Imaging assessment of spine infection.

This article comprehensively reviews current imaging concepts in spinal infection with primary focus on infectious spondylodiscitis (IS) as well as the less common entity of facet joint septic arthritis (FSA). This review encompasses the multimodality imaging appearances (radiographs, CT, MRI, and nuclear imaging) of spinal infection-both at initial presentation and during treatment-to aid the radiologist in guiding diagnosis and successful management. We discuss the pathophysiology of spinal infection in various patient populations (including the non-instrumented and postoperative spine) as well as the role of imaging-guided biopsy. We also highlight several non-infectious entities that can mimic IS (both clinically and radiologically) that should be considered during image interpretation to avoid misdiagnosis. These potential mimics include the following: Modic type 1 degenerative changes, acute Schmorl's node, neuropathic spondyloarthropathy, radiation osteitis, and inflammatory spondyloarthropathy (SAPHO syndrome).

How Does Pruning Impact Long-Tailed Multi-Label Medical Image Classifiers?

Pruning has emerged as a powerful technique for compressing deep neural networks, reducing memory usage and inference time without significantly affecting overall performance. However, the nuanced ways in which pruning impacts model behavior are not well understood, particularly for long-tailed, multi-label datasets commonly found in clinical settings. This knowledge gap could have dangerous implications when deploying a pruned model for diagnosis, where unexpected model behavior could impact patient well-being. To fill this gap, we perform the first analysis of pruning's effect on neural networks trained to diagnose thorax diseases from chest X-rays (CXRs). On two large CXR datasets, we examine which diseases are most affected by pruning and characterize class "forgettability" based on disease frequency and co-occurrence behavior. Further, we identify individual CXRs where uncompressed and heavily pruned models disagree, known as pruning-identified exemplars (PIEs), and conduct a human reader study to evaluate their unifying qualities. We find that radiologists perceive PIEs as having more label noise, lower image quality, and higher diagnosis difficulty. This work represents a first step toward understanding the impact of pruning on model behavior in deep long-tailed, multi-label medical image classification. All code, model weights, and data access instructions can be found at https://github.com/VITA-Group/PruneCXR.

How Does Pruning Impact Long-Tailed Multi-label Medical Image Classifiers?

Pruning has emerged as a powerful technique for compressing deep neural networks, reducing memory usage and inference time without significantly affecting overall performance. However, the nuanced ways in which pruning impacts model behavior are not well understood, particularly for long-tailed, multi-label datasets commonly found in clinical settings. This knowledge gap could have dangerous implications when deploying a pruned model for diagnosis, where unexpected model behavior could impact patient well-being. To fill this gap, we perform the first analysis of pruning's effect on neural networks trained to diagnose thorax diseases from chest X-rays (CXRs). On two large CXR datasets, we examine which diseases are most affected by pruning and characterize class "forgettability" based on disease frequency and co-occurrence behavior. Further, we identify individual CXRs where uncompressed and heavily pruned models disagree, known as pruning-identified exemplars (PIEs), and conduct a human reader study to evaluate their unifying qualities. We find that radiologists perceive PIEs as having more label noise, lower image quality, and higher diagnosis difficulty. This work represents a first step toward understanding the impact of pruning on model behavior in deep long-tailed, multi-label medical image classification. All code, model weights, and data access instructions can be found at https://github.com/VITA-Group/PruneCXR.

Adherus Dural Sealant in Endoscopic Skull Base Surgery: Safety, Imaging Characteristics, and Sinonasal Quality of Life.

Objective This study was aimed to compare the safety profiles, magnetic resonance imaging (MRI) findings, and sinonasal outcome test (SNOT-22) scores of Adherus dural sealant, a novel tissue glue designed for skull base surgery. Design Present study is a prospective case series. Setting The research work took place at a tertiary-care academic medical center. Participants Consecutive series of 26 patients undergoing endoscopic skull base surgery (ESBS) with Adherus was compared with a control group of 24 patients matched for tumor type and size with DuraSeal as a sealant. Main Outcome Measures Postoperative complication rates, imaging characteristics, and postoperative SNOT-22 scores were measured and compared. Results No postoperative cerebrospinal fluid (CSF) leaks, intracranial hemorrhages, or mucoceles were observed in either cohort. Adherus was more likely to be identifiable on immediate postoperative MRI (50 vs. 20.8%, p = 0.032). In patients in whom a nasoseptal flap was utilized, the flap was opposed to the skull base in all cases regardless of sealant selected. Postoperative SNOT-22 total (17.25 [±10.81] vs. 14.85 [±14.22], p = 0.609) and subdomain scores were similar between the two groups. Conclusions Adherus dural sealant appears to be a safe alternative to Duraseal in ESBS with comparable quality of life outcomes and imaging findings. These preliminary results are promising but should be examined in a larger population with long-term follow-up.

