Deep phenotyping the cervical spine: automatic characterization of cervical degenerative phenotypes based on T2-weighted MRI.

PURPOSE: Radiological degenerative phenotypes provide insight into a patient's overall extent of disease and can be predictive for future pathological developments as well as surgical outcomes and complications. The objective of this study was to develop a reliable method for automatically classifying sagittal MRI image stacks of cervical spinal segments with respect to these degenerative phenotypes. METHODS: We manually evaluated sagittal image data of the cervical spine of 873 patients (5182 motion segments) with respect to 5 radiological phenotypes. We then used this data set as ground truth for training a range of multi-class multi-label deep learning-based models to classify each motion segment automatically, on which we then performed hyper-parameter optimization. RESULTS: The ground truth evaluations turned out to be relatively balanced for the labels disc displacement posterior, osteophyte anterior superior, osteophyte posterior superior, and osteophyte posterior inferior. Although we could not identify a single model that worked equally well across all the labels, the 3D-convolutional approach turned out to be preferable for classifying all labels. CONCLUSIONS: Class imbalance in the training data and label noise made it difficult to achieve high predictive power for underrepresented classes. This shortcoming will be mitigated in the future versions by extending the training data set accordingly. Nevertheless, the classification performance rivals and in some cases surpasses that of human raters, while speeding up the evaluation process to only require a few seconds.

An externally validated deep learning model for the accurate segmentation of the lumbar paravertebral muscles.

PURPOSE: Imaging studies about the relevance of muscles in spinal disorders, and sarcopenia in general, require the segmentation of the muscles in the images which is very labour-intensive if performed manually and poses a practical limit to the number of investigated subjects. This study aimed at developing a deep learning-based tool able to fully automatically perform an accurate segmentation of the lumbar muscles in axial MRI scans, and at validating the new tool on an external dataset. METHODS: A set of 60 axial MRI images of the lumbar spine was retrospectively collected from a clinical database. Psoas major, quadratus lumborum, erector spinae, and multifidus were manually segmented in all available slices. The dataset was used to train and validate a deep neural network able to segment muscles automatically. Subsequently, the network was externally validated on images purposely acquired from 22 healthy volunteers. RESULTS: The median Jaccard index for the individual muscles calculated for the 22 subjects of the external validation set ranged between 0.862 and 0.935, demonstrating a generally excellent performance of the network, although occasional failures were noted. Cross-sectional area and fat fraction of the muscles were in agreement with published data. CONCLUSIONS: The externally validated deep neural network was able to perform the segmentation of the paravertebral muscles in an accurate and fully automated manner, although it is not without limitations. The model is therefore a suitable research tool to perform large-scale studies in the field of spinal disorders and sarcopenia, overcoming the limitations of non-automated methods.

ISSLS Prize in Bioengineering Science 2021: in vivo sagittal motion of the lumbar spine in low back pain patients-a radiological big data study.

PURPOSE: We investigated the flexion-extension range of motion and centre of rotation of lumbar motion segments in a large population of 602 patients (3612 levels), and the associations between lumbar motion and other parameters such as sex, age and intervertebral disc degeneration. METHODS: Lumbar radiographs in flexion-extension of 602 patients suffering from low back pain and/or suspect instability were collected; magnetic resonance images were retrieved and used to score the degree of disc degeneration for a subgroup of 354 patients. Range of motion and centre of rotation were calculated for all lumbosacral levels with in-house software allowing for high degree of automation. Associations between motion parameters and age, sex, spinal level and disc degeneration were then assessed. RESULTS: The median range of motion was 6.6° (range 0.1-28.9°). Associations between range of motion and age as well as spinal level, but not sex, were found. Disc degeneration determined a consistent reduction in the range of motion. The centre of rotation was most commonly located at the centre of the lower endplate or slightly lower. With progressive degeneration, centres of rotation were increasingly dispersed with no preferential directions. CONCLUSION: This study constitutes the largest analysis of the in vivo lumbar motion currently available and covers a wide range of clinical scenarios in terms of age and degeneration. Findings confirmed that ageing determines a reduction in the mobility independently of degeneration and that in degenerative levels, centres of rotation are dispersed around the centre of the intervertebral space.

