Search

Background and Purpose: Artificial intelligence models in radiology are frequently developed and validated using data sets from a single institution and are rarely tested on independent, external data sets, raising questions about their generalizability and applicability in clinical practice. The American Society of Functional Neuroradiology (ASFNR) organized a multicenter artificial intelligence competition to evaluate the proficiency of developed models in identifying various pathologies on NCCT, assessing age-based normality and estimating medical urgency., Materials and Methods: In total, 1201 anonymized, full-head NCCT clinical scans from 5 institutions were pooled to form the data set. The data set encompassed studies with normal findings as well as those with pathologies, including acute ischemic stroke, intracranial hemorrhage, traumatic brain injury, and mass effect (detection of these, task 1). NCCTs were also assessed to determine if findings were consistent with expected brain changes for the patient's age (task 2: age-based normality assessment) and to identify any abnormalities requiring immediate medical attention (task 3: evaluation of findings for urgent intervention). Five neuroradiologists labeled each NCCT, with consensus interpretations serving as the ground truth. The competition was announced online, inviting academic institutions and companies. Independent central analysis assessed the performance of each model. Accuracy, sensitivity, specificity, positive and negative predictive values, and receiver operating characteristic (ROC) curves were generated for each artificial intelligence model, along with the area under the ROC curve., Results: Four teams processed 1177 studies. The median age of patients was 62 years, with an interquartile range of 33 years. Nineteen teams from various academic institutions registered for the competition. Of these, 4 teams submitted their final results. No commercial entities participated in the competition. For task 1, areas under the ROC curve ranged from 0.49 to 0.59. For task 2, two teams completed the task with area under the ROC curve values of 0.57 and 0.52. For task 3, teams had little-to-no agreement with the ground truth., Conclusions: To assess the performance of artificial intelligence models in real-world clinical scenarios, we analyzed their performance in the ASFNR Artificial Intelligence Competition. The first ASFNR Competition underscored the gap between expectation and reality; and the models largely fell short in their assessments. As the integration of artificial intelligence tools into clinical workflows increases, neuroradiologists must carefully recognize the capabilities, constraints, and consistency of these technologies. Before institutions adopt these algorithms, thorough validation is essential to ensure acceptable levels of performance in clinical settings., (© 2024 by American Journal of Neuroradiology.)

In this review, concepts of algorithmic bias and fairness are defined qualitatively and mathematically. Illustrative examples are given of what can go wrong when unintended bias or unfairness in algorithmic development occurs. The importance of explainability, accountability, and transparency with respect to artificial intelligence algorithm development and clinical deployment is discussed. These are grounded in the concept of "primum no nocere" (first, do no harm). Steps to mitigate unfairness and bias in task definition, data collection, model definition, training, testing, deployment, and feedback are provided. Discussions on the implementation of fairness criteria that maximize benefit and minimize unfairness and harm to neuroradiology patients will be provided, including suggestions for neuroradiologists to consider as artificial intelligence algorithms gain acceptance into neuroradiology practice and become incorporated into routine clinical workflow., (© 2023 by American Journal of Neuroradiology.)

Background and Purpose: Multi-parametric MRI, provides a variety of biomarkers sensitive to white matter integrity, However, spinal cord MRI data in pediatrics is rare compared to adults. The purpose of this work was 3-fold: 1) to develop a processing pipeline for atlas-based generation of the typically developing pediatric spinal cord WM tracts, 2) to derive atlas-based normative values of the DTI indices for various WM pathways, and 3) to investigate age-related changes in the obtained normative DTI indices along the extracted tracts., Materials and Methods: DTI scans of 30 typically developing subjects (age range, 6-16 years) were acquired on a 3T MR imaging scanner. The data were registered to the PAM50 template in the Spinal Cord Toolbox. Next, the DTI indices for various WM regions were extracted at a single section centered at the C3 vertebral body in all the 30 subjects. Finally, an ANOVA test was performed to examine the effects of the following: 1) laterality, 2) functionality, and 3) age, with DTI-derived indices in 34 extracted WM regions., Results: A postprocessing pipeline was developed and validated to delineate pediatric spinal cord WM tracts. The results of ANOVA on fractional anisotropy values showed no effect for laterality ( P  = .72) but an effect for functionality ( P < .001) when comparing the 30 primary WM labels. There was a significant ( P  < .05) effect of age and maturity of the left spinothalamic tract on mean diffusivity, radial diffusivity, and axial diffusivity values., Conclusions: The proposed automated pipeline in this study incorporates unique postprocessing steps followed by template registration and quantification of DTI metrics using atlas-based regions. This method eliminates the need for manual ROI analysis of WM tracts and, therefore, increases the accuracy and speed of the measurements., (© 2021 by American Journal of Neuroradiology.)

