all people.”

Overall, African Americans in the United States do have a higher risk of ischemic stroke than their white counterparts, but known risk factors do not completely explain these differences. If providers could get a better understanding of the potential underlying causes of these strokes, they could possible do a better job in addressing these racial disparities, he said.

To delve deeper into this question, Kamel’s team enrolled 2,391 African American and white ischemic stroke patients into a study and examined electrocardiograms (ECG) and echocardiograms that focused on the structure and function of the heart’s left atrium. Using those tests, they measured the left atrium diameter and analyzed the amount of fibrosis and other abnormalities in the left atrium that have been associated with stroke.

Based on their evaluations, they determined the average diameter of the left atrium in study participants was 3.65 cm – 3.59 cm for African Americans and 3.69 for white patients. However, even after the team adjusted for age, body mass index, and other chronic conditions, such as high blood pressure, African American patients exhibited more markers for left atrial fibrosis on ECG.

“These findings reveal systematic differences in left atrium between [African American] versus white patients with ischemic stroke,” he said, “so clinicians should be aware that structural or functional changes in the left atrium may not present similarly in all stroke populations.”

Even though these tests and images support the need for updated diagnostic standards, Kamel said, further research is still needed on whether a better understanding around the many ways cardiac disease can manifest and clot-reducing drugs can reduce racial and ethnic disparities around stroke.

Left atrial diameter and fibrosis differences between African American and white patients could play a role the risk of ischemic stroke.

Heart Scans Reveal Structure May Impact Racial and Ethnic Stroke Risk Disparities

Cooking with wood or other biomass fuels can put people at greater risk for significant lung damage, and CT scans can reveal the impact pollutants and biotoxins make.

Results of a multidisciplinary study into the effect of cookstove pollutants, led by Eric A. Hoffman, Ph.D., University of Iowa radiology medicine, and biomedical engineering professor, will be presented during next week’s Radiological Society of North America (RSNA) annual meeting.

Although public health initiatives actively promote the use of cleaner-burning fuels, such as liquified petroleum gas, approximately 3 billion people worldwide still rely on biofuels, such as wood or dried brush, to cook. And, the pollutants produced by this type of cooking contribute to household air pollution-related illness, killing approximately 4 million people annually.

Being able to assess the effects on the lungs is important, said study co-author Abhilash Kizhakke Puliyakote, Ph.D., a postdoctoral researcher from the University of California San Diego School of Medicine. But, existing tests do not do an adequate job in evaluation.

“The extent of damage from biomass fuels is not really well captured by traditional tests,” he said. “You need more advanced, sensitive techniques like CT imaging. The key advantage to using imaging is that it’s so sensitive that you can detect subtle, regional changes before they progress to full blown disease, and you can follow disease progression over short periods of time.”

For their study, the team, which also included researchers from Periyar Maniammai Institute of Science and Technology, analyzed the impact of cookstove pollutants in 23 people who cooked with liquified petroleum gas or wood biomass in Thanjavur, India. After measuring the concentration of pollutants in their homes, the team tested their lung function using spirometry.

In addition, the team captured CT scans both when a participant inhaled and when he or she exhaled, measuring the difference between the images to determine lung function. Their analysis revealed that participants who cooked with wood biomass had greater concentrated exposures to pollutants and bacterial endotoxins than those who used petroleum liquified gas.

Wood biomass-burning study enrollees also faced an additional problem – they had significantly higher levels of air trapping in their lungs. This condition, associated with lung disease and often seen in tobacco smokers, occurs when part of the lungs cannot take in enough oxygen or expel carbon dioxide. In roughly one-third of the biomass users, 50 percent of the air they inhaled was ultimately trapped in their lungs, the team said.

According to Puliyakote, the lack of emphysema seen in this group points to a greater impact on the small airways of the lungs. Even without any obvious symptoms or breathing problems, he said, lungs can still experience injury and inflammation. Consequently, it is possible that many people can have undetected and unaddressed problems.