The impact of age, sex, disc height loss and T1 slope on the upper and lower cervical lordosis: a large-scale radiologic study.

PURPOSE: To clarify the relative influence of age, sex, disc height loss and T1 slope on upper (Occiput-C2) and lower cervical lordosis (C2-C7). METHODS: Standing lateral cervical radiographs of 865 adult subjects were evaluated. The presence and severity of disc height loss from C2/C3 to C6/C7 (a total of 4325 discs) were assessed using a validated grading system. The total disc height loss score for each subject was calculated as the sum of the score of each disc space. Sagittal radiographic parameters included: occipital slope, occiput-C2 (Oc-C2) lordosis, C2-C7 lordosis and T1 slope. Multivariable regression analyses were performed to examine the relative influence of the multiple factors on upper and lower cervical lordosis. RESULTS: This study included 360 males and 505 females, with a mean age of 40.2 ± 16.0 years (range, 20-95 years). Linear multivariate regression analyses showed that greater age, male sex, greater T1 slope were each found to be significantly and independently associated with greater C2-C7 lordosis, whereas total disc height loss score was negatively associated with C2-C7 lordosis. T1 slope had the most independent influence on C2-C7 lordosis among these factors. Age, sex and disc height loss were not independently associated with Oc-C2 lordosis. CONCLUSIONS: Results from our large-scale radiologic analysis may enhance the understanding of the factors that affect cervical lordosis, indicating that age, sex, disc height loss and T1 slope were each independently associated with C2-C7 lordosis. However, age, sex and disc height loss were not independently associated with upper cervical lordosis.

Radiographic cervical spine degenerative findings: a study on a large population from age 18 to 97 years.

PURPOSE: The aims of this study were (1) to determine the prevalence of radiographic cervical disc degeneration in a large population of patients aged from 18 to 97 years; (2) to investigate individually the prevalence and distribution of height loss, osteophyte formation, endplate sclerosis and spondylolisthesis; and (3) to describe the patterns of cervical disc degeneration. METHODS: A retrospective study was performed. Standard lateral cervical spine radiographs in standing, neutral position of 1581 consecutive patients (723 males, 858 females) with an average age of 41.2 ± 18.2 years were evaluated. Cervical disc degeneration was graded from C2/C3 to C6/C7 based on a validated quantitative grading system. The prevalence and distribution of radiographic findings were evaluated and associations with age were investigated. RESULTS: 53.9% of individuals had radiographic disc degeneration and the most affected level was C5/C6. The presence and severity of disc degeneration were found to be significantly associated with age both in male and female subjects. The most frequent and severe occurrences of height loss, osteophyte formation, and endplate sclerosis were at C5/C6, whereas spondylolisthesis was most observed at C4/C5. Age was significantly correlated with radiographic degenerative findings. Contiguous levels degeneration pattern was more likely found than skipped level degeneration. The number of degenerated levels was also associated with age. CONCLUSIONS: The presence and severity of radiographic disc degeneration increased with aging in the cervical spine. Older age was associated with greater number of degenerated disc levels. Furthermore, the correlations between age and the degree of degenerative findings were stronger at C5/C6 and C6/C7 than at other cervical spinal levels.

Sagittal wedging of intervertebral discs and vertebral bodies in the cervical spine and their associations with age, sex and cervical lordosis: A large-scale morphological study.

