Advanced Imaging of the Peripheral Nerves, From the AJR "How We Do It" Special Series.

Advanced imaging of peripheral nerves is occupying an increasingly important role in the diagnostic workup of peripheral nerve disorders. Advances in MR neurography (MRN) and high-resolution ultrasound have addressed historical challenges in peripheral nerve imaging related to nerves' small size and non-linear course, and difficult differentiation from surrounding tissues. Modern MRN depicts neuromuscular anatomy with exquisite contrast resolution, and MRN has become the workhorse imaging modality for peripheral nerve evaluation. MRN protocols vary across institutions and are adjusted in individual patients, although commonly include nerve-selective sequences and diffusion tensor imaging tractography. Ultrasound offers a dynamic, real-time high-resolution assessment of peripheral nerves, and is widely accessible and less costly than MRN. Ultrasound has greater ability to interrogate peripheral nerves at the fascicular level and provides complementary information to MRN. However, ultrasound of peripheral nerves requires substantial skill and experience and is operator-dependent. The two modalities have distinct advantages and disadvantages, and the selection between these depends on the clinical context. This article provides an overview of advanced imaging techniques used for evaluation of peripheral nerves, with attention to MRN and high-resolution ultrasound. We draw on our institutional experience in performing both modalities to highlight technical considerations for optimizing examinations.

A 20-year Follow-up of the International Early Lung Cancer Action Program (I-ELCAP).

Background The low-dose CT (≤3 mGy) screening report of 1000 Early Lung Cancer Action Program (ELCAP) participants in 1999 led to the International ELCAP (I-ELCAP) collaboration, which enrolled 31 567 participants in annual low-dose CT screening between 1992 and 2005. In 2006, I-ELCAP investigators reported the 10-year lung cancer-specific survival of 80% for 484 participants diagnosed with a first primary lung cancer through annual screening, with a high frequency of clinical stage I lung cancer (85%). Purpose To update the cure rate by determining the 20-year lung cancer-specific survival of participants diagnosed with first primary lung cancer through annual low-dose CT screening in the expanded I-ELCAP cohort. Materials and Methods For participants enrolled in the HIPAA-compliant prospective I-ELCAP cohort between 1992 and 2022 and observed until December 30, 2022, Kaplan-Meier survival analysis was used to determine the 10- and 20-year lung cancer-specific survival of participants diagnosed with first primary lung cancer through annual low-dose CT screening. Eligible participants were aged at least 40 years and had current or former cigarette use or had never smoked but had been exposed to secondhand tobacco smoke. Results Among 89 404 I-ELCAP participants, 1257 (1.4%) were diagnosed with a first primary lung cancer (684 male, 573 female; median age, 66 years; IQR, 61-72), with a median smoking history of 43.0 pack-years (IQR, 29.0-60.0). Median follow-up duration was 105 months (IQR, 41-182). The frequency of clinical stage I at pretreatment CT was 81% (1017 of 1257). The 10-year lung cancer-specific survival of 1257 participants was 81% (95% CI: 79, 84) and the 20-year lung cancer-specific survival was 81% (95% CI: 78, 83), and it was 95% (95% CI: 91, 98) for 181 participants with pathologic T1aN0M0 lung cancer. Conclusion The 10-year lung cancer-specific survival of 80% reported in 2006 for I-ELCAP participants enrolled in annual low-dose CT screening and diagnosed with a first primary lung cancer has persisted, as shown by the updated 20-year lung cancer-specific survival for the expanded I-ELCAP cohort. © RSNA, 2023 See also the editorials by Grenier and by Sequist and Olazagasti in this issue.

Deep learning for automated, interpretable classification of lumbar spinal stenosis and facet arthropathy from axial MRI.