“For people exposed to biomass smoke for any extended duration, it is critical to have a complete assessment of lung function by healthcare professionals to ensure that any potential injury can be resolved with appropriate interventions,” he said.

In addition to offering insights into how biomass smoke from cooking acts on the lungs, this study’s findings can also add knowledge around the impact of the wildfires across the United States, as well.

CT scans reveal impaired lung function in individuals who use biomass, such as wood or wildfires, to cook.

Cooking with Wood? Smoke Pollutants Could Make It Harder to Breathe

Northwestern University researchers have created an artificial intelligence (AI) platform that can outperform thoracic radiologists in the detection of COVID-19 on chest X-rays.

Although this new AI tool, called DeepCOVID-XR, is not intended to be used in place of RT-PCR testing, the investigators said, it can spot evidence of COVID-19 on these images 10 times faster and with 1 percent-to-6 percent greater accuracy than providers can.

“We are not aiming to replace actual testing,” said senior study author Aggelos Katsaggelos, Ph.D., electrical and computer engineering professor who is also an AI expert with a courtesy appointment in the radiology department. “X-rays are routine, safe, and inexpensive. It would take seconds for our system to screen a patient and determine if that patient needs to be isolated.”

Rather, the intent, the team said, is to potentially identify patients with suspected viral infection earlier, prompting faster isolation to protect healthcare workers. It could also help detect patients who are not currently being assessed for COVID-19 who should be tested.

AI Tool Scores COVID-19 Lung Infection Severity

The team will publish their study today in 

 The algorithm, which is still in the research phase, is publicly available for other institutions to train it with new data, the team said.

To create and train their algorithm, the team gathered 17,002 chest X-rays – the largest published dataset of chest X-rays used to train an AI system during the pandemic. Among those images were 5,445 from COVID-19-positive patients seen across several sites throughout the Northwestern Memorial Healthcare System.

The team pulled 300 random images from Lake Forest Hospital and compared the performance of DeepCOVID-XR against five experienced cardiothoracic fellowship-trained radiologists. Overall, it took each radiologist 2.5-to-3 hours to examine each set of images. But, the AI system was able to do it in 18 minutes. The tool also produced slightly better accuracy – 82 percent compared to the 76 percent-to-81 percent from the radiologists.

The team stressed that this performance came from specialty providers.

“These are experts who are sub-specialty trained in reading chest imaging,” said Ramsey Wehbe, M.D., a cardiologist and postdoctoral fellow in AI at the Northwestern Medicine Bluhm Cardiovascular Institute. “Whereas the majority of chest X-rays are read by general radiologists or initially interpreted by non-radiologists, such as the treating clinician. A lot of times decisions are made based off that initial interpretation.”

Using this AI algorithm could also be a more cost-effective and accessible way of identifying patients who should be isolated.

“Radiologists are expensive and not always available,” Katsaggelos said. “X-rays are inexpensive and already a common element of routine care. This could potentially save money and time – especially because timing is so critical when working with COVID-19.”

The team did point out that, while the AI algorithm can be more effective with patients who exhibit COVID-19 signs on their images, it cannot improve the identification of the number of patients who do not show signs. That shortcoming highlights the limits of the system.

“In those cases, the AI system will not flag the patient as positive,” Wehbe said. “But, neither would a radiologist. Clearly, there is a limit to radiologic diagnosis of COVID-19, which is why we wouldn’t use this to replace testing.”

Investigators from Northwestern University have developed an algorithm that can identify evidence of COVID-19 on chest X-rays in a fraction of the time. 



Anxiety can do more than augment daily stress levels. New research with brain MRIs shows it is also associated with a faster transition from mild cognitive impairment to Alzheimer’s disease.

Deaths associated with Alzheimer’s disease have more than doubled since 2000, and many of these individuals start down the path by suffering mild symptoms, such as declines in memory and thinking skills. Many also exhibit signs of anxiety, but little is known about the role anxiety might actually play in cognitive decline.

In research that will be presented next week during the Radiological Society of North America (RSNA) annual meeting, investigators from the Medical University of South Carolina (MUSC) explored that question.