Many recent studies have focused on the functional and clinical importance of cervical lordosis. However, there is little accurate knowledge of the anatomical parameters that constitute cervical lordosis (i.e., the sagittal wedging angles of intervertebral discs and vertebral bodies) and their associations with age and sex. Standing lateral cervical radiographs of 1020 subjects (424 males, 596 females) with a mean age of 36.6 ± 17.0 years (range 7-95 years) were evaluated retrospectively. Cervical lordosis, the sum of intervertebral disc wedging angles from C2/C3 to C6/C7 and the sum of vertebral body wedging angles from C3 to C7 were measured. The sum of intervertebral disc wedging and the sum of vertebral body wedging were 20.6° ± 14.7° and -12.8° ± 10.3°, respectively. The sum of intervertebral disc wedging increased significantly with age and was significantly greater in males than females, whereas there was no sex-related difference in the sum of vertebral body wedging. The sum of intervertebral disc wedging was negatively correlated with sum of vertebral body wedging. Wedging of discs contributed to C2-C7 cervical lordosis more significantly than wedging of vertebral bodies. There were moderate positive correlations between cervical lordosis and intervertebral disc wedging angles at C3/C4, C4/C5 and C5/C6; weak correlations were observed at C2/C3 and C6/C7. This study constitutes the largest currently available analysis comprehensively documenting the anatomical characteristics of sagittal wedging of intervertebral discs and vertebral bodies in the cervical spine. The findings could improve understanding of the internal architecture of cervical lordosis among clinicians.

Fully automated radiological analysis of spinal disorders and deformities: a deep learning approach.

PURPOSE: We present an automated method for extracting anatomical parameters from biplanar radiographs of the spine, which is able to deal with a wide scenario of conditions, including sagittal and coronal deformities, degenerative phenomena as well as images acquired with different fields of view. METHODS: The location of 78 landmarks (end plate centers, hip joint centers, and margins of the S1 end plate) was extracted from three-dimensional reconstructions of 493 spines of patients suffering from various disorders, including adolescent idiopathic scoliosis, adult deformities, and spinal stenosis. A fully convolutional neural network featuring an additional differentiable spatial to numerical (DSNT) layer was trained to predict the location of each landmark. The values of some parameters (T4-T12 kyphosis, L1-L5 lordosis, Cobb angle of scoliosis, pelvic incidence, sacral slope, and pelvic tilt) were then calculated based on the landmarks' locations. A quantitative comparison between the predicted parameters and the ground truth was performed on a set of 50 patients. RESULTS: The spine shape predicted by the models was perceptually convincing in all cases. All predicted parameters were strongly correlated with the ground truth. However, the standard errors of the estimated parameters ranged from 2.7° (for the pelvic tilt) to 11.5° (for the L1-L5 lordosis). CONCLUSIONS: The proposed method is able to automatically determine the spine shape in biplanar radiographs and calculate anatomical and posture parameters in a wide scenario of clinical conditions with a very good visual performance, despite limitations highlighted by the statistical analysis of the results. These slides can be retrieved under Electronic Supplementary Material.

Interlaboratory comparison of femur surface reconstruction from CT data compared to reference optical 3D scan.

BACKGROUND: The present study contrasts the accuracy of different reconstructed models with distinctive segmentation methods performed by various experts. Seven research groups reconstructed nine 3D models of one human femur based on an acquired CT image using their own computational methods. As a reference model for accuracy assessment, a 3D surface scan of the human femur was created using an optical measuring system. Prior to comparison, the femur was divided into four areas; "neck and greater trochanter", "proximal metaphysis", "diaphysis", and "distal metaphysis". The deviation analysis was carried out in GEOMAGIC studio v.2013 software. RESULTS: The results revealed that the highest deviation errors occurred in "neck and greater trochanter" area and "proximal metaphysis" area with RMSE of 0.84 and 0.83 mm respectively. CONCLUSION: In conclusion, this study shows that the average deviation of reconstructed models prepared by experts with various methods, skills and software from the surface 3D scan is lower than 0.79 mm, which is not a significant discrepancy.

Asymmetrical intrapleural pressure distribution: a cause for scoliosis? A computational analysis.