Background and Purpose: The National Institute of Neurological Disorders and Stroke common data elements initiative was created to provide a consistent method for recording and reporting observations related to neurologic diseases in clinical trials. The purpose of this study is to validate the subset of common data elements related to MR imaging evaluation of acute spinal cord injury., Materials and Methods: Thirty-five cervical and thoracic MR imaging studies of patients with acute spinal cord injury were evaluated independently in 2 rounds by 5 expert reviewers. Intra- and interrater agreement were calculated for 17 distinct MR imaging observations related to spinal cord injury. These included ordinal, categoric, and continuous measures related to the length and location of spinal cord hemorrhage and edema as well as spinal canal and cord measurements. Level of agreement was calculated using the interclass correlation coefficient and kappa., Results: The ordinal common data elements spinal cord injury elements for lesion center and rostral or caudal extent of edema or hemorrhage demonstrated agreement ranging from interclass correlation coefficient 0.68 to 0.99. Reproducibility ranged from 0.95 to 1.00. Moderate agreement was observed for absolute length of hemorrhage and edema (0.54 to 0.60) with good reproducibility (0.78 to 0.83). Agreement for the Brain and Spinal Injury Center score showed the lowest interrater agreement with an overall kappa of 0.27 (0.20, 0.34). For 7 of the 8 variables related to spinal cord injury, agreement improved between the first and second evaluation. Continuous diameter measures of the spinal cord and spinal canal using interclass correlation coefficient varied substantially (0.23 to 0.83)., Conclusions: Agreement was more consistent for the ordinal measures of spinal cord injury than continuous measures. Good to excellent agreement on length and location of spinal cord hemorrhage and edema can be achieved with ordinal measures alone., (© 2021 by American Journal of Neuroradiology.)

Background and Purpose: Prior research has shown substantial shifts in procedure shares between specialty groups providing lumbar punctures. Our aim was to analyze national trends in lumbar punctures among the Medicare population from 2010 to 2018., Materials and Methods: Medicare Part B Physician/Supplier Procedure Summary Master Files from 2010 to 2018 were analyzed for all Current Procedural Terminology, Version 4 codes related to lumbar punctures (62270 and 62272). Lumbar puncture procedure volume and utilization rates were assessed and stratified by place of service and specialty background of the providers., Results: From 2010 to 2018, the overall number of lumbar puncture procedures essentially has not changed (92,579 versus 92,533). Radiologists hold the largest and an increasing procedure share of diagnostic and overall lumbar punctures (overall share, 45.7% in 2010 [ n  = 42,296] versus 52.3% in 2018 [ n  = 48,414]). Advanced practice providers have increased their procedure share (3.7% in 2010 [ n  = 3388] versus 8.4% in 2018 [ n  = 7785], + 129.8% procedure volume). Emergency medicine physicians and neurologists have a decreasing procedure share (21.8% versus 15.3% and 12.5% versus 8.8%, respectively). The inpatient hospital setting remains the largest place of service for lumbar punctures, recording a 5.3% increase in procedure share. The emergency department lumbar puncture volume has declined, with a 7.4% decrease in the overall procedure share. Similarly, the hospital outpatient department procedure volume has increased (+4%), while the private office volume has decreased (-1.7%)., Conclusions: During the past decade, lumbar puncture procedures among the Medicare population have remained stable, with a shift in procedure volume from the emergency department and private offices to the hospital setting, which has mainly affected radiologists and advanced practice providers., (© 2021 by American Journal of Neuroradiology.)