Predictors of postoperative motor function in rolandic meningiomas.

OBJECTIVE: Resection of supratentorial meningiomas is generally considered a low-risk procedure, but tumors involving the rolandic cortex present a unique challenge. The rate of motor function deterioration associated with resecting such tumors is not well described in the literature. Thus, the authors sought to report the rates and predictors of postoperative motor deficit following the resection of rolandic meningiomas to assist with patient counseling and surgical decision-making. METHODS: An institution's pathology database was screened for meningiomas removed between 2000 and 2017, and patients with neuroradiological evidence of rolandic involvement were identified. Parameters screened as potential predictors included patient age, sex, preoperative motor severity, tumor location, tumor origin (falx vs convexity), histological grade, FLAIR signal (T2-weighted MRI), venous involvement (T1-weighted MRI with contrast), intratumoral hemorrhage, embolization, and degree of resection (Simpson grade). Variables of interest included preoperative weakness and postoperative motor decline (novel or worsened permanent deficit). The SPSS univariate and bivariate analysis functions were used, and statistical significance was determined with alpha < 0.05. RESULTS: In 89 patients who had undergone resection of convexity (80.9%) or parasagittal (19.1%) rolandic meningiomas, a postoperative motor decline occurred in 24.7%. Of 53 patients (59.6%) with preoperative motor deficits, 60.3% improved, 13.2% were unchanged, and 26.4% worsened following surgery. Among the 36 patients without preoperative deficits, 22.2% developed new weakness. Predictors of preoperative motor deficit included tumor size (41.6 vs 33.2 cm3, p = 0.040) and presence of FLAIR signal (69.8% vs 50.0%, p = 0.046). Predictors of postoperative motor decline were preoperative motor deficit (47.2% vs 22.2%, p = 0.017), minor (compared with severe) preoperative weakness (25.6% vs 21.4%, p < 0.001), and preoperative embolization (54.5% vs 20.5%, p = 0.014). Factors that trended toward significance included parafalcine tumor origin (41.2% vs 20.8% convexity, p = 0.08), significant venous involvement (44.4% vs 23.5% none, p = 0.09), and Simpson grade II+ (34.2% vs 17.6% grade I, p = 0.07). CONCLUSIONS: Resection of rolandic area meningiomas carries a high rate of postoperative morbidity and deserves special preoperative planning. Large tumor size, peritumoral edema, preoperative embolization, parafalcine origin, and venous involvement may further increase the risk. Alternative surgical strategies, such as aggressive internal debulking, may prevent motor decline in a subset of high-risk patients.

Emergency Radiology "Boot Camp": Educating Emergency Medicine Residents Using E-learning Radiology Modules.

OBJECTIVES: There is an overall paucity of literature on the radiologic education of emergency medicine (EM) clinicians. Given the fact that many EM clinicians preliminarily review images for their patients, we hypothesized that a brief imaging curriculum could be efficacious in teaching basic and relevant radiologic interpretation. METHODS: We designed a 4-hour "radiology boot camp" for a group of 20 EM residents (from all years of training) covering several subject specific e-learning modules. They completed precourse and postcourse quizzes to evaluate the efficacy of these modules. These modules included interactive PowerPoint-based tutorials, games, and imaging decision support simulators. Matched results from the pre- and posttests were analyzed using paired t test. An additional questionnaire was administered to the EM residents to evaluate their perception of the educational experience. RESULTS: The precourse and postcourse quizzes demonstrated a statistically significant level of improved knowledge due to the educational modules (p < 0.0001). In addition, all of the participants believed the modules were a useful learning experience (100%) and a vast majority described them as a valuable resource for future reference (95%). CONCLUSION: We demonstrate a model for providing an easy and effective means of educating EM residents on basic imaging interpretation and utilization, using e-learning modules.