PURPOSE: The mechanical link between the pleural physiology and the development of scoliosis is still unresolved. The intrapleural pressure (IPP) which is distributed across the inner chest wall has yet been widely neglected in etiology debates. With this study, we attempted to investigate the mechanical influence of the IPP distribution on the shape of the spinal curvature. METHODS: A finite element model of pleura, chest and spine was created based on CT data of a patient with no visual deformities. Different IPP distributions at a static end of expiration condition were investigated, such as the influence of an asymmetry in the IPP distribution between the left and right hemithorax. The results were then compared to clinical data. RESULTS: The application of the IPP resulted in a compressive force of 22.3 N and a flexion moment of 2.8 N m at S1. An asymmetrical pressure between the left and right hemithorax resulted in lateral deviation of the spine towards the side of the reduced negative pressure. In particular, the pressure within the dorsal section of the rib cage had a strong influence on the vertebral rotation, while the pressure in medial and ventral region affected the lateral displacement. CONCLUSIONS: An asymmetrical IPP caused spinal deformation patterns which were comparable to deformation patterns seen in scoliotic spines. The calculated reaction forces suggest that the IPP contributes in counterbalancing the weight of the intrathoracic organs. The study confirms the potential relevance of the IPP for spinal biomechanics and pathologies, such as adolescent idiopathic scoliosis.

Characteristic morphological patterns within adolescent idiopathic scoliosis may be explained by mechanical loading.

PURPOSE: Adolescent idiopathic scoliosis (AIS) is a three-dimensional deformity of the spine which exhibits morphological changes during growth. The goal of this study was to identify morphological patterns that could be explained by different loading patterns for AIS. METHODS: Computed tomography data of 21 patients with diagnosed AIS and 48 patients without any visual spinal abnormalities were collected prospectively. The bony structures were reconstructed, and landmarks were placed on characteristic morphological points on the spine. Multiple morphological parameters were calculated based on the distances between the landmarks. The intra- and inter-observer variability for each parameter was estimated. Differences between healthy and scoliotic spines were statistically analysed using the t test for unpaired data, with a significance level of α = 0.01. RESULTS: Within the healthy group, an out-of-plane rotation of the vertebrae in the transverse plane was measured (2.6° ± 4.1° at T2). Relating the length of the spinal curvature to the T1-S1 height of the spine revealed that scoliotic spines were significantly longer. However, the endplate area in the AIS group was significantly smaller once compared to the curvature length. The relation between the left and right pedicle areas varied between 2.5 ± 0.79 and 0.4 ± 0.19, while the ratio of the facet articular surfaces varied within 2.3 ± 0.5 and 0.5 ± 0.2. CONCLUSIONS: This study identified a certain morphological pattern along the spine, which reveals a distinct load path prevalent within AIS. The data suggested that the spine adapts to the asymmetric load conditions and the spine is not deformed by asymmetric growth disturbance. These slides can be retrieved under Electronic Supplementary Material.

Case Studies of a Simulation Workflow to Improve Bone Healing Assessment in Impending Non-Unions.

Background: The healing potential of a fracture is determined by mechanical and biological factors. Simulation-based workflows can help assess these factors to assist in predicting non-unions. The aim of this study was the introduction of two use cases for a novel patient-specific simulation workflow based on clinically available information. Methods: The used software is an extension of the "Ulm Bone Healing model" and was applied in two cases with non-union development after fracture fixation to show its principal feasibility. The clinical and radiographic information, starting from initial treatment, were used to feed the simulation process. Results: The simulation predicted non-union development and axial deviation in a mechanically driven non-union. In the case of a biological non-union, a slow, incomplete healing course was correctly identified. However, the time offset in callus bridging was discordant between the simulation and the distinctly slower healing response in the clinical case. Conclusions: The simulation workflow presented in the two clinical use cases allowed for the identification of fractures at risk for impending non-union immediately after the initial fixation based on available clinical and radiographic information. Further validation in a large non-union cohort is needed to increase the model's precision, especially in biologically challenging cases, and show its validity as a screening instrument.