OBJECTIVES: To evaluate a deep learning model for automated and interpretable classification of central canal stenosis, neural foraminal stenosis, and facet arthropathy from lumbar spine MRI. METHODS: T2-weighted axial MRI studies of the lumbar spine acquired between 2008 and 2019 were retrospectively selected (n = 200) and graded for central canal stenosis, neural foraminal stenosis, and facet arthropathy. Studies were partitioned into patient-level train (n = 150), validation (n = 20), and test (n = 30) splits. V-Net models were first trained to segment the dural sac and the intervertebral disk, and localize facet and foramen using geometric rules. Subsequently, Big Transfer (BiT) models were trained for downstream classification tasks. An interpretable model for central canal stenosis was also trained using a decision tree classifier. Evaluation metrics included linearly weighted Cohen's kappa score for multi-grade classification and area under the receiver operator characteristic curve (AUROC) for binarized classification. RESULTS: Segmentation of the dural sac and intervertebral disk achieved Dice scores of 0.93 and 0.94. Localization of foramen and facet achieved intersection over union of 0.72 and 0.83. Multi-class grading of central canal stenosis achieved a kappa score of 0.54. The interpretable decision tree classifier had a kappa score of 0.80. Pairwise agreement between readers (R1, R2), (R1, R3), and (R2, R3) was 0.86, 0.80, and 0.74. Binary classification of neural foraminal stenosis and facet arthropathy achieved AUROCs of 0.92 and 0.93. CONCLUSION: Deep learning systems can be performant as well as interpretable for automated evaluation of lumbar spine MRI including classification of central canal stenosis, neural foraminal stenosis, and facet arthropathy. KEY POINTS: • Interpretable deep-learning systems can be developed for the evaluation of clinical lumbar spine MRI. Multi-grade classification of central canal stenosis with a kappa of 0.80 was comparable to inter-reader agreement scores (0.74, 0.80, 0.86). Binary classification of neural foraminal stenosis and facet arthropathy achieved favorable and accurate AUROCs of 0.92 and 0.93, respectively. • While existing deep-learning systems are opaque, leading to clinical deployment challenges, the proposed system is accurate as well as interpretable, providing valuable information to a radiologist in clinical practice.

Deep Learning for Multi-Tissue Segmentation and Fully Automatic Personalized Biomechanical Models from BACPAC Clinical Lumbar Spine MRI.

STUDY DESIGN: In vivo retrospective study of fully automatic quantitative imaging feature extraction from clinically acquired lumbar spine magnetic resonance imaging (MRI). OBJECTIVE: To demonstrate the feasibility of substituting automatic for human-demarcated segmentation of major anatomic structures in clinical lumbar spine MRI to generate quantitative image-based features and biomechanical models. SETTING: Previous studies have demonstrated the viability of automatic segmentation applied to medical images; however, the feasibility of these networks to segment clinically acquired images has not yet been demonstrated, as they largely rely on specialized sequences or strict quality of imaging data to achieve good performance. METHODS: Convolutional neural networks were trained to demarcate vertebral bodies, intervertebral disc, and paraspinous muscles from sagittal and axial T1-weighted MRIs. Intervertebral disc height, muscle cross-sectional area, and subject-specific musculoskeletal models of tissue loading in the lumbar spine were then computed from these segmentations and compared against those computed from human-demarcated masks. RESULTS: Segmentation masks, as well as the morphological metrics and biomechanical models computed from those masks, were highly similar between human- and computer-generated methods. Segmentations were similar, with Dice similarity coefficients of 0.77 or greater across networks, and morphological metrics and biomechanical models were similar, with Pearson R correlation coefficients of 0.69 or greater when significant. CONCLUSIONS: This study demonstrates the feasibility of substituting computer-generated for human-generated segmentations of major anatomic structures in lumbar spine MRI to compute quantitative image-based morphological metrics and subject-specific musculoskeletal models of tissue loading quickly, efficiently, and at scale without interrupting routine clinical care.

Technology and Tool Development for BACPAC: Qualitative and Quantitative Analysis of Accelerated Lumbar Spine MRI with Deep-Learning Based Image Reconstruction at 3T.

