Convolutional Neural Network-Based Automated Segmentation of the Spinal Cord and Contusion Injury: Deep Learning Biomarker Correlates of Motor Impairment in Acute Spinal Cord Injury

SCIseg: Automatic Segmentation of T2-weighted Intramedullary Lesions in Spinal Cord Injury

Purpose

To develop a deep learning tool for the automatic segmentation of T2-weighted intramedullary lesions in spinal cord injury (SCI).

Material and Methods

This retrospective study included a cohort of SCI patients from three sites enrolled between July 2002 and February 2023. A deep learning model, SCIseg, was trained in a three-phase process involving active learning for the automatic segmentation of intramedullary SCI lesions and the spinal cord. The data consisted of T2-weighted MRI acquired using different scanner manufacturers with heterogeneous image resolutions (isotropic/anisotropic), orientations (axial/sagittal), lesion etiologies (traumatic/ischemic/hemorrhagic) and lesions spread across the cervical, thoracic and lumbar spine. The segmentations from the proposed model were visually and quantitatively compared with other open-source baselines. Wilcoxon signed-rank test was used to compare quantitative MRI biomarkers (lesion volume, lesion length, and maximal axial damage ratio) computed from manual lesion masks and those obtained automatically with SCIseg predictions.

Results

MRI data from 191 SCI patients (mean age, 48.1 years ± 17.9 [SD]; 142 males) were used for model training and evaluation. SCIseg achieved the best segmentation performance for both the cord and lesions. There was no statistically significant difference between lesion length and maximal axial damage ratio computed from manually annotated lesions and those obtained using SCIseg.

Conclusion

Automatic segmentation of intramedullary lesions commonly seen in SCI replaces the tedious manual annotation process and enables the extraction of relevant lesion morphometrics in large cohorts. The proposed model segments lesions across different etiologies, scanner manufacturers, and heterogeneous image resolutions. SCIseg is open-source and accessible through the Spinal Cord Toolbox.

Artificial intelligence CAD tools in trauma imaging: a scoping review from the American Society of Emergency Radiology (ASER) AI/ML Expert Panel

BACKGROUND

AI/ML CAD tools can potentially improve outcomes in the high-stakes, high-volume model of trauma radiology. No prior scoping review has been undertaken to comprehensively assess tools in this subspecialty.

PURPOSE

To map the evolution and current state of trauma radiology CAD tools along key dimensions of technology readiness.

METHODS

Following a search of databases, abstract screening, and full-text document review, CAD tool maturity was charted using elements of data curation, performance validation, outcomes research, explainability, user acceptance, and funding patterns. Descriptive statistics were used to illustrate key trends.

RESULTS

A total of 4052 records were screened, and 233 full-text articles were selected for content analysis. Twenty-one papers described FDA-approved commercial tools, and 212 reported algorithm prototypes. Works ranged from foundational research to multi-reader multi-case trials with heterogeneous external data. Scalable convolutional neural network-based implementations increased steeply after 2016 and were used in all commercial products; however, options for explainability were narrow. Of FDA-approved tools, 9/10 performed detection tasks. Dataset sizes ranged from < 100 to > 500,000 patients, and commercialization coincided with public dataset availability. Cross-sectional torso datasets were uniformly small. Data curation methods with ground truth labeling by independent readers were uncommon. No papers assessed user acceptance, and no method included human-computer interaction. The USA and China had the highest research output and frequency of research funding.

CONCLUSIONS

Trauma imaging CAD tools are likely to improve patient care but are currently in an early stage of maturity, with few FDA-approved products for a limited number of uses. The scarcity of high-quality annotated data remains a major barrier.

Magnetic resonance image segmentation of the compressed spinal cord in patients with degenerative cervical myelopathy using convolutional neural networks

PURPOSE

Spinal cord segmentation is the first step in atlas-based spinal cord image analysis, but segmentation of compressed spinal cords from patients with degenerative cervical myelopathy is challenging. We applied convolutional neural network models to segment the spinal cord from T2-weighted axial magnetic resonance images of DCM patients. Furthermore, we assessed the correlation between the cross-sectional area segmented by this network and the neurological symptoms of the patients.