Innovative Hypofractionated Stereotactic Regimen Achieves Excellent Local Control with No Radiation Necrosis: Promising Results in the Management of Patients with Small Recurrent Inoperable GBM.

Management of recurrent glioblastoma multiforme (GBM) remains a challenge. Several institutions reported that a single fraction of ≥ 20 Gy for small tumor burden results in excellent local control; however, this is at the expense of a high incidence of radiation necrosis (RN). Therefore, we developed a hypofractionation pattern of 33 Gy/3 fractions, which is a radiobiological equivalent of 20 Gy, with the aim to lower the incidence of RN. We reviewed records of 21 patients with recurrent GBM treated with hypofractionated stereotactic radiation therapy (HFSRT) to their 22 respective lesions. Sixty Gy fractioned external beam radiotherapy was performed as first-line treatment. Median time from primary irradiation to HFSRT was 9.6 months (range: 3.1 - 68.1 months). In HFSRT, a median dose of 33 Gy in 11 Gy fractions was delivered to the 80% isodose line that encompassed the target volume. The median tumor volume was 1.07 cm3 (range: 0.11 - 16.64 cm3). The median follow-up time after HFSRT was 9.3 months (range: 1.7 - 33.6 months). Twenty-one of 23 lesions treated (91.3%) achieved local control while 2/23 (8.7%) progressed. Median time to progression outside of the treated site was 5.2 months (range: 2.2 - 9.6 months). Progression was treated with salvage chemotherapy. Five of 21 patients (23.8%) were alive at the end of this follow-up; two patients remain disease-free. The remaining 16/21 patients (76.2%) died of disease. Treatment was well tolerated by all patients with no acute CTC/RTOG > Grade 2. There was 0% incidence of RN. A prospective trial will be underway to validate these promising results.

Can left ventricular function be assessed on non-ECG-gated CT?

To evaluate whether non-gated computed tomography (CT) can assess left ventricular (LV) function, 101 patients with both CT and echocardiography were selected, with ejection fraction <50% on echocardiography used as a reference standard. CTs were blindly reevaluated, and qualitative assessment of LV dysfunction on CT correlated with echocardiographic dysfunction, odds ratio of 21.0 (95% confidence interval=6.55-71.0), specificity of 86% (56/65). Systolic and diastolic images were identified on CT, and the ratio of systolic to diastolic LV internal diameters and ratio of LV to RV internal diameter were determined, both showing correlation with LV dysfunction on echocardiography (P<.0001). Non-gated CT can be used to predict LV dysfunction.

Utilization guidelines for reducing radiation exposure in the evaluation of aneurysmal subarachnoid hemorrhage: A practice quality improvement project.

OBJECTIVE: The purpose of this study was to reduce the cumulative radiation exposure from CT of patients with aneurysmal subarachnoid hemorrhage. MATERIALS AND METHODS: Baseline data on 30 patients with aneurysmal subarachnoid hemorrhage were collected retrospectively for all CT examinations of the head performed throughout the hospital course. Radiation exposure estimates were obtained by recording dose-length products for each examination. As a departmental practice quality improvement project, an imaging protocol was implemented that included utilization guidelines to reduce radiation exposure in CTA and CT perfusion examinations performed to detect vasospasm in patients with aneurysmal subarachnoid hemorrhage. Ten months after implementation of this protocol, data on 30 additional patients were analyzed. Means, medians, and SD estimates were compared for cumulative radiation exposure and absolute numbers of each examination. RESULTS: Sixty patients were included in the study: 30 patients at baseline and 30 patients after implementation of the quality improvement plan. These patients underwent 435 CT examinations: 248 examinations at baseline and 187 examinations with the new protocol. With the new algorithm, the mean number of CT examinations per patient was 5.8 compared with 7.8 at baseline, representing a decrease of 25.6%. The number of CT perfusion examinations per patient decreased 32.1%. Overall, there was a 12.1% decrease in cumulative radiation exposure (p > 0.05). CONCLUSION: With the structured imaging algorithm, the cumulative radiation exposure and number of CT examinations of the head decreased among patients with aneurysmal subarachnoid hemorrhage because utilization guidelines defined the appropriate imaging time points for detection of vasospasm. Application of these methods to other patient populations with high use of CT may reduce cumulative radiation exposure while the clinical benefits of imaging are maintained.