Hidden rib fractures are a diagnostic issue that frequently confronts emergency radiologists.
An increase in imaging exams coupled with more images per exam can raise the risk of diagnostic error. In addition, where chest x-rays are typically the first line of diagnostic imaging, they reportedly miss more than half of rib fractures, according to some estimates.
Ensuring that few to no rib fractures are missed during interpretation drives the work of innovative hospital emergency departments that have deployed fracture detection software, post-processing, and/or advanced visualization tools with CT scans to increase sensitivity for fracture detection, reduce reading time, and improve outcomes for trauma patients.
Hidden rib fractures are a diagnostic issue that frequently confronts emergency radiologists.
An increase in imaging exams coupled with more images per exam can raise the risk of diagnostic error. In addition, where chest x-rays are typically the first line of diagnostic imaging, they reportedly miss more than half of rib fractures, according to some estimates.
Ensuring that few to no rib fractures are missed during interpretation drives the work of innovative hospital emergency departments that have deployed fracture detection software, post-processing, and/or advanced visualization tools with CT scans to increase sensitivity for fracture detection, reduce reading time, and improve outcomes for trauma patients.
Radiologists pressed
Marc Succi, MD, an attending radiologist at Mass General Brigham (MGB) Emergency Radiology in Boston, told AuntMinnie.com that rib fractures can be difficult to detect on radiographs. MGB sees about 110,000 emergency patients annually.
Like other institutions working to overcome these challenges, MGB has long instituted an optimized "CT trauma scan" and continues to refine protocols for speed and sensitivity for a number of acute traumatic injuries, including fractures.
"Obvious displaced fractures you can see on chest or rib radiograph series," Succi said, "but CT scan is often needed to diagnose subtle fractures, including nondisplaced rib fractures, and a host of other acute injuries. Whether or not this changes treatment course and for what proportion of patients can be debated, as well as the impact of obtaining more CTs on turnaround times and workflow."
However, Succi said generally, utilizing either rib radiograph series or CT, or both, for rib fracture detection can depend on resources available and the acuity of the setting and practice preferences -- for example, many urgent care clinics may only have radiography available, not CT.
Marc Succi, MD.
Missed rib fractures as a diagnostic error on CT trauma scans can at least be partially attributed to the issue of radiologists pressed for time due to an increasing number of exams and images per exam that they are responsible for interpreting, according to Succi.
At MGB's emergency locations, most patients coming in with some level of trauma, especially given the potential lack of clear history accompanying trauma patients, will get initial radiographs and then some combination of trauma or other optimized CT, Succi explained.
"Thin-slice CT scans have helped because often these fractures are subtle and can be obscured or missed by thicker slices," Succi said. He is often reading pan-scan CT for trauma (some refer to it as whole-body CT [WBCT]) -- that focuses the study on the head, cervical spine, chest, abdomen, and pelvis to detect all kinds of traumatic injuries, fractures among them.
Institutions use tailored protocols, Succi added, saying that some locations have changed their protocols over the years to alter contrast timing for the purpose of detecting pulmonary artery embolism or aortic injury on the same scan, for example. Depending on patient needs and circumstances, altering contrast timing can prevent repeat scans, Succi said.
"The context of how your hospital practices is important," he added.
Problem
Hospital and health system leaders concerned about rib fracture-connected organ and aortic injuries, prolonged hospital stays, and readmissions may be interested in a meta-analysis published January 27. Researchers explored diagnostic accuracy, rib fracture detection (RFD), and why proper diagnosis of rib fractures is critical for improving clinical outcomes of trauma patients.
Inadequately controlled rib cage pain and breathing mechanics due to rib fractures can put patients at risk of atelectasis, pleural effusion, and pneumonia -- with prolonged stays in the intensive care unit as a result, explained Christopher Collins and colleagues from California University of Science and Medicine, Loma Linda University Health, and Arrowhead Regional Medical Center in Colton, CA.
Notably, major pulmonary complications affect an estimated 33% of elderly and frail patients, with three or more rib fractures associated with increased mortality, the team noted for the Journal of Imaging Informatics in Medicine. Importantly, though, rib fractures should be managed properly.
In pediatrics, rib fractures are highly predictive of nonaccidental trauma in children under 3 years old. Rib fracture detection on pediatric radiographs is challenging because fractures can be obliquely oriented to the imaging detector, obfuscated by other structures, incomplete, and nondisplaced. The problem has led hospital trauma teams and radiologists to explore high-sensitivity methods for automated rib fracture detection, as described in a study published April 2024 in Scientific Reports.
Sensitive CT
Improving the sensitivity of fracture detection to ensure that few to no fractures are missed during interpretation drives the work of researchers from around the world. Statistics associated with missed rib fractures vary.
Where chest x-rays are typically the first line of diagnostic imaging, they reportedly miss more than half of rib fractures. Sensitivity may be as low as 15%, Collins and colleagues noted in their research.
In children younger than 3, studies have shown that up to two-thirds of rib fractures may be missed during initial interpretation, according to researchers at Michigan State University in East Lansing, Helen DeVos Children's Hospital in Grand Rapids, and Seattle Children's Hospital.