METHODS

The CNN architecture was built using U-Net and DeepLabv3 + and PyTorch. The CNN was trained on 2762 axial slices from 174 patients, and an additional 517 axial slices from 33 patients were held out for validation and 777 axial slices from 46 patients for testing. The performance of the CNN was evaluated on a test dataset with Dice coefficients as the outcome measure. The ratio of CSA at the maximum compression level to CSA at the C2 level, as segmented by the CNN, was calculated. The correlation between the spinal cord CSA ratio and the Japanese Orthopaedic Association score in DCM patients from the test dataset was investigated using Spearman's rank correlation coefficient.

RESULTS

The best Dice coefficient was achieved when U-Net was used as the architecture and EfficientNet-b7 as the model for transfer learning. Spearman's r_s between the spinal cord CSA ratio and the JOA score of DCM patients was 0.38 (p = 0.007), showing a weak correlation.

CONCLUSION

Using deep learning with magnetic resonance images of deformed spinal cords as training data, we were able to segment compressed spinal cords of DCM patients with a high concordance with expert manual segmentation. In addition, the spinal cord CSA ratio was weakly, but significantly, correlated with neurological symptoms. Our study demonstrated the first steps needed to implement automated atlas-based analysis of DCM patients.

Automated Cervical Spinal Cord Segmentation in Real-World MRI of Multiple Sclerosis Patients by Optimized Hybrid Residual Attention-Aware Convolutional Neural Networks

Magnetic resonance (MR) imaging is the most sensitive clinical tool in the diagnosis and monitoring of multiple sclerosis (MS) alterations. Spinal cord evaluation has gained interest in this clinical scenario in recent years, but, unlike the brain, there is a more limited choice of algorithms to assist spinal cord segmentation. Our goal was to investigate and develop an automatic MR cervical cord segmentation method, enabling automated and seamless spinal cord atrophy assessment and setting the stage for the development of an aggregated algorithm for the extraction of lesion-related imaging biomarkers. The algorithm was developed using a real-world MR imaging dataset of 121 MS patients (96 cases used as a training dataset and 25 cases as a validation dataset). Transversal, 3D T1-weighted gradient echo MR images (TE/TR/FA = 1.7-2.7 ms/5.6-8.2 ms/12°) were acquired in a 3 T system (Signa HD, GEHC) as standard of care in our clinical practice. Experienced radiologists supervised the manual labelling, which was considered the ground-truth. The 2D convolutional neural network consisted of a hybrid residual attention-aware segmentation method trained to delineate the cervical spinal cord. The training was conducted using a focal loss function, based on the Tversky index to address label imbalance, and an automatic optimal learning rate finder. Our automated model provided an accurate segmentation, achieving a validation DICE coefficient of 0.904 ± 0.101 compared with the manual delineation. An automatic method for cervical spinal cord segmentation on T1-weighted MR images was successfully implemented. It will have direct implications serving as the first step for accelerating the process for MS staging and follow-up through imaging biomarkers.

Determining the short-term neurological prognosis for acute cervical spinal cord injury using machine learning

It is challenging to predict neurological outcomes of acute spinal cord injury (SCI) considering issues such as spinal shock and injury heterogeneity. Deep learning-based radiomics (DLR) were developed to quantify the radiographic characteristics automatically using a convolutional neural network (CNN), and to potentially allow the prognostic stratification of patients. We aimed to determine the functional prognosis of patients with cervical SCI using machine learning approach based on MRI and to assess the ability to predict the neurological outcomes. We retrospectively analyzed the medical records of SCI patients (n=215) who had undergone MRI and had an American Spinal cord Injury Association Impairment Scale (AIS) assessment at 1 month after injury, enrolled with a total of 294 MR images. Sagittal T2-weighted MR images were used for the CNN training and validation. The deep learning framework TensorFlow was used to construct the CNN architecture. After we calculated the probability of the AIS grade using the DLR, we built the identification model based upon the random forest using 3 features: the probability of each AIS grade obtained by the DLR method, age, and the initial AIS grade at admission. We performed a statistical evaluation between the actual and predicted AIS. The accuracy, precision, recall and f1 score of the ensemble model based on the DLR and RF were 0.714, 0.590, 0.565 and 0.567, respectively. The present study demonstrates that prediction of the short-term neurological outcomes for acute cervical spinal cord injury based on MRI using machine learning is feasible.

