Abstract
Objectives:
To determine the interrater reliability between an automated and manual measure of lesion damage following spinal cord injury (SCI) using T2-weighted magnetic resonance images (MRI).
Methods:
Twenty-one MRIs were collected from patients who had completed rehabilitation at Craig Hospital. Manual measurements of midsagittal tissue bridges were conducted by an experienced rater using OsiriX (Pixmeo Sarl, Geneva, Switzerland), and automated measures were taken using the SCIsegV2 automated function through the Spinal Cord Toolbox (SCT). Manual and automated measurements were compared using intraclass correlation coefficients (ICC). Percentage agreement and Cohen's kappa statistic were calculated to compare detection of midsagittal tissue bridges.
Results:
ICCs between the manual and automated measures were excellent (ICC 0.94, 95% CI 0.84-0.97, P < .001, for ventral tissue bridges; ICC 0.99, 95% CI 0.97-0.99, P < .001, for dorsal tissue bridges). Percentage agreement between raters was 90.8% for ventral, dorsal, and any midsagittal tissue bridge. Cohen's kappa for the detection of tissue bridges showed substantial agreement between the two raters for ventral, dorsal, and any tissue bridges (0.81, P < .001; 0.79, P < .001; and 0.81, P < .001, respectively).
Conclusion:
Measurements of midsagittal tissue bridges between manual and automated raters are reliable. Automated measurements may help to expedite research related to midsagittal tissue bridges and functional outcomes for individuals with SCI.
Keywords: automated measurement, magnetic resonance imaging, reliability, spinal cord injury
Introduction
Spinal cord injury (SCI) is a life-changing event that has substantial effects on an individual, their families, and their communities at large. It may impact an individual's mobility, cardiovascular function, autonomic function, and overall quality of life.1 Accurate predictors of recovery following SCI are vital for determining appropriate medical and rehabilitative interventions. The current standard for classifying injury and determining prognosis following SCI is the International Standards for the Neurological Classification of SCI (ISNCSCI).2 The ISNCSCI uses a standardized assessment of motor and sensory scores to determine the neurological classification of an individual following SCI. Those with similar scores in the acute phase of injury, however, may show variability in their long-term functional recovery.3 Thus, current standards for prognosis may be limited in their ability to accurately predict long-term outcomes.
Imaging biomarkers have been evaluated as potential predictors for neurological recovery in individuals with SCI.4–8 Magnetic resonance imaging (MRI) can be used to measure spinal cord lesion characteristics following injury. These characteristics provide insight into the extent of damage to the spinal cord and the presence of spared neural tissue. Quantifying spared neural tissue may provide insight into patients’ potential for functional recovery. A promising quantitative biomarker of spared neural tissue, midsagittal tissue bridges, are measured on midsagittal T2-weighted (T2w) MRIs.4,5,9–13
Midsagittal tissue bridges appear as hypointense spaces on T2w images, adjacent to the relatively hyperintense intramedullary lesion and the surrounding cerebrospinal fluid, and can be appreciated on both the dorsal and ventral side of the lesion.5 Tissue bridges may emerge around 3 weeks post injury and persist through the chronic stage.10,13,14 These bridges represent spared neural tissue permissive of electrophysiological information flow and may indicate more potential for recovery.4,5,11–13,15 A cross-sectional study10 and a longitudinal cohort study16 have linked midsagittal tissue bridges with walking ability for individuals with SCIs. Tissue bridges may have prognostic value for improvements in sensorimotor function from baseline over the recovery process.13
Midsagittal tissue bridges, so far quantified manually, show promise for predicting long-term clinical outcomes. Although manual measurement has been shown to be reliable,13,17,18 it can be time-consuming and requires specialized training to perform. In an effort to decrease the burden associated with this process, an automated tool has been recently developed that segments and quantifies lesion characteristics from sagittal MRI scans using Spinal Cord Toolbox (SCT).7
SCT is a free, open-source set of tools used for the analysis of spinal cord MRIs.19 Included within SCT are several automated measurements of spinal cord lesion characteristics, including midsagittal tissue bridges.7,8 Automated measures may provide a more standardized and streamlined way to measure midsagittal tissue bridges and other characteristics. Implementation of these measures could therefore greatly benefit research efforts related to the study of spinal cord lesions. The reliability and validity of automated midsagittal tissue bridge measurements was established using 15 participants and found no significant differences between manual and automated measurements.7 These measures, however, have yet to be validated externally. Accordingly, this study aims to evaluate the interrater reliability of this tool using an external dataset.