“We know that volume loss in certain areas of the brain is a factor that predicts progression to Alzheimer’s disease,” said senior study author Maria Vittoria Spampinato, M.D., MUSC radiology professor. “In this study, we wanted to see if anxiety had an effect on brain structure, or if the effect of anxiety was independent from brain structure in favoring the progression of disease.”

To get a better understanding of the impact of anxiety, Spampinato’s team examined brain MRI scans of 339 patients included in the Alzheimer’s Disease Neuroimaging Initiative 2 cohort. The average age was 72, and each patient had a baseline diagnosis of mild cognitive impairment. Of the group, 72 progressed to Alzheimer’s and the remaining 267 stayed cognitively stable.

Using these scans, the team identified baseline hippocampus and entorhinal cortex volumes. These areas are integral to forming memories, they said. In addition, they tested study participants for the presence of the ApoE4 allele, the most prevalent genetic risk factor for Alzheimer’s disease. For anxiety measurements and assessments, they conducted clinical surveys.

Their evaluation confirmed some expectations. Patients who progressed through to Alzheimer’s did, indeed, have significantly lower hippocampus and entorhinal cortex volumes, as well as a greater incidence of the ApoE4 allele.

The team also made an additional discovery – anxiety was independently associated with cognitive decline. Those study participants with anxiety symptoms did progress to Alzheimer’s faster regardless of whether they had the ApoE4 allele or had brain volume loss. Being able to identify this link, they said, opens the door to improving how patients with mild cognitive impairment are screened and managed.

“We need to better understand the association between anxiety disorders and cognitive decline,” Spampinato said. “We don’t know yet if the anxiety is a symptom – in other words, their memory is getting worse and they become anxious – or if anxiety contributes to cognitive decline. If we were able in the future to find that anxiety is actually causing progression, then we should more aggressively screen for anxiety disorders in the elderly.”

For example, said Jenny L. Ulber, MUSC medical student and first author, it could be helpful to conduct an assessment for anxiety disorders alongside the depression evaluations completed on elderly patients who are in the hospital.

“Middle-aged and elderly individuals with high levels of anxiety may benefit from intervention,” she said, “whether it be pharmacological or cognitive behavior therapy, with the goal of slowing cognitive decline.”

Their study did have a limitation of relying on MRI scans conducted at one moment in time. For future work – and to continue to better understand this link between anxiety and brain structure – the team plans to study MRIs collected after initial scans.

“We’re now interested in looking at changes over time to see if anxiety has an effect one way or the other on how fast the brain damage progresses,” Spampinato said. “We will also take a closer look at gender differences in the association between anxiety and cognitive decline.”

Brain MRI scans reveal that anxiety is independently associated with a faster progression from mild cognitive impairment to Alzheimer’s disease.

Anxiety Goes Beyond Stress – It Can Speed Up Alzheimer’s

A common weight-loss surgery in adolescents – sleeve gastrectomy – might induce weight loss, but, according to CT and MRI scans, it also weakens the patient’s bones.

In a sleeve gastrectomy procedure, 75 percent of the patient’s stomach is removed to control food intake, promoting weight loss. As a result, the stomach becomes tube- or sleeve-shaped, and from 2005 to 2014, the number of these procedures performed on adolescents has increased 100-fold, said a team from Harvard Medical School.

“Childhood obesity is a major public health issue that has increased over the last 10 years,” said lead investigator Miriam A. Bredella, M.D., radiology professor at Harvard and vice chair of the Massachusetts General Hospital radiology department. “Sleeve gastrectomy is the most common bariatric surgery procedure performed in children and adults.”

Bredella’s team will present their findings next week during the Radiological Society of North America (RSNA) annual meeting.

It is known that bariatric surgery has long-term bone effects in adults that can lead to a higher risk of fracture. The team set out to determine if the same thing happens to teen-agers who have the surgery during their critical years of growth and development when they are accruing bone mass.