A pipeline to quantify spinal cord atrophy with deep learning: Application to differentiation of MS and NMOSD patients

PURPOSE

Quantitative measurement of various anatomical regions of the brain and spinal cord (SC) in MRI images are used as unique biomarkers to consider progress and effects of demyelinating diseases of the central nervous system. This paper presents a fully-automated image processing pipeline which quantifies the SC volume of MRI images.

METHODS

In the proposed pipeline, after conducting some pre-processing tasks, a deep convolutional network is utilized to segment the spinal cord cross-sectional area (SCCSA) of each slice. After full segmentation, certain extra slices interpolate between each two adjacent slices using the shape-based interpolation method. Then, a 3D model of the SC is reconstructed, and, by counting the voxels of it, the SC volume is calculated. The performance of the proposed method for the SCCSA segmentation is evaluated on 140 MRI images. Subsequently, to demonstrate the application of the proposed pipeline, we study the differentiations of SC atrophy between 38 Multiple Sclerosis (MS) and 25 Neuromyelitis Optica Spectrum Disorder (NMOSD) patients.

RESULTS

The experimental results of the SCCSA segmentation indicate that the proposed method, adapted by Mask R-CNN, presented the most satisfactory result with the average Dice coefficient of 0.96. For this method, statistical metrics including sensitivity, specificity, accuracy, and precision are 97.51%, 99.98%, 99.92%, and 98.04% respectively. Moreover, the t-test result (p-value = 0.00089) verified a significant difference between the SC atrophy of MS and NMOSD patients.

CONCLUSION

The pipeline efficiently quantifies the SC volume of MRI images and can be utilized as an affordable computer-aided tool for diagnostic purposes.

Cervical and thoracic cord atrophy in multiple sclerosis phenotypes: Quantification and correlation with clinical disability

•

Spinal cord atrophy is prevalent across multiple sclerosis phenotypes.

•

It correlates with disability, especially in relapsing-remitting patients.

•

This correlation can be demonstrated both cross-sectionally and longitudinally.

•

Cervical atrophy is highly associated with disability and disease progression.

•

Thoracic atrophy contributes to improved correlation and radiological subgrouping.

Objective

We sought to characterize spinal cord atrophy along the entire spinal cord in the major multiple sclerosis (MS) phenotypes, and evaluate its correlation with clinical disability.

Methods

Axial T₁-weighted images were automatically reformatted at each point along the cord. Spinal cord cross‐sectional area (SCCSA) were calculated from C1-T10 vertebral body levels and profile plots were compared across phenotypes. Average values from C2-3, C4-5, and T4-9 regions were compared across phenotypes and correlated with clinical scores, and then categorized as atrophic/normal based on z-scores derived from controls, to compare clinical scores between subgroups. In a subset of relapsing-remitting cases with longitudinal scans these regions were compared to change in clinical scores.

Results

The cross-sectional study consisted of 149 adults diagnosed with relapsing-remitting MS (RRMS), 49 with secondary-progressive MS (SPMS), 58 with primary-progressive MS (PPMS) and 48 controls. The longitudinal study included 78 RRMS cases. Compared to controls, all MS groups had smaller average regions except RRMS in T4-9 region. In all MS groups, SCCSA from all regions, particularly the cervical cord, correlated with most clinical measures. In the RRMS cohort, 22% of cases had at least one atrophic region, whereas in progressive MS the rate was almost 70%. Longitudinal analysis showed correlation between clinical disability and cervical cord thinning.