Methods
This was a retrospective secondary analysis of de-identified data. T2w sagittal MRI scans were acquired from participants with traumatic cervical and thoracic SCIs admitted to the local inpatient rehabilitation hospital. Inclusion criteria included postoperative cervical or thoracic scans with a visible lesion that could be manually measured. Images were excluded if there was excessive artifact from surgical hardware such that lesion characteristics could not be measured. Initially, 25 images were selected for examination based on a convenience sample of available images that were not used to train SCIsegV2. Four of the images were excluded due to excessive artifact. None of the participants’ images in this study were used during the deep learning process used to generate the automated function. The dataset used in this study was approved by the local institutional review board. Specifics of the MRI are as follows: General Electric 1.5T Signa scanner using a sagittal T2w fat-saturated fast recovery fast spin-echo sequence (acquisition matrix size = 448 × 256, flip angle = 90°, repetition time = 4037 ms, echo time = 106.56 ms, number of excitations = 4, echo train length = 25, field of view = 260 mm x 260 mm, slice thickness = 3 mm, space between slices = 4 mm, voxel size = 0.47 x 0.47 x 3 mm, acquisition time = 1.39 minutes).
Measurement protocol
Our laboratory has worked with board-certified neuroradiologists for previously published tissue bridge investigations.8 For this study, one research rater with 4 years of experience (W. T.) was assigned to perform manual measurements of midsagittal tissue bridges, while a separate research rater with 3 years of experience (J. C.) was assigned to perform automated measurements using SCIsegV2. The automated approach was not used to guide the manual selection of the midsagittal slice. Rather, to guide the manual approach and to maintain consistency between the raters with regards to the most appropriate midsagittal slice to be used for assessment, the following steps were taken: First, the ideal midsagittal slice was defined as the one with the greatest amount of observable continuous spinal cord. Second, if multiple slices appeared to show similar continuity of spinal cord, these slices were co-registered with axial MRI to determine which sagittal slice was located most towards the midline. Third, if this was not conclusive, the most appropriate slice was determined based on image quality to ensure that the lesion and potential tissue bridges could be accurately assessed.
Manual measurement
On the midsagittal slice, manual measurement of tissue bridges was completed using the line tool within OsiriX image processing software (Pixmeo, Geneva Switzerland). The lesion was visually identified as the area of hyperintensity within the spinal cord. The rater then identified if there were observable tissue bridges both ventral and dorsal to the lesion. If present, ventral and dorsal tissue bridges were calculated as the smallest distance between the lesion and the surrounding cerebrospinal fluid, both of which appear as hyperintense relative to the tissue bridges (see Figure 1).
Figure 1.
Example of manual measurement of midsagittal tissue bridges on a T2-weighted image. Ventral (green) and dorsal (yellow) tissue bridges are quantified as the shortest distance between the hyperintense spinal cord lesion and surrounding cerebrospinal fluid.
Automated segmentation
Automated segmentation was completed using the SCT SCIsegV2 function (Spinal Cord Toolbox version 6.4). The automated rater was blinded to the results of the manually measured segmentation and vice versa. The automated process of segmentation through the SCT is detailed by Karthik et al.8 It involves automated segmentation of the spinal cord followed by automated segmentation of the lesion (see Figure 2). A separate command is then used to automatically quantify lesion metrics, including midsagittal tissue bridges (see Figure 3). To maintain consistency, comparisons were made only to the midsagittal slice as determined by the protocol described above. Quality checks were completed for each subject to ensure proper measurement.
Figure 2.
Example of the automated measurement process. From left to right, original T2-weighted midsagittal image, automatic segmentation of the spinal cord, and segmentation of the lesion.
Figure 3.
Example of manual and automated measurements: manually measured tissue bridges on the left, readout of automated measurements on the right (participant 1).
Data analysis
Statistics were calculated using Stata statistical analysis software (version 18.5; College Station, TX). Automated measures of midsagittal tissue bridges were compared to manually measured midsagittal tissue bridges in terms of (1) presence of any tissue bridge, (2) presence of ventral tissue bridges, (3) presence of dorsal tissue bridges, and (4) total width of these tissue bridges.