OBJECTIVES: To evaluate whether combining fast acquisitions with deep-learning reconstruction can provide diagnostically useful images and quantitative assessment comparable to standard-of-care acquisitions for lumbar spine magnetic resonance imaging (MRI). METHODS: Eighteen patients were imaged with both standard protocol and fast protocol using reduced signal averages, each protocol including sagittal fat-suppressed T2-weighted, sagittal T1-weighted, and axial T2-weighted 2D fast spin-echo sequences. Fast-acquisition data was additionally reconstructed using vendor-supplied deep-learning reconstruction with three different noise reduction factors. For qualitative analysis, standard images as well as fast images with and without deep-learning reconstruction were graded by three radiologists on five different categories. For quantitative analysis, convolutional neural networks were applied to sagittal T1-weighted images to segment intervertebral discs and vertebral bodies, and disc heights and vertebral body volumes were derived. RESULTS: Based on noninferiority testing on qualitative scores, fast images without deep-learning reconstruction were inferior to standard images for most categories. However, deep-learning reconstruction improved the average scores, and noninferiority was observed over 24 out of 45 comparisons (all with sagittal T2-weighted images while 4/5 comparisons with sagittal T1-weighted and axial T2-weighted images). Interobserver variability increased with 50 and 75% noise reduction factors. Deep-learning reconstructed fast images with 50% and 75% noise reduction factors had comparable disc heights and vertebral body volumes to standard images (r2≥ 0.86 for disc heights and r2≥ 0.98 for vertebral body volumes). CONCLUSIONS: This study demonstrated that deep-learning-reconstructed fast-acquisition images have the potential to provide noninferior image quality and comparable quantitative assessment to standard clinical images.

Magnetic Resonance Imaging of the Lumbar Spine: Recommendations for Acquisition and Image Evaluation from the BACPAC Spine Imaging Working Group.

Management of patients suffering from low back pain (LBP) is challenging and requires development of diagnostic techniques to identify specific patient subgroups and phenotypes in order to customize treatment and predict clinical outcome. The Back Pain Consortium (BACPAC) Research Program Spine Imaging Working Group has developed standard operating procedures (SOPs) for spinal imaging protocols to be used in all BACPAC studies. These SOPs include procedures to conduct spinal imaging assessments with guidelines for standardizing the collection, reading/grading (using structured reporting with semi-quantitative evaluation using ordinal rating scales), and storage of images. This article presents the approach to image acquisition and evaluation recommended by the BACPAC Spine Imaging Working Group. While the approach is specific to BACPAC studies, it is general enough to be applied at other centers performing magnetic resonance imaging (MRI) acquisitions in patients with LBP. The herein presented SOPs are meant to improve understanding of pain mechanisms and facilitate patient phenotyping by codifying MRI-based methods that provide standardized, non-invasive assessments of spinal pathologies. Finally, these recommended procedures may facilitate the integration of better harmonized MRI data of the lumbar spine across studies and sites within and outside of BACPAC studies.

Diagnostic Performance of CT-Guided Bone Biopsies in Patients with Suspected Osteomyelitis of the Appendicular and Axial Skeleton with a Focus on Clinical and Technical Factors Associated with Positive Microbiology Culture Results.