Conclusions

Spinal cord atrophy was prevalent across MS phenotypes, with regional measures from the RRMS cohort and the progressive cohort, including SPMS and PPMS, being correlated with disability. Longitudinal changes in the spinal cord were documented in RRMS cases, making it a potential marker for disease progression. While cervical SCCSA correlated with most disability and progression measures, inclusion of thoracic measurements improved this correlation and allowed for better subgrouping of spinal cord phenotypes. Cord atrophy is an important and easily obtainable imaging marker of clinical and sub-clinical progression in all MS phenotypes, and such measures can play a key role in patient selection for clinical trials.

Diagnostic blood RNA profiles for human acute spinal cord injury

Diagnosis of spinal cord injury (SCI) severity at the ultra-acute stage is of great importance for emergency clinical care of patients as well as for potential enrollment into clinical trials. The lack of a diagnostic biomarker for SCI has played a major role in the poor results of clinical trials. We analyzed global gene expression in peripheral white blood cells during the acute injury phase and identified 197 genes whose expression changed after SCI compared with healthy and trauma controls and in direct relation to SCI severity. Unsupervised coexpression network analysis identified several gene modules that predicted injury severity (AIS grades) with an overall accuracy of 72.7% and included signatures of immune cell subtypes. Specifically, for complete SCIs (AIS A), ROC analysis showed impressive specificity and sensitivity (AUC: 0.865). Similar precision was also shown for AIS D SCIs (AUC: 0.938). Our findings indicate that global transcriptomic changes in peripheral blood cells have diagnostic and potentially prognostic value for SCI severity.

Injury volume extracted from MRI predicts neurologic outcome in acute spinal cord injury: A prospective TRACK-SCI pilot study

Conventional MRI measures of traumatic spinal cord injury severity largely rely on 2-dimensional injury characteristics such as intramedullary lesion length and cord compression. Recent advances in spinal cord (SC) analysis have led to the development of a robust anatomic atlas incorporated into an open-source platform called the Spinal Cord Toolbox (SCT) that allows for quantitative volumetric injury analysis. In the current study, we evaluate the prognostic value of volumetric measures of spinal cord injury on MRI following registration of T2-weighted (T2w) images and segmented lesions from acute SCI patients with a standardized atlas. This IRB-approved prospective cohort study involved the image analysis of 60 blunt cervical SCI patients enrolled in the TRACK-SCI clinical research protocol. Axial T2w MRI data obtained within 24 h of injury were processed using the SCT. Briefly, SC MRIs were automatically segmented using the sct_deepseg_sc tool in the SCT and segmentations were manually corrected by a neuro-radiologist. Lesion volume data were used as predictor variables for correlation with lower extremity motor scores at discharge. Volumetric MRI measures of T2w signal abnormality comprising the SCI lesion accurately predict lower extremity motor scores at time of patient discharge. Similarly, MRI measures of injury volume significantly correlated with motor scores to a greater degree than conventional 2-D metrics of lesion size. The volume of total injury and of injured spinal cord motor regions on T2w MRI is significantly and independently associated with neurologic outcome at discharge after injury.

Diverse Applications of Artificial Intelligence in Neuroradiology

Recent advances in artificial intelligence (AI) and deep learning (DL) hold promise to augment neuroimaging diagnosis for patients with brain tumors and stroke. Here, the authors review the diverse landscape of emerging neuroimaging applications of AI, including workflow optimization, lesion segmentation, and precision education. Given the many modalities used in diagnosing neurologic diseases, AI may be deployed to integrate across modalities (MR imaging, computed tomography, PET, electroencephalography, clinical and laboratory findings), facilitate crosstalk among specialists, and potentially improve diagnosis in patients with trauma, multiple sclerosis, epilepsy, and neurodegeneration. Together, there are myriad applications of AI for neuroradiology."