Patient demographics including age, sex, location of injury (cervical vs. thoracic injury), American Spinal Injury Association Impairment Scale (AIS) classification, and time between injury and imaging were collected. To calculate agreement on the presence or absence of tissue bridges, we calculated percentage agreement and Cohen's kappa (κ). The interrater reliability of midsagittal tissue bridge width measurements between raters was assessed using intraclass correlation coefficients with a two-way mixed, average measures model. A two-way mixed-effects model was used as both raters (manual rater and automated measurement) were defined, and each subject was rated by both raters. Means and mean differences were calculated between the raters for ventral, dorsal, and total tissue bridge widths. A Bland-Altman plot was created to visualize the agreement between raters across tissue bridge widths. Significance values were set to P < .05.
Results
Of 21 participants, the average age was 43 (SD 17) years. The participants included 16 males and 5 females, and 19 sustained cervical level injuries and 2 sustained thoracic level injuries. The sample included 12 individuals with AIS A, 2 with AIS B, 2 with AIS C, and 5 with AIS D. The average time between injury and MRI collection was 43.42 days (SD 30.24). Participant demographics are shown in Table 1.
Table 1.
Summary of participant demographics
| Participant ID number | Age, years | Sex | Cervical or thoracic (C/T) | AIS classification | Time between injuryand MRI, days |
|---|---|---|---|---|---|
| 1 | 30 | Male | C | D | 20 |
| 2 | 59 | Female | C | D | 23 |
| 3 | 36 | Male | C | B | 70 |
| 4 | 28 | Female | C | B | 24 |
| 5 | 30 | Male | C | A | 30 |
| 6 | 67 | Female | C | A | 67 |
| 7 | 67 | Male | C | D | 46 |
| 8 | 41 | Male | C | D | 45 |
| 9 | 35 | Male | C | A | 23 |
| 10 | 32 | Male | C | A | 139 |
| 11 | 26 | Male | C | A | 30 |
| 12 | 64 | Female | C | A | 55 |
| 13 | 26 | Male | C | A | 39 |
| 14 | 69 | Male | C | D | 21 |
| 15 | 28 | Male | T | A | 31 |
| 16 | 19 | Male | C | A | 29 |
| 17 | 62 | Male | C | C | 21 |
| 18 | 35 | Male | C | A | 104 |
| 19 | 48 | Female | C | C | 42 |
| 20 | 27 | Female | T | A | 28 |
| 21 | 56 | Male | C | A | 25 |
Note: AIS = American Spinal Injury Association Impairment Scale; MRI = magnetic resonance imaging.
Results for the percentage agreement between manual and automated measures can be found in Table 2. Percentage agreement for the presence of any tissue bridge was calculated at 90.48% (Cohen's kappa 0.81, P < .001), for the presence of ventral tissue bridges at 90.48%, and for the presence of dorsal tissue bridges at 90.48%. Cohen's kappa was 0.81 (P < .001) for presence of any tissue bridge, 0.81 (P < .001) for presence of ventral tissue bridges, and 0.79 (P < .001) for dorsal tissue bridges.
Table 2.
Agreement between rater and SCIsegV2 for presence or absence of midsagittal tissue bridges
| Participant ID | Presence of any tissue bridge | Presence of ventral tissue bridges | Presence of dorsal tissue bridges | |||
|---|---|---|---|---|---|---|
| Automated | Manual | Automated | Manual | Automated | Manual | |
| 1 | Yes | Yes | Yes | Yes | Yes | Yes |
| 2 | Yes | Yes | Yes | Yes | No | Yes |
| 3 | No | No | No | No | No | No |
| 4 | Yes | Yes | Yes | Yes | Yes | Yes |
| 5 | Yes | No | Yes | No | No | No |
| 6 | Yes | Yes | Yes | Yes | Yes | Yes |
| 7 | Yes | Yes | Yes | Yes | Yes | Yes |
| 8 | Yes | Yes | Yes | Yes | Yes | Yes |
| 9 | Yes | Yes | Yes | Yes | No | No |
| 10 | No | No | No | No | No | No |
| 11 | No | No | No | No | No | No |
| 12 | No | No | No | No | No | No |
| 13 | Yes | Yes | Yes | Yes | Yes | Yes |
| 14 | No | No | No | No | No | No |
| 15 | No | No | No | No | No | No |
| 16 | No | No | No | No | No | No |
| 17 | Yes | Yes | Yes | Yes | No | Yes |
| 18 | No | No | No | No | No | No |
| 19 | No | No | No | No | No | No |
| 29 | Yes | No | Yes | No | No | No |
| 21 | No | No | No | No | No | No |
| Percent agreement | 90.48%, kappa=0.811, P < .001 | 90.48%, kappa=0.810, P < .001 | 90.48%, kappa=.788, P < .001 | |||
Interrater reliability between the manual rater and automated measurements of total, ventral, and dorsal bridge width was excellent (ICC 0.94, 95% CI 0.84-0.97, P < .001; ICC 0.99, 95% CI 0.97-0.99, P < .001; ICC 0.96, 95% CI 0.91-0.98, P < .001, respectively) and can be seen in Table 3. Mean differences between the automated and manual raters were -0.13 mm (SD 0.44) for ventral tissue bridges, -0.02 mm (SD 0.13) for dorsal tissue bridges, and -0.16 mm (SD 0.48) for total tissue bridges. Agreement between automated and manual raters for combined midsagittal tissue bridge widths can be visualized in a Bland-Altman plot (see Figure 4).