Automatic grading of intervertebral disc degeneration in lumbar dog spines.

Background: Intervertebral disc degeneration is frequent in dogs and can be associated with symptoms and functional impairments. The degree of disc degeneration can be assessed on T2-weighted MRI scans using the Pfirrmann classification scheme, which was developed for the human spine. However, it could also be used to quantify the effectiveness of disc regeneration therapies. We developed and tested a deep learning tool able to automatically score the degree of disc degeneration in dog spines, starting from an existing model designed to process images of human patients. Methods: MRI midsagittal scans of 5991 lumbar discs of dog patients were collected and manually evaluated with the Pfirrmann scheme and a modified scheme with transitional grades. A deep learning model was trained to classify the disc images based on the two schemes and tested by comparing its performance with the model processing human images. Results: The determination of the Pfirrmann grade showed sensitivities higher than 83% for all degeneration grades, except for grade 5, which is rare in dog spines, and high specificities. In comparison, the correspondent human model had slightly higher sensitivities, on average 90% versus 85% for the canine model. The modified scheme with the fractional grades did not show significant advantages with respect to the original Pfirrmann grades. Conclusions: The novel tool was able to accurately and reliably score the severity of disc degeneration in dogs, although with a performance inferior than that of the human model. The tool has potential in the clinical management of disc degeneration in canine patients as well as in longitudinal studies evaluating regenerative therapies in dogs used as animal models of human disorders.

Computer-Based Mechanobiological Fracture Healing Model Predicts Non-Union of Surgically Treated Diaphyseal Femur Fractures.

As non-unions are still common, a predictive assessment of healing complications could enable immediate intervention before negative impacts for the patient occur. The aim of this pilot study was to predict consolidation with the help of a numerical simulation model. A total of 32 simulations of patients with closed diaphyseal femoral shaft fractures treated by intramedullary nailing (PFNA long, FRN, LFN, and DePuy Synthes) were performed by creating 3D volume models based on biplanar postoperative radiographs. An established fracture healing model, which describes the changes in tissue distribution at the fracture site, was used to predict the individual healing process based on the surgical treatment performed and full weight bearing. The assumed consolidation as well as the bridging dates were retrospectively correlated with the clinical and radiological healing processes. The simulation correctly predicted 23 uncomplicated healing fractures. Three patients showed healing potential according to the simulation, but clinically turned out to be non-unions. Four out of six non-unions were correctly detected as non-unions by the simulation, and two simulations were wrongfully diagnosed as non-unions. Further adjustments of the simulation algorithm for human fracture healing and a larger cohort are necessary. However, these first results show a promising approach towards an individualized prognosis of fracture healing based on biomechanical factors.

Endplate abnormalities, Modic changes and their relationship to alignment parameters and surgical outcomes in the cervical spine.

Modic changes (MC) and endplate abnormalities (EA) have been shown to impact preoperative symptoms and outcomes following spinal surgery. However, little is known about how these phenotypes impact cervical alignment. This study aimed to evaluate the impact that these phenotypes have on preoperative, postoperative, and changes in cervical alignment in patients undergoing anterior cervical discectomy and fusion (ACDF). We performed a retrospective study of prospectively collected data of ACDF patients at a single institution. Preoperative magnetic resonance imagings (MRIs) were used to assess for the MC and EA. Patients were subdivided into four groups: MC-only, EA-only, the combined Modic-Endplate-Complex (MEC), and patients without either phenotype. Pre and postoperative MRIs were used to assess alignment parameters. Associations with imaging phenotypes and alignment parameters were assessed, and statistical significance was set at p < 0.5. A total of 512 patients were included, with 84 MC-only patients, 166 EA-only patients, and 71 patients with MEC. Preoperative MC (p = 0.031) and the MEC (p = 0.039) had significantly lower preoperative T1 slope compared to controls. Lower preoperative T1 slope was a risk factor for MC (p = 0.020) and MEC (p = 0.029) and presence of MC (Type II) and the MEC (Type III) was predictive of lower preoperative T1 slope. There were no differences in postoperative alignment measures or patient reported outcome measures. MC and endplate pathologies such as the MEC appear to be associated with worse cervical alignment at baseline relative to patients without these phenotypes. Poor alignment may be an adaptive response to these degenerative findings or may be a risk factor for their development.