Simultaneous voxel-wise analysis of brain and spinal cord morphometry and microstructure within the SPM framework

To validate a simultaneous analysis tool for the brain and cervical cord embedded in the statistical parametric mapping (SPM) framework, we compared trauma-induced macro- and microstructural changes in spinal cord injury (SCI) patients to controls. The findings were compared with results obtained from existing processing tools that assess the brain and spinal cord separately. A probabilistic brain-spinal cord template (BSC) was generated using a generative semi-supervised modelling approach. The template was incorporated into the pre-processing pipeline of voxel-based morphometry and voxel-based quantification analyses in SPM. This approach was validated on T1-weighted scans and multiparameter maps, by assessing trauma-induced changes in SCI patients relative to controls and comparing the findings with the outcome from existing analytical tools. Consistency of the MRI measures was assessed using intraclass correlation coefficients (ICC). The SPM approach using the BSC template revealed trauma-induced changes across the sensorimotor system in the cord and brain in SCI patients. These changes were confirmed with established approaches covering brain or cord, separately. The ICC in the brain was high within regions of interest, such as the sensorimotor cortices, corticospinal tracts and thalamus. The simultaneous voxel-wise analysis of brain and cervical spinal cord was performed in a unique SPM-based framework incorporating pre-processing and statistical analysis in the same environment. Validation based on a SCI cohort demonstrated that the new processing approach based on the brain and cord is comparable to available processing tools, while offering the advantage of performing the analysis simultaneously across the neuraxis.

Spinal cord atrophy in a primary progressive multiple sclerosis trial: Improved sample size using GBSI

•

The GBSI provided clinically meaningful measurements of spinal cord atrophy, with low sample size.

•

Deriving spinal cord atrophy from brain MRI using the GBSI is easier than spinal cord MRI.

•

Spinal cord atrophy on GBSI could be used as a secondary outcome measure.

Background

We aimed to evaluate the implications for clinical trial design of the generalised boundary-shift integral (GBSI) for spinal cord atrophy measurement.

Methods

We included 220 primary-progressive multiple sclerosis patients from a phase 2 clinical trial, with baseline and week-48 3DT1-weighted MRI of the brain and spinal cord (1 × 1 × 1 mm³), acquired separately. We obtained segmentation-based cross-sectional spinal cord area (CSA) at C1-2 (from both brain and spinal cord MRI) and C2-5 levels (from spinal cord MRI) using DeepSeg, and, then, we computed corresponding GBSI.

Results

Depending on the spinal cord segment, we included 67.4–98.1% patients for CSA measurements, and 66.9–84.2% for GBSI. Spinal cord atrophy measurements obtained with GBSI had lower measurement variability, than corresponding CSA. Looking at the image noise floor, the lowest median standard deviation of the MRI signal within the cerebrospinal fluid surrounding the spinal cord was found on brain MRI at the C1-2 level. Spinal cord atrophy derived from brain MRI was related to the corresponding measures from dedicated spinal cord MRI, more strongly for GBSI than CSA. Spinal cord atrophy measurements using GBSI, but not CSA, were associated with upper and lower limb motor progression.

Discussion

Notwithstanding the reduced measurement variability, the clinical correlates, and the possibility of using brain acquisitions, spinal cord atrophy using GBSI should remain a secondary outcome measure in MS studies, until further advancements increase the quality of acquisition and reliability of processing.

Automatic Vertebral Body Segmentation Based on Deep Learning of Dixon Images for Bone Marrow Fat Fraction Quantification