Table 3.
Comparison between rater and SCIsegV2 for measures of ventral, dorsal, and total tissue bridges in millimeters
| Participant ID | Ventral tissue bridge width, mm | Dorsal tissue bridge width, mm | Total tissue bridge width, mm | |||
|---|---|---|---|---|---|---|
| Automated | Manual | Automated | Manual | Automated | Manual | |
| 1 | 1.01 | 0.74 | 2.01 | 2.1 | 3.02 | 2.84 |
| 2 | 0.55 | 1.75 | 0 | 0.19 | 0.55 | 1.94 |
| 3 | 0 | 0 | 0 | 0 | 0 | 0 |
| 4 | 0.45 | 0.69 | 0.45 | 0.91 | 0.9 | 1.6 |
| 5 | 0.47 | 0 | 0 | 0 | 0.47 | 0 |
| 6 | 0.5 | 0.6 | 0.47 | 0.47 | 0.97 | 1.07 |
| 7 | 2.03 | 2.77 | 0.51 | 0.24 | 2.54 | 3.01 |
| 8 | 1.39 | 1.87 | 1.86 | 1.7 | 3.25 | 3.57 |
| 9 | 0.5 | 0.18 | 0 | 0 | 0.5 | 0.18 |
| 10 | 0 | 0 | 0 | 0 | 0 | 0 |
| 11a | 0 | 0 | 0 | 0 | 0 | 0 |
| 12 | 0 | 0 | 0 | 0 | 0 | 0 |
| 13 | 1.38 | 2.14 | 0.5 | 0.57 | 1.88 | 2.71 |
| 14 | 0 | 0 | 0 | 0 | 0 | 0 |
| 15 | 0 | 0 | 0 | 0 | 0 | 0 |
| 16 | 0 | 0 | 0 | 0 | 0 | 0 |
| 17 | 2.6 | 3.47 | 0 | 0.13 | 2.6 | 3.6 |
| 18 | 0 | 0 | 0 | 0 | 0 | 0 |
| 19 | 0 | 0 | 0 | 0 | 0 | 0 |
| 20 | 0.54 | 0 | 0 | 0 | 0.54 | 0 |
| 21 | 0 | 0 | 0 | 0 | 0 | 0 |
| Mean, mm | 0.54 (SD 0.74) |
0.68 (SD 1.06) |
0.28 (SD 0.58) |
0.30 (SD 0.59) |
0.82 (SD 1.12) |
0.98 (SD 1.37) |
| Mean difference, mm | -0.13 (SD 0.44) | -0.02 (SD 0.13) | -0.16 (SD 0.47) | |||
| Intraclass correlation coefficient (ICC) | ICC 0.94, 95% CI 0.84-0.97, P < .001 | ICC 0.99, 95% CI 0.97-0.99, P < .001 | ICC 0.96, 95% CI 0.91-0.98, P < .001 | |||
In one of the samples, “NaN” response was given for presence of lesion, however we considered this to be a zero due to clear completeness of lesion in the image.
Figure 4.
Bland Altman plot displaying the agreement between automated and manual raters for ventral and dorsal midsagittal tissue bridge widths. Note that 23 points are plotted at zero for both the x and y axes.