PURPOSE: To assess diagnostic performance of CT-guided percutaneous needle bone biopsy (CTNBB) in patients with suspected osteomyelitis and analyze whether certain clinical or technical factors were associated with positive microbiology results. MATERIALS AND METHODS: All CTNBBs performed in a single center for suspected osteomyelitis of the appendicular and axial skeleton during 2003-2018 were retrospectively reviewed. Specific inclusion criteria were clinical and radiologic suspicion of osteomyelitis. Standard of reference was defined using outcome of surgical histopathology and microbiology culture and clinical and imaging follow-up. Technical and clinical data (needle size, comorbidities, clinical factors, laboratory values, blood cultures) were collected. Logistic regression was performed to assess associations between technical and clinical data and microbiology biopsy outcome. RESULTS: A total of 142 CTNBBs were included (46.5% female patients; age ± SD 46.10 y ± 22.8), 72 (50.7%) from the appendicular skeleton and 70 (49.3%) from the axial skeleton. CTNBB showed a sensitivity of 42.5% (95% confidence interval [CI], 32.0%-53.6%) in isolating the causative pathogen. A higher rate of positive microbiology results was found in patients with intravenous drug use (odds ratio [OR] = 5.15; 95% CI, 1.2-21.0; P = .022) and elevated white blood cell count ≥ 10 × 109/L (OR = 3.9; 95% CI, 1.62-9.53; P = .002). Fever (≥ 38°C) was another clinical factor associated with positive microbiology results (OR = 3.6; 95% CI, 1.3-9.6; P = .011). CONCLUSIONS: CTNBB had a low sensitivity of 42.5% for isolating the causative pathogen. Rate of positive microbiology samples was significantly higher in patients with IV drug use, elevated white blood cell count, and fever.

Neurogenic thoracic outlet syndrome: current diagnostic criteria and advances in MRI diagnostics.

Neurogenic thoracic outlet syndrome (nTOS) is caused by compression of the brachial plexus as it traverses from the thoracic outlet to the axilla. Diagnosing nTOS can be difficult because of overlap with other complex pain and entrapment syndromes. An nTOS diagnosis is made based on patient history, physical exam, electrodiagnostic studies, and, more recently, interpretation of MR neurograms with tractography. Advances in high-resolution MRI and tractography can confirm an nTOS diagnosis and identify the location of nerve compression, allowing tailored surgical decompression. In this report, the authors review the current diagnostic criteria, present an update on advances in MRI, and provide case examples demonstrating how MR neurography (MRN) can aid in diagnosing nTOS. The authors conclude that improved high-resolution MRN and tractography are valuable tools for identifying the source of nerve compression in patients with nTOS and can augment current diagnostic modalities for this syndrome.

Visualization of nerve fibers and their relationship to peripheral nerve tumors by diffusion tensor imaging.

OBJECT The majority of growing and/or symptomatic peripheral nerve tumors are schwannomas and neurofibromas. They are almost always benign and can usually be resected while minimizing motor and sensory deficits if approached with the proper expertise and techniques. Intraoperative electrophysiological stimulation and recording techniques allow the surgeon to map the surface of the tumor in an effort to identify and thus avoid damaging functioning nerve fibers. Recently, MR diffusion tensor imaging (DTI) techniques have permitted the visualization of axons, because of their anisotropic properties, in peripheral nerves. The object of this study was to compare the distribution of nerve fibers as revealed by direct electrical stimulation with that seen on preoperative MR DTI. METHODS The authors conducted a retrospective chart review of patients with a peripheral nerve or nerve root tumor between March 2012 and January 2014. Diffusion tensor imaging and intraoperative data had been prospectively collected for patients with peripheral nerve tumors that were resected. Preoperative identification of the nerve fiber location in relation to the nerve tumor surface as seen on DTI studies was compared with the nerve fiber's intraoperative localization using electrophysiological stimulation and recordings. RESULTS In 23 patients eligible for study there was good correlation between nerve fiber location on DTI and its anatomical location seen intraoperatively. Diffusion tensor imaging demonstrated the relationship of nerve fibers relative to the tumor with 95.7% sensitivity, 66.7% specificity, 75% positive predictive value, and 93.8% negative predictive value. CONCLUSIONS Preoperative DTI techniques are useful in helping the peripheral nerve surgeon to both determine the risks involved in resecting a nerve tumor and plan the safest surgical approach.

Practical Applications of Artificial Intelligence in Spine Imaging: A Review.

Artificial intelligence (AI), a transformative technology with unprecedented potential in medical imaging, can be applied to various spinal pathologies. AI-based approaches may improve imaging efficiency, diagnostic accuracy, and interpretation, which is essential for positive patient outcomes. This review explores AI algorithms, techniques, and applications in spine imaging, highlighting diagnostic impact and challenges with future directions for integrating AI into spine imaging workflow.