Fueled by new techniques, computational tools, and broader availability of imaging data, artificial intelligence has the potential to transform the practice of neuroradiology. The recent exponential increase in publications related to artificial intelligence and the central focus on artificial intelligence at recent professional and scientific radiology meetings underscores the importance. There is growing momentum behind leveraging artificial intelligence techniques to improve workflow and diagnosis and treatment and to enhance the value of quantitative imaging techniques. This article explores the reasons why neuroradiologists should care about the investments in new artificial intelligence applications, highlights current activities and the roles neuroradiologists are playing, and renders a few predictions regarding the near future of artificial intelligence in neuroradiology., (© 2020 by American Journal of Neuroradiology.)

The appropriate imaging of patients with headache presents a number of important and vexing challenges for clinicians. Despite a number of guidelines and studies demonstrating a lack of cost-effectiveness, clinicians continue to image patients with chronic nonfocal headaches, and the trend toward imaging is increasing. The reasons are complex and include the fear of missing a clinically significant lesion and litigation, habitual and standard of care practices, lack of tort reform, regulatory penalties and potential impact on one's professional reputation, patient pressures, and financial motivation. Regulatory and legislative reforms are needed to encourage best practices without fear of professional sanctions when following the guidelines. The value of negative findings on imaging tests requires better understanding because they appear to provide some measure of societal value. Clinical decision support tools and machine intelligence may offer additional guidance and improve quality and cost-efficient management of this challenging patient population., (© 2020 by American Journal of Neuroradiology.)

The American Society of Neuroradiology has teamed up with the American College of Radiology and the Radiological Society of North America to create a catalog of neuroradiology common data elements that addresses specific clinical use cases. Fundamentally, a common data element is a question, concept, measurement, or feature with a set of controlled responses. This could be a measurement, subjective assessment, or ordinal value. Common data elements can be both machine- and human-generated. Rather than redesigning neuroradiology reporting, the goal is to establish the minimum number of "essential" concepts that should be in a report to address a clinical question. As medicine shifts toward value-based service compensation methodologies, there will be an even greater need to benchmark quality care and allow peer-to-peer comparisons in all specialties. Many government programs are now focusing on these measures, the most recent being the Merit-Based Incentive Payment System and the Medicare Access Children's Health Insurance Program Reauthorization Act of 2015. Standardized or structured reporting is advocated as one method of assessing radiology report quality, and common data elements are a means for expressing these concepts. Incorporating common data elements into clinical practice fosters a number of very useful downstream processes including establishing benchmarks for quality-assurance programs, ensuring more accurate billing, improving communication to providers and patients, participating in public health initiatives, creating comparative effectiveness research, and providing classifiers for machine learning. Generalized adoption of the recommended common data elements in clinical practice will provide the means to collect and compare imaging report data from multiple institutions locally, regionally, and even nationally, to establish quality benchmarks., (© 2019 by American Journal of Neuroradiology.)