Discussion
In this study, we sought to determine the interrater reliability of an automated measure of spinal midsagittal tissue bridges after SCI as compared to manual segmentations, using an external dataset that SCIsegV2 has not been trained on.7 As far as this team is aware, no other validations of this tool have been performed using external datasets. Manual segmentation-derived MRI biomarkers have been shown to be reliable and are considered the gold standard.13,17,18 This study found excellent interrater reliability between the automated and manual measurements of midsagittal tissue bridges. This is consistent with the finding of no statistically significant differences between the measurements of midsagittal tissue bridges computed using different methods by the development team.7
Quantified around 3 weeks post injury or beyond, the measurement of midsagittal tissue bridges provides valuable information that may help to guide rehabilitation planning and improve prognostication following SCI, in the subacute and chronic phases of injury.13 The quantification of spared neural tissue may be useful for understanding an individual's potential for neurologic recovery. Estimates of spared corticospinal tracts, for example, have been shown to be related to motor outcomes, while more intact dorsal columns have been shown to relate to sensory scores.4,15 These biomarkers may, therefore, help to guide novel interventions targeting neurological recovery. Recently, innovative neuromodulatory interventions such as spinal stimulation have shown promise in restoring motor function in individuals with motor complete SCI.20–22 Not all individuals, however, respond to these interventions. Biomarkers such as midsagittal tissue bridges could help to determine responders and nonresponders to such interventions.23
Our study findings reinforced those made by the development team,7 who found strong interrater reliability between their automated measures and manual segmentations of midsagittal tissue bridges. The external validation of this measure provides stronger support for its use in future research. Since evidence has shown that the mere presence or absence of tissue bridges may be clinically meaningful,10 we explored both the agreement on the absence or presence of tissue bridges as a binary variable and their quantitative measure (i.e., width). Furthermore, measuring reliability using only the continuous measurement of tissue bridge widths may miss this meaningful distinction. While agreement between the raters for presence or absence of tissue bridges was substantial, there was greater divergence of agreement at smaller widths (around 0.5 mm and below). This could be due to our image resolution. Previous literature suggests that meaningful functional recovery occurs as tissue bridges meet or exceed 1.5 mm.12,13 Thus, discrepancies at these lower values may not be as much of an issue clinically. Further, previous literature has compared manual segmentation between raters and shown excellent interrater reliability.13
It should be noted that in several instances there was a discrepancy between the manual rater and the automated measure with regard to which slice constituted the midsagittal slice. In some instances, the algorithm demonstrated clear errors in identifying the midsagittal slice, showing an image with minimal spinal cord present. Thus in some of these instances, it incorrectly showed that no lesion existed in the slice. To provide reliable results, we decided to standardize the determination of midsagittal slices by manual raters for both approaches (see methods above). Because the SCIsegV2 tool also outputs data for parasagittal slices, we were able to use that data to compare to the manual rater. It is therefore important that users of this tool perform quality control measures to ensure accurate data collection. Additionally, in one instance, segmentation of the spinal cord was incomplete. This appeared to be due to an extensive lesion and thus no intact cord to measure. As a result, the algorithm was unable to identify lesions. It was assumed that midsagittal tissue bridges were not present, as no visible spinal cord was available. In these instances, the readout of the automated measure may not provide appropriate lesion measurements. We do not consider this to be a significant issue, as images such as these clearly indicate severe injury in which tissue bridges were absent.
While reliability between automated and manual raters was found to be excellent, it is less clear whether the difference between them is clinically significant. There appeared to be a larger difference in measurements of ventral tissue bridges (mean -0.13 mm, SD 0.44) than the differences in measures of dorsal tissue bridges (mean -.02 mm, SD 0.13). A multicenter retrospective cohort study found that for every 1 mm of total tissue bridge width, there was an average of 5.7% to 12.6% of maximal sensorimotor recovery over a 12-month period.13 Our study found a mean difference in total tissue bridge width of -0.16 mm (SD 0.48). Based on this, it would seem that, on average, the observed differences between automated measurements and manual measurements were less than clinically meaningful measurements. Further investigations into tissue bridges characteristics are indeed warranted to further our understanding of the clinical significance of these measures.
The SCIsegV2 feature is a new tool available in the SCT, version 6.4, and will likely continue to be updated. This will presumably improve the overall accuracy of the tool over time. It will be appropriate to continue to validate these measures in the future to ensure ongoing accuracy of this tool. The automated measurement of SCI-related biomarkers offers exciting prospective opportunities with regards to research and clinical care. Quantitative MRI characteristics such as midsagittal tissue bridges provide promising steps towards more accurate and less biased prognostication of neurological recovery in SCI.