Wnt pathway activation predicts increased risk of tumor recurrence in patients with stage I nonsmall cell lung cancer.

OBJECTIVE: To determine the incidence of Wnt pathway activation in patients with stage I NSCLC and its influence on lung cancer recurrence. BACKGROUND: Despite resection, the 5-year recurrence with localized stage I nonsmall cell lung cancer (NSCLC) is 18.4%-24%. Aberrant Wnt signaling activation plays an important role in a wide variety of tumor types. However, there is not much known about the role the Wnt pathway plays in patients with stage I lung cancer. METHODS: Tumor and normal lung tissues from 55 patients following resection for stage I NSCLC were subjected to glutathione S-transferase (GST) E-cadherin pulldown and immunoblot analysis to assess levels of uncomplexed ß-catenin, a reliable measure of Wnt signaling activation. The ß-catenin gene was also screened for oncogenic mutations in tumors with activated Wnt signaling. Cancer recurrence rates were correlated in a blinded manner in patients with Wnt pathway-positive and -negative tumors. RESULTS: Tumors in 20 patients (36.4%) scored as Wnt positive, with only 1 exhibiting a ß-catenin oncogenic mutation. Patients with Wnt-positive tumors experienced a significantly higher rate of overall cancer recurrence than those with Wnt-negative tumors (30.0% vs. 5.7%, P = 0.02), with 25.0% exhibiting distal tumor recurrence compared with 2.9% in the Wnt-negative group (P = 0.02). CONCLUSIONS: Wnt pathway activation occurred in a substantial fraction of Stage I NSCLCs, which was rarely due to mutations. Moreover, Wnt pathway activation was associated with a significantly higher rate of tumor recurrence. These findings suggest that Wnt pathway activation reflects a more aggressive tumor phenotype and identifies patients who may benefit from more aggressive therapy in addition to resection.

A model to create an efficient and equitable admission policy for patients arriving to the cardiothoracic ICU.

OBJECTIVE: To develop queuing and simulation-based models to understand the relationship between ICU bed availability and operating room schedule to maximize the use of critical care resources and minimize case cancellation while providing equity to patients and surgeons. DESIGN: Retrospective analysis of 6-month unit admission data from a cohort of cardiothoracic surgical patients, to create queuing and simulation-based models of ICU bed flow. Three different admission policies (current admission policy, shortest-processing-time policy, and a dynamic policy) were then analyzed using simulation models, representing 10 yr worth of potential admissions. Important output data consisted of the "average waiting time," a proxy for unit efficiency, and the "maximum waiting time," a surrogate for patient equity. SETTING: A cardiothoracic surgical ICU in a tertiary center in New York, NY. PATIENTS: Six hundred thirty consecutive cardiothoracic surgical patients admitted to the cardiothoracic surgical ICU. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Although the shortest-processing-time admission policy performs best in terms of unit efficiency (0.4612 days), it did so at expense of patient equity prolonging surgical waiting time by as much as 21 days. The current policy gives the greatest equity but causes inefficiency in unit bed-flow (0.5033 days). The dynamic policy performs at a level (0.4997 days) 8.3% below that of the shortest-processing-time in average waiting time; however, it balances this with greater patient equity (maximum waiting time could be shortened by 4 days compared to the current policy). CONCLUSIONS: Queuing theory and computer simulation can be used to model case flow through a cardiothoracic operating room and ICU. A dynamic admission policy that looks at current waiting time and expected ICU length of stay allows for increased equity between patients with only minimum losses of efficiency. This dynamic admission policy would seem to be a superior in maximizing case-flow. These results may be generalized to other surgical ICUs.

Comparison of Lumbosacral Fusion Grade in Patients after Transforaminal and Anterior Lumbar Interbody Fusion with Minimum 2-Year Follow-Up.