For their study, the team enrolled 52 adolescents who had moderate-to-severe obesity with an average body mass index (BMI) of 45. A BMI over 30 is considered obese. The average age was 17.5 years, and 38 of the participants were girls. Bredella’s team split the group in half – 26 participants underwent sleeve gastrectomy, and 26 did not.

To detect any changes, all participants had a quantitative CT to detect volumetric bone mineral density changes in the lumbar spine performed both before any surgery and 12 months later. The test is particularly accurate for this measurement, the team said. In addition, the team also performed proton MR spectroscopy on all participants to quantify the amount of bone marrow fat in the lumbar spine. Recent studies have shown that bone marrow fat can be a bone quality biomarker because it responds to nutritional changes.

Based on the results of these scans, the team determined that adolescents who had the surgery lost 34 kilograms/75 pounds, and the participants who did not have the procedure saw no significant weight loss. But, the teens who had the surgery also saw a significant increase in bone marrow fat with a corresponding decrease in bone density in the lumbar spine.

“Adolescents who underwent sleeve gastrectomy had bone loss and an increase in bone marrow fat, despite marked loss of body fat,” Bredella said. “While weight-loss surgery is successful for weight loss and improving metabolic disorders, it has negative effects on bone.”

This post-surgery change was not surprising, though, she said, because greater bone density comes with greater weight-bearing. And, alongside, those drops in bone density, the surgery also disrupted hormones and nutrients that are integral to bone health.

Knowing how this surgery affects young patients can be important to potentially safeguarding their bone quality as they age, Bredella said.

“We need to identify mechanisms that will help prevent bone loss in these patients and to make adolescents with obesity more aware of bone health,” she said. “Adolescence is the critical time for bone mass accrual, and any process that interferes with bone accrual during this time can have dire consequences later in life.”

Images reveal that adolescents experienced a loss in volumetric bone mineral density after sleeve gastrectomy and extreme weight loss.

CT, MRI Show Weaker Bones in Teens Post-Weight Loss Surgery

Radiologists who interpret scans of non-displaced fractures to the ulna in female patients can play a role in identifying a victim of intimate partner violence, a new study says.

In research presented this week during the Radiological Society of North America (RSNA) annual meeting, investigators from Brigham & Women’s Hospital shared their findings that they say supports need to screen for intimate partner violence in women who have these types of injuries.

These non-displaced fractures typically occur when someone uses their forearm to shield their face from being struck, said senior study author and radiologist Bharti Khurana, M.D. They are mostly seen in men, so when a woman presents with a break like this, it should raise a red flag.

“I would see these types of injuries in men, but one in a while, I would see them in women,” she said, discussing her own practice experience. “I never correlated it with intimate partner violence until recently.”

Radiology Spotlights Disturbing Uptick of Intimate Partner Violence During Pandemic

But, after sharing her suspicions with an orthopedic surgeon colleague, she decided to study this issue. After searching the electronic medical records from six hospitals, looking specifically for ulnar fractures in women ages 18 to 50, Khurana’s team identified 62 women who fit the description. Their average age was 31, and 12 had been confirmed as victims of intimate partner violence. Another eight were suspected.

Based on her analysis, she determined that intimate partner violence was strongly associated with X-ray that included a minimally displaced fracture.

“The radiological characteristics we were looking at were the location of the fracture, the pattern of the fracture in terms of how it broke, and the displacement of the fracture,” said lead study author David Sing, M.D., an orthopedic surgery resident at Boston Medical Center. “Out of all those things, what we usually saw was a minimally displaced fracture, meaning the bone is broken all the way through but has not shifted significantly.”

In many cases, Khurana explained, women who have ulnar fractures due to intimate partner violence typically blame a fall. However, falls are much more likely to result in a break in the arm’s other bone, the radius.

“It’s actually rare to break your ulna in a fall,” Khurana said. “If a radiology is seeing an ulnar fracture that is non-displaced, and the woman says she had a fall, it’s actually quite concerning for intimate partner violence.”