High-Intensity Zones on MRI of the Cervical Spine in Patients: Epidemiology and Association With Pain and Disability.

STUDY DESIGN: Retrospective cohort study. OBJECTIVES: This study aimed to address the prevalence, distribution, and clinical significance of cervical high-intensity zones (HIZs) on magnetic resonance imaging (MRI) with respect to pain and other patient-reported outcomes in the setting of patients that will undergo an anterior cervical discectomy and fusion (ACDF) procedure. METHODS: A retrospective cohort study of ACDF patients surgically treated at a single center from 2008 to 2015. Based on preoperative MRI, HIZ subtypes were identified as either traditional T2-hyperintense, T1-hypointense ("single-HIZs"), or combined T1- and T2-hyperintense ("dual-HIZs"), and their level-specific prevalence was assessed. Preoperative symptoms, patient-reported outcomes, and disc degeneration pathology were assessed in relation to HIZs and HIZ subtypes. RESULTS: Of 861 patients, 58 demonstrated evidence of HIZs in the cervical spine (6.7%). Single-HIZs and dual-HIZs comprised 63.8% and 36.2% of the overall HIZs, respectively. HIZs found outside of the planned fusion segment reported better preoperative Neck Disability Index (NDI; P = .049) and Visual Analogue Scale (VAS) Arm (P = .014) scores relative to patients without HIZs. Furthermore, patients with single-HIZs found inside the planned fusion segment had worse VAS Neck (P = .045) and VAS Arm (P = .010) scores. In general, dual-HIZ patients showed no significant differences across all clinical outcomes. CONCLUSIONS: This is the first study to evaluate the clinical significance of HIZs in the cervical spine, noting level-specific and clinical outcome-specific variations. Single-HIZs were associated with significantly more pain when located inside the fusion segment, while dual-HIZs showed no associations with patient-reported outcomes. The presence of single-HIZs may correlate with concurrent spinal pathologies and should be more closely evaluated.

The Modic-endplate-complex phenotype in cervical spine patients: Association with symptoms and outcomes.

This study describes a novel, combined Modic changes (MC) and structural endplate abnormality phenotype of the cervical spine, which we have termed the Modic-Endplate-Complex (MEC), and its association with preoperative symptoms and outcomes in anterior cervical discectomy and fusion (ACDF) patients. This was a retrospective study of prospectively collected data at a single institution. Preoperative cervical magnetic resonance imagings were used to assess the presence of MC and endplate abnormalities. Patients were divided into four groups: MC-only, endplate abnormality-only, the MEC and controls. The MEC was defined as the presence of both a MC and endplate abnormality in the cervical spine. Phenotypes were further stratified by location and compared to controls. Associations with patient-reported outcome measures were assessed using regression controlling for baseline characteristics. A total of 628 patients were included, with 84 MC-only, 166 endplate abnormality-only, and 187 MEC patients. Both MC (p < 0.001) and endplate abnormalities (p < 0.001) were independently associated with one another. MC at the adjacent level (p = 0.018), endplate abnormalities (regardless of location) (p = 0.001), and the MEC within the fusion segment (p = 0.027) were all associated with higher Neck Disability Index scores. Both MC within the fusion segment (p = 0.008) and endplate abnormalities within the fusion segment (p = 0.017) associated with lower Veteran's Rand 12-item scores. MC and structural endplate abnormalities commonly manifest concomitantly in patients indicated for ACDF for degenerative pathology. Patients with the endplate pathology, including the MEC phenotype, reported significantly higher levels of postoperative disability following ACDF. These findings add valuable data to the prognostic assessment of degenerative cervical spine patients.