OBJECTIVE: Generally, anterior lumbar interbody fusion (ALIF) was believed superior to transforaminal lumbar interbody fusion (TLIF) in induction of fusion. However, many studies have reported comparable results in lumbosacral fusion rate between the two approaches. This study aimed to evaluate the realistic lumbosacral arthrodesis rates following ALIF and TLIF in patients with degenerative spondylolisthesis as measured by CT and radiology. METHODS: Ninety-six patients who underwent single-level L5-S1 fusion through ALIF (n = 48) or TLIF (n = 48) for degenerative spondylolisthesis at the Spine Center, University of California San Francisco, between October 2014 and December 2017 were retrospectively evaluated. Fusion was independently evaluated and categorized as solid fusion, indeterminate fusion, or pseudarthroses by two radiologists using the modified Brantigan-Steffee-Fraser (mBSF) grade. Clinical data on sex, age, body mass index, Meyerding grade, smoking status, follow-up times, complications, and radiological parameters including disc height, disc angle, segmental lordosis, and overall lumbar lordosis were collected. The fusion results and clinical and radiographic data were statistically compared between the ALIF and TLIF groups by using t-test or chi-square test. RESULTS: The mean follow-up period was 37.5 (ranging from 24 to 51) months. Clear, solid radiographic fusions were higher in the ALIF group compared with the TLIF group at the last follow-up (75% vs 47.9%, p = 0.006). Indeterminate fusion occurred in 20.8% (10/48) of ALIF cases and in 43.8% (21/48) of TLIF cases (p = 0.028). Radiographic pseudarthrosis was not significantly different between the TLIF and ALIF groups (16.7% vs 8.3%; p = 0.677). In subgroup analysis of the patients without bone morphogenetic protein (BMP), the solid radiographic fusion rate was significantly higher in the ALIF group than that in the TLIF group (78.6% vs 45.5%; p = 0.037). There were no differences in sex, age, body mass index, Meyerding grade, smoking status, or follow-up time between the two groups (p > 0.05). The ALIF group had more improvement in disc height (7.8 mm vs 4.7 mm), disc angle (5.2° vs 1.5°), segmental lordosis (7.0° vs 2.5°), and overall lumbar lordosis (4.7° vs 0.7°) compared with the TLIF group (p < 0.05). Overall complication rates were similar between the TLIF and ALIF groups (10.4% vs 8.33%; p > 0.999). CONCLUSIONS: With a minimum 2-year radiographic analysis of arthrodesis at lumbosacral level by radiologists, the rate of solid radiographic fusions was higher in the ALIF group compared with the TLIF group, whereas the TLIF group had a higher rate of indeterminate fusion. Radiographic pseudarthrosis did not differ significantly between the TLIF and ALIF groups.

Variant Sciatic Nerve Anatomy in Relation to the Piriformis Muscle on Magnetic Resonance Neurography: A Potential Etiology for Extraspinal Sciatica.

OBJECTIVE: To assess the prevalence and clinical implications of variant sciatic nerve anatomy in relation to the piriformis muscle on magnetic resonance neurography (MRN), in patients with lumbosacral neuropathic symptoms. MATERIALS AND METHODS: In this retrospective single-center study, 254 sciatic nerves, from 127 patients with clinical and imaging findings compatible with extra-spinal sciatica on MRN between 2003 and 2013, were evaluated for the presence and type of variant sciatic nerves, split sciatic nerve, abnormal T2-signal hyperintensity, asymmetric piriformis size and increased nerve caliber, and summarized using descriptive statistics. Two-tailed chi-square tests were performed to compare the anatomical variant type and clinical symptoms between imaging and clinical characteristics. RESULTS: Sixty-four variant sciatic nerves were identified with an equal number of right and left variants. Bilateral variants were noted in 15 cases. Abnormal T2-signal hyperintensity was seen significantly more often in variant compared to conventional anatomy (40/64 vs. 82/190; p = 0.01). A sciatic nerve split was seen significantly more often in variant compared to conventional anatomy (56/64 vs. 20/190; p < 0.0001). Increased nerve caliber, abnormal T2-signal hyperintensity, and asymmetric piriformis size were significantly associated with the clinically symptomatic side compared to the asymptomatic side (98:2, 98:2, and 97:3, respectively; p < 0.0001 for all). Clinical symptoms were correlated with variant compared to conventional sciatic nerve anatomy (64% vs. 46%; p = 0.01). CONCLUSION: Variant sciatic nerve anatomy, in relation to the piriformis muscle, is frequently identified with MRN and is more likely to be associated with nerve signal changes and symptomatology.