This finding was bolstered by the fact that four out of eight intimate partner violence victims reported their injuries as the outcome of a fall, but the ones who were not sustained the ulnar fractures from car accidents or an accidental strike, such as hitting a tree while skiing.

The frequency of this type of injury among victims of intimate partner violence underscores the need for greater screening, she said. In only 40 percent of confirmed or suspected cases did providers do a formal evaluation or screening. Consequently, in many cases, the radiologist who sees such a fracture on an image can help improve these efforts by alerting the emergency department physician or orthopedic surgeon that an ulnar fractur frequently accompanies intimate partner violence, giving the ordering physician the opportunity to dig deeper into the patient’s history. In fact, she said historical imaging analysis by itself raised the red flag with 75 percent of confirmed intimate partner violence patients.

“Careful analysis of previous imaging exams may also help radiologists confirm their suspicion of intimate partner violence,” said study co-author Rahul Gujrathi, M.D., a Brigham & Women’s radiology fellow.

The ability to recognize the signs of intimate partner violence is particularly critical now since the rate of 

 has increased significantly during the pandemic. A quick diagnosis can be substantially beneficial for the patient.

“The sooner we can address and change the behavior, the better,” she said. “Just like radiologists want to diagnose cancer as early as possible, it’s the same thing with this. If we diagnose early, we have a better chance to break the cycle of violence.”

Radiologists have the opportunity to identify women suffering intimate partner violence by detecting a specific type of forearm fracture. 

With Forearm Fracture in Women, Suspect Domestic Violence

At some point during my more-hectic-than-usual Thanksgiving work week, I got messaged from our telerad support-staff: A patient whose spinal MR I had read wanted to talk to me.

Your mileage may vary, but in my practice it’s vanishingly rare for a patient to speak directly to a rad who interpreted their stuff. Excepting things with a personal touch, such as interventional or other fluoro, diagnostic mammo, etc.

I could go into all sorts of reasons for this, but a lot of them can be summed up: prior to the referral for imaging, there’s been an interaction between a patient and his physician. When that physician decides that it’s time for imaging, s/he’s creating a new interaction with a radiologist.

The referrer is the only one in the mix with a full understanding of the circumstances necessitating imaging (think of the difference between a patient’s chart in the referrer’s files and the typical “reason for imaging” that the rad receives). Thus, he or she is the best person to explain or discuss results of that imaging with the patient. Creating a new interaction between patient and rad, excluding the referrer, opens wide the door to potential misinformation—not only for the patient, but also for the rad to potentially enter into the medical record. More than a few referrers would be perturbed, if not downright angry, to learn that this had happened.

So, when I received this message, it got me on high alert. Something potentially unkosher was afoot, and I’d best tread carefully. I had our support-staff advise the patient that he should speak with his physician, who had all of the relevant info, and if, then, needed, his doc could speak with me.

Support-staff successfully did this, mentioning to me along the way that it seemed like the patient wanted me to amend my report, which had described his spondylosis as being mild, to say it was “moderate.” Regarding getting his insurance to pay for something, perhaps surgery. (By the way, I had another look at the images, just in case, and there was nothing on them to justify such a change from me.)

Fast-forward a couple of hours, and his physician was on the line. I took a deep breath, ready to be pressured to alter my interpretation in order to get the insurer to play ball. Happily, the physician was not after this—indeed, had no idea why the patient had contacted her office to get her to speak with me until I told her. She was mildly horrified at her patient’s request, and reassured me that there was no way she would ask such a thing of a rad. A few pleasantries later, we returned to our respective workdays.

Aside from my internal fretting, I can’t think of an easier/better way this episode could have gone, thanks to the telerad support-staff who ran interference for me. If, instead, there were no buffer at all to people ringing me up at my workstation, the patient would have easily gotten ahold of m, and caught me flat-footed. And, depending on how insistent he was, no amount of diplomacy might have enabled me to stand my ground without him going away mad.

He wouldn’t be the only one able to directly contact me. It could be any random patient (or referrer), whose studies might have been read by other rads. Family members of patients. Insurance companies. Recruiters. Political pollsters. Nigerian princes, you name it.