A Deep Learning Model for the Accurate and Reliable Classification of Disc Degeneration Based on MRI Data.

OBJECTIVES: Although magnetic resonance imaging-based formalized grading schemes for intervertebral disc degeneration offer improved reproducibility compared with purely subjective ratings, their intrarater and interrater reliability are not nearly good enough to be able to detect small to medium effects in clinical longitudinal studies. The aim of this study thus was to develop a method that enables automatic and therefore reproducible and reliable evaluation of disc degeneration based on conventional clinical image data and Pfirrmann's grading scheme. MATERIALS AND METHODS: We propose a classifier based on a deep convolutional neural network that we trained on a large, manually evaluated data set of 1599 patients (7948 intervertebral discs). To improve upon the status quo, we focused on the quality of the training data and performed extensive hyperparameter optimization. We assessed the potential benefits of optimizing loss functions beyond common cross-entropy loss, such as soft kappa loss, ordinal cross-entropy loss, or regression losses. We furthermore experimented with ways to mitigate class imbalance by pooling classes or using class-weighted loss functions. During model development and hyperparameter optimization, we used a fixed 90%/10% training/validation set split. To estimate real-world prediction performance, we performed 10-fold cross-validation. RESULTS: The evaluated image data results in a Gaussian degeneration grade distribution, and thus grades 1 and 5 are slightly underrepresented in the training set. Our default cross-entropy-based classifier achieves a reliability of κ = 0.92 (Cohen κ), an average sensitivity of 90.2%, and an average precision of 92.5%. In 99.2% of validation cases, the network's prediction deviates at most 1 Pfirrmann grades from the ground truth. Framed as an ordinal regression problem, the mean absolute error between the ground truth and the prediction is 0.08 Pfirrmann grade with a correlation of r = 0.96. The results of the 10-fold cross validation confirm those performance estimates, indicating no substantial overfitting. More sophisticated loss functions, class-based loss weighting, or class pooling did not lead to improved classification performance overall. CONCLUSIONS: With a reliability of κ > 0.9, our system clearly outperforms average human interrater as well as intrarater reliability. With an average sensitivity of more than 90%, our classifier also surpasses state-of-the-art machine learning solutions for automatically grading disc degeneration.

Cervical spine MRI phenotypes and prediction of pain, disability and adjacent segment degeneration/disease after ACDF.

Degenerative spine imaging findings have been extensively studied in the lumbar region and are associated with pain and adverse clinical outcomes after surgery. However, few studies have investigated the significance of these imaging "phenotypes" in the cervical spine. Patients with degenerative cervical spine pathology undergoing anterior cervical discectomy and fusion (ACDF) from 2008 to 2015 were retrospectively and prospectively assessed using preoperative MRI for disc degeneration, narrowing, and displacement, high-intensity zones, endplate abnormalities, Modic changes, and osteophyte formation from C2-T1. Points were assigned for these phenotypes to generate a novel Cervical Phenotype Index (CPI). Demographics were evaluated for association with phenotypes and the CPI using forward stepwise regression. Bootstrap sampling and multiple imputations assessed phenotypes and the CPI in association with patient-reported outcomes (Neck Disability Index [NDI], Visual Analog Scale [VAS]-neck, VAS-arm) and adjacent segment degeneration (ASDeg) and disease (ASDz). Of 861 patients, disc displacement was the most common (99.7%), followed by osteophytes (92.0%) and endplate abnormalities (57.3%). Most findings were associated with age and were identified at similar cervical vertebral levels; at C5-C7. Imaging phenotypes demonstrated both increased and decreased associations with adverse patient-reported outcomes and ASDeg/Dz. However, the CPI consistently predicted worse NDI (P = .012), VAS-neck (P = .007), and VAS-arm (P = .013) scores, in addition to higher odds of ASDeg (P = .002) and ASDz (P = .004). The CPI was significantly predictive of postoperative symptoms of pain/disability and ASDeg/Dz after ACDF, suggesting that the totality of degenerative findings may be more clinically relevant than individual phenotypes and that this tool may help prognosticate outcomes after surgery.