Diagnosing ulnar neuropathy at the elbow using magnetic resonance neurography.

INTRODUCTION: Early diagnosis of ulnar neuropathy at the elbow is important. Magnetic resonance neurography (MRN) images peripheral nerves. We evaluated the usefulness of elbow MRN in diagnosing ulnar neuropathy at the elbow. METHODS: The MR neurograms of 21 patients with ulnar neuropathy were reviewed retrospectively. MRN was performed prospectively on 10 normal volunteers. The MR neurograms included axial T1 and axial T2 fat-saturated and/or axial STIR sequences. The sensitivity and specificity of MRN in detecting ulnar neuropathy were determined. RESULTS: The mean ulnar nerve size in the symptomatic and normal groups was 0.12 and 0.06 cm(2) (P < 0.001). The mean relative signal intensity in the symptomatic and normal groups was 2.7 and 1.4 (P < 0.01). When using a size of 0.08 cm(2), sensitivity was 95% and specificity was 80%. DISCUSSION: Ulnar nerve size and signal intensity were greater in patients with ulnar neuropathy. MRN is a useful test in evaluating ulnar neuropathy at the elbow.

Automatic detection and voxel-wise mapping of lumbar spine Modic changes with deep learning.

Background: Modic changes (MCs) are the most prevalent classification system for describing magnetic resonance imaging (MRI) signal intensity changes in the vertebrae. However, there is a growing need for novel quantitative and standardized methods of characterizing these anomalies, particularly for lesions of transitional or mixed nature, due to the lack of conclusive evidence of their associations with low back pain. This retrospective imaging study aims to develop an interpretable deep learning-based detection tool for voxel-wise mapping of MCs. Methods: Seventy-five lumbar spine MRI exams that presented with acute-to-chronic low back pain, radiculopathy, and other symptoms of the lumbar spine were enrolled. The pipeline consists of two deep convolutional neural networks to generate an interpretable voxel-wise Modic map. First, an autoencoder was trained to segment vertebral bodies from T1-weighted sagittal lumbar spine images. Next, two radiologists segmented and labeled MCs from a combined T1- and T2-weighted assessment to serve as ground truth for training a second autoencoder that performs segmentation of MCs. The voxels in the detected regions were then categorized to the appropriate Modic type using a rule-based signal intensity algorithm. Post hoc, three radiologists independently graded a second dataset with the aid of the model predictions in an artificial (AI)-assisted experiment. Results: The model successfully identified the presence of changes in 85.7% of samples in the unseen test set with a sensitivity of 0.71 (±0.072), specificity of 0.95 (±0.022), and Cohen's kappa score of 0.63. In the AI-assisted experiment, the agreement between the junior radiologist and the senior neuroradiologist significantly improved from Cohen's kappa score of 0.52 to 0.58 (p < 0.05). Conclusions: This deep learning-based approach demonstrates substantial agreement with radiologists and may serve as a tool to improve inter-rater reliability in the assessment of MCs.

Institution-wide shape analysis of 3D spinal curvature and global alignment parameters.