Limitations
This study sought to explore the performance of an automated measurement of midsagittal tissue bridges. The SCIsegV2 function implemented in SCT provides additional measurements including lesion volume, axial damage ratios, and lesion lengths. Further investigation and comparison of these additional measurements is warranted.
Several subjects were excluded from this study which contained significant artifacts, and therefore lesions were difficult to appreciate. This seemed to be more common for images taken in the thoracic region. Thus, overall sample size was limited to images that were clear enough to analyze and were more skewed towards images in the cervical region of the spinal cord (cervical images = 19, thoracic images = 2).
Another limitation is only using one manual rater as a comparison to the automated tool. While adding more raters could have potentially improved interrater reliability as found in previous work,17 it may also have led to decreased disagreement.
Using this available clinical sagittal T2-weighted dataset, another limitation is the 4-mm space between slices, which may have influenced the selection of the midsagittal slice. We tried to mitigate this by selecting the sagittal slice with the most observable spinal cord, but we certainly recognize this limitation.
Another limitation is the inability to assign greater clinical value to either manual or automated measurements. Automated measurements were compared to manual measures because, up until now, this was the method available. We do not, however, know which measure more accurately predicts clinical outcomes. Further research will indeed be warranted.
Conclusion
This study found that SCIsegV2 is a reliable measurement tool for midsagittal tissue bridges, showing excellent interrater reliability with manual raters on an external dataset. We do recommend that appropriate steps be taken while using the tool. Discrepancies with regard to identification of the midsagittal slice, as well as limited applicability to images with significant artifact, should be considered and require thorough quality control.
Funding Statement
A.C.S. was supported by Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health – K01HD106928 and the Boettcher Foundation's Webb-Waring Biomedical Research Program. K.A.W. and A.C.S. were supported by NIH National Institute of Neurological Disorders and Stroke of the National Institutes of Health R01NS128478. W.A.T. was supported by the Foundation for Physical Therapy Research Promotion of Doctoral Studies program.
Footnotes
Conflicts of Interest
The authors declare no conflicts of interest.
Financial Support
A.C.S. was supported by Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health – K01HD106928 and the Boettcher Foundation's Webb-Waring Biomedical Research Program. K.A.W. and A.C.S. were supported by NIH National Institute of Neurological Disorders and Stroke of the National Institutes of Health R01NS128478. W.A.T. was supported by the Foundation for Physical Therapy Research Promotion of Doctoral Studies program.
REFERENCES
- 1.Cripps RA, Lee BB, Wing P, Weerts E, Mackay J, Brown D. A global map for traumatic spinal cord injury epidemiology: Towards a living data repository for injury prevention. Spinal Cord. 2011;49(4):493–501. doi: 10.1038/sc.2010.146. doi: [DOI] [PubMed] [Google Scholar]
- 2.Kirshblum SC, Burns SP, Biering-Sorensen F, et al. International standards for neurological classification of spinal cord injury (revised 2011) J Spinal Cord Med. 2011;34(6):535–546. doi: 10.1179/204577211X13207446293695. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Gündoğdu İ, Akyüz M, Öztürk EA, Çakcı FA. Can spinal cord injury patients show a worsening in ASIA impairment scale classification despite actually having neurological improvement? The limitation of ASIA Impairment Scale Classification. Spinal Cord. 2014;52(9):667–670. doi: 10.1038/sc.2014.89. doi: [DOI] [PubMed] [Google Scholar]
- 4.Vallotton K, Huber E, Sutter R, Curt A, Hupp M, Freund P. Width and neurophysiologic properties of tissue bridges predict recovery after cervical injury. Neurology. 2019;92(24):e2793. doi: 10.1212/WNL.0000000000007642. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Pfyffer D, Huber E, Sutter R, Curt A, Freund P. Tissue bridges predict recovery after traumatic and ischemic thoracic spinal cord injury. Neurology. 2019;93(16):e1550. doi: 10.1212/WNL.0000000000008318. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Smith AC, Kenneth A Weber, II, Parrish TB, et al. Ambulatory function in motor incomplete spinal cord injury: A magnetic resonance imaging study of spinal cord edema and lower extremity muscle morphometry. Spinal Cord. 2017;55(7):672. doi: 10.1038/sc.2017.18. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Karthik EN, Valošek J, Farner L, et al. SCIsegV2: A Universal Tool for Segmentation of Intramedullary Lesions in Spinal Cord Injury. [Accessed October 8, 2024];2024 July 24; https://arxiv.org/abs/2407.17265v1 arXiv.org. [Google Scholar]