Background and Purpose: The zonally magnified oblique multislice EPI (ZOOM-EPI) diffusion-weighted sequence has been visually shown to provide superior MR diffusion image quality compared with the full-FOV single-shot EPI sequence (non-ZOOM-EPI) in the adult cervical spinal cord. The purpose of this study was to examine the diffusion tensor imaging indices in the normal human cervical spinal cord between ZOOMED and non-ZOOMED DTI acquisitions and determine whether DTI values are comparable between direct and indirect age-matched groups., Materials and Methods: Fifty-four subjects 23-58 years of age (9 direct age-matched and 45 indirect age-matched) were scanned using a 1.5T scanner. Diffusion tensor indices including fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity were generated from the DTI dataset. These DTI values were calculated for both ZOOM and non-ZOOM acquisitions and compared at each intervertebral disc level. The variability of the DTI values for ZOOM and non-ZOOM sequences was measured using a coefficient of variation within direct and indirect age-matched controls., Results: The mean diffusivity, axial diffusivity, and radial diffusivity values obtained along the cervical spinal cord in the age-matched controls showed a significant decrease using the ZOOM sequence ( P = .05, P = .002, and P < .001). Mean fractional anisotropy showed a significant increase ( P = .04) using the ZOOM sequence. The indirect age-matched controls showed a statistically significant increase in fractional anisotropy ( P = .03) and a decrease in mean diffusivity ( P = .002), axial diffusivity ( P < .001), and radial diffusivity ( P = .002) using the ZOOM sequence. Less variability has been shown in DTI using the ZOOM sequence compared with the non-ZOOM sequence in both direct and indirect age groups. The ZOOM sequence exhibited higher SNR (SNR ZOOM = 22.84 ± 7.59) compared with the non-ZOOM sequence (SNR non-ZOOM = 19.7 ± 7.05). However, when we used a 2-tailed t test assuming unequal variances, the ZOOM sequence did not demonstrate a statistically significant increase., Conclusions: ZOOM DTI acquisition methods provide superior image quality and precision over non-ZOOM techniques and are recommended over conventional full-FOV single-shot EPI DTI for clinical applications in cervical spinal cord imaging., (© 2018 by American Journal of Neuroradiology.)

Background and Purpose: More than 75 million Americans have less than adequate health literacy skills according to the National Center for Education Statistics. Readability scores are used as a measure of how well populations read and understand patient education materials. The purpose of this study was to assess the readability of Web sites dedicated to patient education for radiologic spine imaging and interventions., Materials and Methods: Eleven search terms relevant to radiologic spine imaging were searched on the public Internet, and the top 10 links for each term were collected and analyzed to determine readability scores by using 10 well-validated quantitative readability assessments from patient-centered education Web sites. The search terms included the following: x-ray spine, CT spine, MR imaging spine, lumbar puncture, kyphoplasty, vertebroplasty, discogram, myelogram, cervical spine, thoracic spine, and lumbar spine., Results: Collectively, the 110 articles were written at an 11.3 grade level (grade range, 7.1-16.9). None of the articles were written at the American Medical Association and National Institutes of Health recommended 3rd-to-7th grade reading levels. The vertebroplasty articles were written at a statistically significant ( P < .05) more advanced level than the articles for x-ray spine, CT spine, and MR imaging spine., Conclusions: Increasing use of the Internet to obtain health information has made it imperative that on-line patient education be written for easy comprehension by the average American. However, given the discordance between readability scores of the articles and the American Medical Association and National Institutes of Health recommended guidelines, it is likely that many patients do not fully benefit from these resources., (© 2017 by American Journal of Neuroradiology.)

Background and Purpose: How early spinal cord injury (SCI) lesions evolve in patients after injury is unknown. The purpose of this study was to characterize the early evolution of spinal cord edema and hemorrhage on MR imaging after acute traumatic SCI., Materials and Methods: We performed a retrospective analysis of 48 patients with clinically complete cervical spine injury. Inclusion criteria were the clear documentation of the time of injury and MR imaging before surgical intervention within 72 hours of injury. The length of intramedullary spinal cord edema and hemorrhage was assessed. The correlation between time to imaging and lesion size was determined by multiple regression analysis. Short-interval follow-up MR imaging was also available for a few patients (n = 5), which allowed the direct visualization of changes in spinal cord edema., Results: MR imaging demonstrated cord edema in 100% of patients and cord hemorrhage in 67% of patients. The mean longitudinal length of cord edema was 10.3 +/- 4.0 U, and the mean length of cord hemorrhage was 2.6 +/- 2.0 U. Increased time to MR imaging correlated to increased spinal cord edema length (P = .002), even after accounting for the influence of other variables. A difference in time to MR imaging of 1.2 days corresponded to an average increase in cord edema by 1 full vertebral level. Hemorrhage length was not affected by time to imaging (P = .825). A temporal increase in the length of spinal cord edema was confirmed in patients with short-interval follow-up MR imaging (P = .003)., Conclusion: Spinal cord edema increases significantly during the early time period after injury, whereas intramedullary hemorrhage is comparatively static.