The spine is an articulated, 3D structure with 6 degrees of translational and rotational freedom. Clinical studies have shown spinal deformities are associated with pain and functional disability in both adult and pediatric populations. Clinical decision making relies on accurate characterization of the spinal deformity and monitoring of its progression over time. However, Cobb angle measurements are time-consuming, are limited by interobserver variability, and represent a simplified 2D view of a 3D structure. Instead, spine deformities can be described by 3D shape parameters, addressing the limitations of current measurement methods. To this end, we develop and validate a deep learning algorithm to automatically extract the vertebral midline (from the upper endplate of S1 to the lower endplate of C7) for frontal and lateral radiographs. Our results demonstrate robust performance across datasets and patient populations. Approximations of 3D spines are reconstructed from the unit normalized midline curves of 20,118 pairs of full spine radiographs belonging to 15,378 patients acquired at our institution between 2008 and 2020. The resulting 3D dataset is used to describe global imbalance parameters in the patient population and to build a statistical shape model to describe global spine shape variations in preoperative deformity patients via eight interpretable shape parameters. The developed method can identify patient subgroups with similar shape characteristics without relying on an existing shape classification system.

CT-guided biopsy of bone and soft-tissue lesions: role of on-site immediate cytologic evaluation.

PURPOSE: To assess the impact of on-site immediate cytologic assessment (ICA) on the diagnostic success rate of computed tomography (CT)-guided percutaneous needle biopsy (PNB) of musculoskeletal lesions and the long-term outcome in inconclusive PNB findings. MATERIALS AND METHODS: A total of 299 CT-guided PNBs of musculoskeletal lesions performed between January 1997 and December 2009 were retrospectively reviewed. The lesions were categorized by their morphology, location, and size, and by biopsy type. The diagnostic success rates, impact of ICA, and outcome in inconclusive PNBs were studied, with final histopathologic findings and/or clinical follow-up as a reference. RESULTS: The overall diagnostic success rate of PNBs was 72.9% (218 of 299). The success rate increased with larger lesions (> 2 cm to 4 cm; P = .009). Biopsies performed with ICA had a higher success rate (77.0% vs 63.3%; P = .015). PNBs had inconclusive results in 109 of 299 cases (36.5%). In 66 of these, repeat open biopsy or clinical follow-up demonstrated 19 malignant/aggressive lesions (28.8%) and 47 benign/nonaggressive lesions (71.2%). CONCLUSIONS: CT-guided PNB had a satisfactory success rate, which significantly increased when performed with ICA. Inconclusive results in PNB were most frequently associated with benign findings during further workup.

Cerebrospinal Fluid-Venous Fistulas: A Systematic Review and Examination of Individual Patient Data.

BACKGROUND: Spontaneous intracranial hypotension (SIH) is usually caused by a spinal cerebrospinal fluid (CSF) leak. CSF-venous fistula is an underdiagnosed cause of spinal CSF leak, as it is challenging to identify on myelography. OBJECTIVE: To review existing literature to summarize common presentations, diagnostic imaging modalities, and current treatment strategies for CSF-venous fistulas. METHODS: We conducted a systematic review using PubMed, Embase, Scopus, and Web of Science databases to identify studies discussing CSF-venous fistulas. Titles and abstracts were screened. Studies meeting prespecified inclusion criteria were reviewed in full. RESULTS: Of 180 articles identified, 16 articles met inclusion criteria. Individual patient data was acquired from 7 studies reporting on 18 patients. CSF-venous fistula most frequently presented as positional headache. Digital subtraction myelography provided greatest detection of CSF-venous fistula in the lateral decubitus position and detected CSF-venous fistula in all individual patient cases. Dynamic computed tomography (CT) myelogram enabled detection and differentiation of CSF-venous fistulas from low-flow epidural leaks. The majority of fistulas were in the thoracic spine and slightly more common on the right. Epidural blood patch (EBP) provided temporary or no relief in all individual patients. Resolution or improvement of clinical symptoms and radiologic normalization were observed in all surgically treated patients. CONCLUSION: Although rare, CSF-venous fistula is an important cause of spinal CSF leak contributing to SIH. Dynamic CT myelogram and digital subtraction myelography, particularly in the lateral decubitus position, are the most accurate and effective diagnostic imaging modalities. EBPs often provide only transient relief, while surgical management is preferred.