Background: Bone marrow fat (BMF) fraction quantification in vertebral bodies is used as a novel imaging biomarker to assess and characterize chronic lower back pain. However, manual segmentation of vertebral bodies is time consuming and laborious. Purpose: (1) Develop a deep learning pipeline for segmentation of vertebral bodies using quantitative water-fat MRI. (2) Compare BMF measurements between manual and automatic segmentation methods to assess performance. Materials and Methods: In this retrospective study, MR images using a 3D spoiled gradient-recalled echo (SPGR) sequence with Iterative Decomposition of water and fat with Echo Asymmetry and Least-squares estimation (IDEAL) reconstruction algorithm were obtained in 57 subjects (28 women, 29 men, mean age, 47.2 ± 12.6 years). An artificial network was trained for 100 epochs on a total of 165 lumbar vertebrae manually segmented from 31 subjects. Performance was assessed by analyzing the receiver operating characteristic curve, precision-recall, F1 scores, specificity, sensitivity, and similarity metrics. Bland-Altman analysis was used to assess performance of BMF fraction quantification using the predicted segmentations. Results: The deep learning segmentation method achieved an AUC of 0.92 (CI 95%: 0.9186, 0.9195) on a testing dataset (n = 24 subjects) on classification of pixels as vertebrae. A sensitivity of 0.99 and specificity of 0.80 were achieved for a testing dataset, and a mean Dice similarity coefficient of 0.849 ± 0.091. Comparing manual and automatic segmentations on fat fraction maps of lumbar vertebrae (n = 124 vertebral bodies) using Bland-Altman analysis resulted in a bias of only -0.605% (CI 95% = -0.847 to -0.363%) and agreement limits of -3.275% and +2.065%. Automatic segmentation was also feasible in 16 ± 1 s. Conclusion: Our results have demonstrated the feasibility of automated segmentation of vertebral bodies using deep learning models on water-fat MR (Dixon) images to define vertebral regions of interest with high specificity. These regions of interest can then be used to quantify BMF with comparable results as manual segmentation, providing a framework for completely automated investigation of vertebral changes in CLBP.

XGBoost, a Machine Learning Method, Predicts Neurological Recovery in Patients with Cervical Spinal Cord Injury

The accurate prediction of neurological outcomes in patients with cervical spinal cord injury (SCI) is difficult because of heterogeneity in patient characteristics, treatment strategies, and radiographic findings. Although machine learning algorithms may increase the accuracy of outcome predictions in various fields, limited information is available on their efficacy in the management of SCI. We analyzed data from 165 patients with cervical SCI, and extracted important factors for predicting prognoses. Extreme gradient boosting (XGBoost) as a machine learning model was applied to assess the reliability of a machine learning algorithm to predict neurological outcomes compared with that of conventional methodology, such as a logistic regression or decision tree. We used regularly obtainable data as predictors, such as demographics, magnetic resonance variables, and treatment strategies. Predictive tools, including XGBoost, a logistic regression, and a decision tree, were applied to predict neurological improvements in the functional motor status (ASIA [American Spinal Injury Association] Impairment Scale [AIS] D and E) 6 months after injury. We evaluated predictive performance, including accuracy and the area under the receiver operating characteristic curve (AUC). Regarding predictions of neurological improvements in patients with cervical SCI, XGBoost had the highest accuracy (81.1%), followed by the logistic regression (80.6%) and the decision tree (78.8%). Regarding AUC, the logistic regression showed 0.877, followed by XGBoost (0.867) and the decision tree (0.753). XGBoost reliably predicted neurological alterations in patients with cervical SCI. The utilization of predictive machine learning algorithms may enhance personalized management choices through pre-treatment categorization of patients.

Predictive Modeling of Outcomes After Traumatic and Nontraumatic Spinal Cord Injury Using Machine Learning: Review of Current Progress and Future Directions

Machine learning represents a promising frontier in epidemiological research on spine surgery. It consists of a series of algorithms that determines relationships between data. Machine learning maintains numerous advantages over conventional regression techniques, such as a reduced requirement for a priori knowledge on predictors and better ability to manage large datasets. Current studies have made extensive strides in employing machine learning to a greater capacity in spinal cord injury (SCI). Analyses using machine learning algorithms have been done on both traumatic SCI and nontraumatic SCI, the latter of which typically represents degenerative spine disease resulting in spinal cord compression, such as degenerative cervical myelopathy. This article is a literature review of current studies published in traumatic and nontraumatic SCI that employ machine learning for the prediction of a host of outcomes. The studies described utilize machine learning in a variety of capacities, including imaging analysis and prediction in large epidemiological data sets. We discuss the performance of these machine learning-based clinical prognostic models relative to conventional statistical prediction models. Finally, we detail the future steps needed for machine learning to become a more common modality for statistical analysis in SCI.