Pretty much anybody reading this has had the unpleasant experience of trying to phone someone—their own doc’s office, maybe, or some other business (banks and airlines leap to mind)—and being in that patient’s position of not being able to get through to the person who seems most suited to solve a problem. There’s an entire industry built around running interference.

The big difference there, however, is that most of that is about wearing down the caller to the point that s/he hangs up, or maybe doesn’t even try in the first place, thus diminishing the business’ overall burden of incoming calls. This often happens with the assistance of a touchtone-maze and recordings that steer the caller to dead-ends, if not an on-hold purgatory.

Legit instances of running interference do exist, however, and I’ve happily been shielded by them for the past 9-plus years of my radiological career. It was such a seismic shift in work environment from my previous jobs that I am increasingly amazed when I hear of other rad groups that haven’t instituted similar measures. So much efficiency is needlessly lost, and so many avoidable problems allowed to occur.

Screening out telemarketers and other outright fluff is a good start, but that’s amateur-hour. It doesn’t take much more training and effort to have phone-answerers direct calls so that rads are only interrupted by queries pertinent to them (cases they personally have read, or are capable of reading/protocoling, for instance).

It can get even more efficient than that. Suppose a referrer wants to talk to a rad about a specific case. If the support-staff can get the relevant information (medical record number, for instance), and shoot that in an instant message to the rad, he can get the case on his screen and message back that he’s ready to talk. If the instant message can be embedded with a hotlink to the case in question: rad clicks on it, and the case (including report) comes right up on his screen, ready for review/discussion.

Coordinate this process with support-staff in the referrer’s office, and neither doc has to get on the phone till they are both prepared to discuss the matter at hand. Referrer tells his staff he wants to talk with the rad about patient’s X’s scan. Staff contacts rad group staff with the request, and rad group instant messages the details to the relevant rad. Rad indicates that s/he got the instant message and has the case open, staff connect her/him with the referrer, and the game is afoot. Just about every second the docs spend on the phone is with each other, discussing the case…no wastage of doctorly time.

 hearing a referrer breathing in their ear as they fumble around to get the case on their screen, including spelling out the patient’s name in a search function, and, then, trying to re-familiarize themselves with the case on the fly. To each his own.

Follow Editorial Board Member Eric Postal, M.D., on Twitter, 

When support staff step in to control communications and keep the workday running smoothly.

Running Interference

A photo of a chest X-ray captured by a smartphone can be enough for providers to diagnose tuberculosis (TB), according to new research.

In a poster presented during this year’s Radiological Society of North America (RSNA) annual meeting, researchers from National Tsing Hua University in Taiwan explained how using a deep-learning TB detection model with those smartphone pictures can impact diagnosis of the disease. This is the first study to pair a deep learning model with a smartphone for this purpose.

This method is critical, said lead study author Po-Chih Kuo, Ph.D., assistant professor of computer science, because early TB diagnosis has historically been difficult in resource-poor countries where radiologists and high-resolution images are not always available. By deploying the algorithm – called TBShoNet – in a smartphone, providers will be better equipped to make a diagnosis on their own.

“We need to extend the opportunities around medical artificial intelligence to resource-limited settings,” Kuo said.

Kuo’s team used a database of 250,044 chest X-rays with 14 pulmonary labels that did not include TB to pre-train the neural network. They, then, recalibrated the model for chest X-ray photographs, using simulation methods to augment the dataset. Adding an additional two-layer neural network that was trained on augmented chest X-ray images to the pre-trained model completed TBShoNet.

To test TBShoNet, Kuo’s team took 662 chest X-ray photographs – 336 TB and 326 normal – with five different smartphones. According to their analysis, the model produced 81 percent sensitivity and 84 percent specificity for TB classification.

Applying a deep-learning model to a photograph of a chest X-ray can help providers in resource-poor areas diagnose the disease.

With Forearm Fracture in Women, Suspect Domestic Violence

Running Interference

Using AI Can Help Diagnose Tuberculosis on a Smartphone