Estimating the three-dimensional vertebral orientation from a planar radiograph: Is it feasible?

We trained a deep neural network for the three-dimensional estimation of the direction of the three anatomical axes (cranio-caudal, anteroposterior and laterolateral) of individual vertebrae from a single sagittal radiographic image acquired from an approximately lateral direction with large deviations from a perfect alignment up to 60 degrees. To this aim, we exploited computed tomography (CT), which can be used to create simulated radiographic projections with different orientations, for the creation of large training and validation datasets. In a set of 21 CT stacks, the location of 5 landmark points was manually determined for L2, L3 and L4, for a total of 63 vertebrae. For each vertebra, 200 simulated projections approximately aligned with sagittal plane but including random perturbations of the projection direction were built, resulting in 12,600 simulated radiographs with the corresponding local directions of the anatomical axes. These data were integrated by 1765 lateral images of vertebrae acquired with a biplanar radiographic imaging system, for which the orientation was calculated by means of three-dimensional reconstruction. The whole dataset was used to train a deep neural network, ResNet-101, customized for the estimation of the three-dimensional components of the axes. The accuracy of the network was qualitatively and quantitatively tested on a large group of simulated radiographic images as well as real lateral images acquired with a biplanar radiographic system for which the direction of the axes was known. Errors were lower than 3 degrees in 76% of the evaluations conducted on the simulated images, and in 86% for the real radiographs. The novel method will be useful to extract three-dimensional information from planar images even in clinical cases in which vertebrae are markedly rotated due to spinal deformities or to an imprecise alignment of the patient with respect to the detector.

Cervical Spine Endplate Abnormalities and Association With Pain, Disability, and Adjacent Segment Degeneration After Anterior Cervical Discectomy and Fusion.

STUDY DESIGN: A retrospective study with prospectively-collected data. OBJECTIVE: To determine how type, location, and size of endplate lesions on magnetic resonance imaging (MRI) may be associated with symptoms and clinical outcomes after anterior cervical discectomy and fusion (ACDF). SUMMARY OF BACKGROUND DATA: Structural endplate abnormalities are important, yet understudied, phenomena in the cervical spine. ACDF is a common surgical treatment for degenerative disc disease; however, adjacent segment degeneration/disease (ASD) may develop. METHODS: Assessed the imaging, symptoms and clinical outcomes of 861 patients who underwent ACDF at a single center. MRI and plain radiographs of the cervical spine were evaluated. Endplate abnormalities on MRI were identified and stratified by type (atypical, typical), location, relation to operative levels, presence at the adjacent level, and size. These strata were assessed for association with presenting symptoms, patient-reported, and postoperative outcomes. RESULTS: Of 861 patients (mean follow-up: 17.4 months), 57.3% had evidence of endplate abnormalities, 39.0% had typical abnormalities, while 18.2% had atypical abnormalities. Patients with any endplate abnormality had greater odds of myelopathy irrespective of location or size, while sensory deficits were associated with atypical lesions (P = 0.016). Typical and atypical abnormalities demonstrated differences in patient-reported outcomes based on location relative to the fused segment. Typical variants were not associated with adverse surgical outcomes, while atypical lesions were associated with ASD (irrespective of size/location; P = 0.004) and reoperations, when a large abnormality was present at the proximal adjacent level (P = 0.025). CONCLUSION: This is the first study to examine endplate abnormalities on MRI of the cervical spine, demonstrating distinct risk profiles for symptoms, patient-reported, and surgical outcomes after ACDF. Patients with typical lesions reported worsening postoperative pain/disability, while those with atypical abnormalities experienced greater rates of ASD and reoperation. This highlights the relevance of a degenerative spine phenotypic assessment, and suggests endplate abnormalities may prognosticate clinical outcomes after surgery. LEVEL OF EVIDENCE: 3.