BK virus infection is most often associated with urologic disease in patients who have undergone renal or bone marrow transplantation. We report a rare case of biopsy-confirmed BK virus encephalitis in an immunocompromised patient with hemorrhagic cystitis, in whom dramatic imaging findings were present despite relatively mild clinical symptoms. MR imaging demonstrated widespread increased signal intensity on T2- and fluid-attenuated inversion recovery-weighted images, with restricted diffusion, in the cerebellum, cerebral white matter, and deep gray matter structures. The simultaneous presence of urologic abnormalities and neurologic deficits in certain immunocompromised patients should prompt consideration of BK virus encephalitis.

Background and Purpose: The incidence of blunt traumatic vertebral artery dissection/thrombosis varies widely in published trauma series and is associated with spinal trauma. The purpose of this study was to determine the frequency of traumatic vertebral artery thrombosis (VAT) in cervically injured patients by using routine MR angiography (MRA) and MR imaging and identify associations with the severity of neurologic injury., Methods: A retrospective review of 1283 patients with nonpenetrating cervical spine fractures with or without an associated spinal cord injury (SCI) was performed. Imaging consisted of routine cervical MR imaging and 2D time-of-flight MRA of the neck. The cervical injury level, neurologic level of injury, and American Spinal Injury Association (ASIA) grade were recorded., Results: In this study, 632 patients met the inclusion criteria, 83 (13%) of whom had VAT on the admission MR imaging/MRA. Fifty-nine percent (49/83) of VAT patients had an associated SCI. VAT was significantly more common in motor-complete patients (ASIA A and B, 20%) than in neurologically intact (ASIA E, 11%) cervical spine-injured patients (P = .019). VAT incidence was not significantly different between motor-incomplete (ASIA C and D, 10%) and neurologically intact (ASIA E, 11%) cervical spine-injured patients (P = .840)., Conclusion: The absence of neurologic symptoms in a patient with cervical spine fracture does not preclude VAT. VAT associated with cervical spinal injury occurs with similar frequency in both neurologically intact (ASIA E) and motor-incomplete patients (ASIA C and D) but is significantly more common in motor-complete SCI (ASIA A and B).

Background and Purpose: Although research with functional MR imaging of the brain has proliferated over the past 5 years, technical limitations, such as motion, chemical shift, and susceptibility artifacts, have impeded such research in the human spinal cord. The purpose of this investigation was to determine whether a reliable functional MR imaging signal can be elicited from the cervical spinal cord during simple motor activity., Methods: Subjects performed three different motor tasks that activate different segments of the spinal cord. Gradient-echo-planar imaging on a 1.5-T clinical unit was used to image cervical spinal cords of human subjects. Another group of subjects was imaged while performing isometric exercise to study the relationship between the blood oxygenation level-dependent (BOLD) signal and applied force., Results: Task-dependent BOLD activity was detected in all subjects. Signal amplitude varied between 0.5% and 7%. Moreover, a linear relationship was found between the applied force and the signal amplitude during isometric exercise. While regions of activation were distributed throughout the spinal cord, concentrated activity was found in the anatomic locations of expected motor innervation., Conclusion: The functional MR imaging signal can be reliably detected with motor activity in the human cervical spinal cord on a 1.5-T clinical unit. The location of neural activation has an anatomic correspondence to the myotome in use. The strength of the BOLD signal is directly proportional to the level of muscular activity.