CT has become viewed as more sensitive, with the added benefit of detecting potential injury to surrounding structures and organs, such as the spleen, liver, or spine. MGB's Succi pointed to a 2022 study out of Beijing, China, that compared the results of AI rib fracture diagnosis using initial CT and follow-up CT.
For the study, a radiologist group (group 1), AI group (group 2), and radiologist with AI group (group 3) reviewed CT images for rib fractures.
Group 3, the radiologist plus AI group, showed high sensitivity and specificity at 91.6% and 99.7%, respectively. The radiologist-only group performed at 82.2% sensitivity and 99.8% specificity.
Overall, sensitivity was higher for AI plus radiologist, compared with either alone. AI alone had low sensitivity and specificity, at 79.4% sensitivity and 84.9% specificity, according to lead author Quanshuai Zhou, MD, from the department of radiology at The Fifth Affiliated Hospital of Sun Yat-sen University in Zhuhai, China, and colleagues.
Automated RFD
With radiologists interpreting more CT images daily than ever before, the risk of an oversight increases, even for experienced professionals. At some locations, emergency radiology volumes are increasing as much as 20% year over year, according to Succi.
"All radiologists have less time per CT scan and per image than ever before to see rib fractures and other acute findings," Succi told AuntMinnie. "Time to detect and accurately interpret findings is decreasing, and pressure to reduce turnaround times is increasing. That's really one of several challenges that radiology is facing at this time."
Toward improving workload conditions and patient outcomes, hospital emergency departments have deployed fracture detection software, post-processing, and/or advanced visualization tools. University Hospitals (UH) radiologists in Cleveland, Ohio, for example, have estimated that automated fracture detection decreased reading time by 27% and increased sensitivity for fracture detection by 10.4%, according to a retrospective study there.
Navid Faraji, MD.
Navid Faraji, MD, a musculoskeletal radiologist at UH said he and his colleagues have been using fracture detection tools clinically for the past couple of years. UH, a level 1 trauma center, interprets approximately 2 million exams per year as a department, many of which are in the emergency setting.
While Faraji serves as associate program director of UH diagnostic radiology residency, he also interprets a large volume of emergency x-rays in his clinical practice; and, using CT at UH depends on the mechanism of injury. Depending on the type of trauma, CT is reserved for cases where pain persists or worsens and cannot be attributed to any other cause when patients are presenting with rib pain.
"CT is a more sensitive means by which to detect rib fractures relative to x-rays due to elimination of superimposed structures and 2D/3D reconstruction techniques," Faraji added. "Fracture detection AI tools can also help us minimize mistakes."
Although most commercial AI products have been designed for radiography, about one-third were tailored to CT as of December 2023, according to a UK-based report published in BJR Open. Pediatric products have lagged behind adult products.
Faraji noted that there are many clinically available U.S. Food and Drug Administration (FDA)-cleared AI tools (technically medical devices) for fracture detection. The data, he said, is promising that AI-assisted additional-reader workflow might be a feasible strategy for reducing errors in rib fracture evaluation in real clinical practice.
"There are some rib fracture detection tools that are being deployed on chest x-rays and CT," Faraji said, "but they're not currently advised to be a standalone tool to assess for the presence or absence of rib fractures."
Faraji emphasized the importance of checking and monitoring any AI tool, especially when using the tool for teaching the new generation of radiologists.
At Yale New Haven Health and its hospitals in and around New Haven, CT, emergency radiologist Joseph Cavallo, MD, is medical director for clinical operations and an assistant director of informatics with a focus on clinical implementation of AI. For the past five years, Cavallo has served as attending radiologist in the emergency radiology department.
Joseph Cavallo, MD, MBA.
Yale New Haven's emergency radiology department interprets approximately 250,000 studies per year and is currently running a rib fracture detection algorithm on CTs, according to Cavallo. The algorithm can alert the reading radiologist about the presence of a fracture and also highlight the location of suspected rib fractures.
"I've found it most helpful when there's a rib fracture on studies that were not ordered specifically for trauma," Cavallo told AuntMinnie, adding that the rib fracture algorithm's value is not necessarily as a triage tool but, instead, it can be useful for decreasing radiologists' search time and increasing overall accuracy.
"We will often get incomplete patient history or history without mention of trauma as a contributing source of pain," Cavallo added. "Whether it's catching lower rib fractures on a CT of the abdomen and pelvis performed for generalized pain or an incidental rib fracture on a chest CTA performed for evaluation of pulmonary embolism, I find value in detecting fractures in cases where a rib fracture is not the focus of my diagnostic search pattern."
When it comes to implementing an AI-based tool, Cavallo recommended characterizing a well-defined clinical problem or pain point upfront, for which you are then able to find a solution.
"You always want to think about your workflows," Cavallo added. "Where can this tool be most impactful? Is it a patient-facing tool? Is it part of the technologist workflow? Is it something to alleviate radiologists' volume or time to read? The more data you have to quantify a problem, the better you can assess or quantify the value of a specific software program."
What lies ahead?
"I think the future of these algorithms is moving beyond detection and alerting toward deeper integration in our workflows, [with] interactive findings displayed within the PACS viewer, integrating AI outputs with data from the EMR and prior exams, automated labeling and measurement, automated reporting processes, et cetera, " Cavallo said. "That is the type of innovation I want to see in the future."