- 8.Naga Karthik E, Valošek J, Smith AC, et al. SCIseg: Automatic Segmentation of intramedullary lesions in spinal cord injury on T2-weighted MRI scans. Radiol Artif Intell. 2024 November 6;:e240005. doi: 10.1148/ryai.240005. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.O’Dell DR, Weber KA, Berliner JC, et al. Midsagittal tissue bridges are associated with walking ability in incomplete spinal cord injury: A magnetic resonance imaging case series. J Spinal Cord Med. 2020;43(2):268–271. doi: 10.1080/10790268.2018.1527079. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Thornton WA, Marzloff G, Ryder S, et al. The presence or absence of midsagittal tissue bridges and walking: A retrospective cohort study in spinal cord injury. Spinal Cord. 2023;61(8):436. doi: 10.1038/s41393-023-00890-6. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pfyffer D, Vallotton K, Curt A, Freund P. Tissue bridges predict neuropathic pain emergence after spinal cord injury. J Neurol Neurosurg Psychiatry. 2020;91(10):1111–1117. doi: 10.1136/jnnp-2020-323150. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pfyffer D, Vallotton K, Curt A, Freund P. Predictive value of midsagittal tissue bridges on functional recovery after spinal cord injury. Neurorehabil Neural Repair. 2021;35(1):33–43. doi: 10.1177/1545968320971787. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pfyffer D, Smith AC, Weber KA, et al. Prognostic value of tissue bridges in cervical spinal cord injury: A longitudinal, multicentre, retrospective cohort study. Lancet Neurol. 2024;23(8):816–825. doi: 10.1016/S1474-4422(24)00173-X. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Emmenegger TM, Pfyffer D, Curt A, et al. Longitudinal motor system changes from acute to chronic spinal cord injury. Eur J Neurol. 2024;31(4):e16196. doi: 10.1111/ene.16196. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Smith AC, O’Dell DR, Albin SR, et al. Lateral corticospinal tract and dorsal column damage: Predictive relationships with motor and sensory scores at discharge from acute rehabilitation after spinal cord injury. Arch Phys Med Rehabil. 2022;103(1):62–68. doi: 10.1016/j.apmr.2021.07.792. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Berliner JC, O’Dell DR, Albin SR, et al. The influence of conventional T2 MRI indices in predicting who will walk outside one year after spinal cord injury. J Spinal Cord Med. 2021;46(3):501. doi: 10.1080/10790268.2021.1907676. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Cummins DP, Connor JR, Heller KA, et al. Establishing the interrater reliability of spinal cord damage manual measurement using magnetic resonance imaging. Spinal Cord Ser Cases. 2019;5:20. doi: 10.1038/s41394-019-0164-1. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Smith AC, O’Dell DR, Thornton WA, et al. Spinal Cord Tissue Bridges Validation Study: Predictive relationships with sensory scores following cervical spinal cord injury. Top Spinal Cord Inj Rehabil. 2021;28(2):111. doi: 10.46292/sci21-00018. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.De Leener B, Lévy S, Dupont SM, et al. SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. NeuroImage. 2017;145(Pt A):24–43. doi: 10.1016/j.neuroimage.2016.10.009. doi: [DOI] [PubMed] [Google Scholar]
- 20.Angeli CA, Edgerton VR, Gerasimenko YP, Harkema SJ. Altering spinal cord excitability enables voluntary movements after chronic complete paralysis in humans. Brain. 2014;137(5):1394. doi: 10.1093/brain/awu038. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Harkema S, Gerasimenko Y, Hodes J, et al. Effect of epidural stimulation of the lumbosacral spinal cord on voluntary movement, standing, and assisted stepping after motor complete paraplegia: A case study. Lancet. 2011;377(9781):1938. doi: 10.1016/S0140-6736(11)60547-3. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Rejc E, Angeli CA, Bryant N, Harkema SJ. Effects of stand and step training with epidural stimulation on motor function for standing in chronic complete paraplegics. J Neurotrauma. 2017;34(9):1787. doi: 10.1089/neu.2016.4516. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rejc E, Smith AC, Kenneth A Weber, II, et al. Spinal cord imaging markers and recovery of volitional leg movement with spinal cord epidural stimulation in individuals with clinically motor complete spinal cord injury. Front Syst Neurosci. 2020;14:559313. doi: 10.3389/fnsys.2020.559313. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]