Background and Purpose: The appearance of the damaged spinal cord after injury correlates with initial neurologic deficit, as determined by the American Spinal Injury Association grade and manual muscle test score, as well as with recovery, as assessed by manual muscle test scores. The purpose of this study was to determine whether the presence of spinal cord hemorrhage and the size and location of spinal cord edema on MR images is predictive of functional recovery in survivors of cervical spinal cord injury (SCI)., Methods: The degree of damage to the cervical spinal cord was measured on the MR images of 49 patients who underwent imaging within 72 hours of sustaining SCI. The effects of hemorrhage and length/location of edema on changes in the value of the motor scale of the functional independence measure (FIM) were assessed on admission to and discharge from rehabilitation., Results: Patients without spinal cord hemorrhage had significant improvement in self-care and mobility scores compared with patients with hemorrhage. There was no significant effect of spinal cord hemorrhage on changes in locomotion and sphincter control scores. The rostral limit of edema positively correlated with admission and discharge self-care scores and with admission mobility and locomotion scores. Edema length had a negative correlation with all FIM scales at admission and discharge., Conclusion: The imaging characteristics of cervical SCI (hemorrhage and edema) are related to levels of physical recovery as determined by the FIM scale. Imaging factors that correlate with poor functional recovery are hemorrhage, long segments of edema, and high cervical locations.

A case of malignant optic glioma of adulthood is imaged in its early and late stages with high-resolution MR. The images show the mass to arise from the right optic nerve before invasion of the optic chiasm.

We describe a case of acute transverse myelitis after the administration of the recombinant form of hepatitis B vaccine. Abnormal enhancement of MR imaging accompanied residual neurologic deficit.

MR appearance of two patients with large, orbital conjunctival epithelium-lined inclusion cysts are presented. Both were complications of ophthalmic surgical procedures that necessitated conjunctival incision.

Purpose: To evaluate the safety and efficacy of gadoteridol as an MR contrast agent in patients with suspected head and neck pathology., Methods: One hundred thirty-three adult patients were studied with MR at 12 clinical trial sites before and after the intravenous administration of 0.10 mmol/kg gadoteridol. After enrollment, patients underwent a complete physical examination. Patient monitoring included vital signs, laboratory tests, and recording of the incidence and duration of adverse events. To evaluate efficacy, postcontrast T1-weighted images were compared with precontrast T1- and T2-weighted images. Investigators having clinical information evaluated 129 studies for efficacy; two readers blinded to clinical information subsequently evaluated 122 studies using the same criteria as the investigators., Results: Eight patients (6.0%) experienced mild adverse events possibly or probably related to contrast administration, all of which resolved without treatment. Two clinically significant laboratory abnormalities considered related or possibly related to the administration of gadoteridol were reported in two patients. Enhancement of pathology was seen in 82.9% of cases evaluated by investigators at the study site and in 78.7% of cases subsequently evaluated by the blinded readers. Postcontrast images were judged by investigators to provide additional diagnostic information over precontrast images in 68.9% of studies. The additional diagnostic information available on postcontrast studies most often consisted of improved visualization of pathology and better definition of lesion borders. The use of this information might have contributed to a change in patient diagnosis in 18.6% of the cases evaluated by the investigators and in 16.4% of the cases reviewed by the blinded readers., Conclusions: Preliminary results show gadoteridol to be a safe and efficacious contrast agent for enhanced MR study of extracranial and extraspinal head and neck pathology.

We report three cases that reveal an array of etiologic and radiologic findings associated with dissection of the proximal segment of the vertebral arteries. Regardless of etiology, the proximal segment may be the principal site of dissection in these vessels.

We retrospectively reviewed the routine spin-echo MR studies of the brain in 12 patients with 13 angiographically demonstrated occlusions and in 14 patients with 16 high-grade stenoses of the carotid arteries. Intraluminal signal that was isointense with adjacent brain on long TR/short TE and long TR/long TE images was 100% specific for atherosclerotic occlusion. Of the 13 proved occlusions, six (46%) had significant degrees of hyperintense intraluminal signal indistinguishable from that observed consequent to slow flow distal to high-grade stenoses. MR detected only five (31%) of the 16 proved high-grade stenoses. Normal flow void does not exclude significant extracranial carotid stenosis. Occlusion cannot always be distinguished from high-grade stenosis when hyperintense intraluminal signal is encountered. However, a reliable diagnosis of atherosclerotic occlusion can be made when isointense intraluminal signal is observed.