Abstract

Synopsis Atrophy of the spinal cord is known to occur in multiple sclerosis (MS). To measure such atrophy, the mean upper cervical cord area (MUCCA) can be assessed. We tested five different (semi-)automated spinal cord segmentation methods (SCT-PropSeg, SCT-DeepSeg, ITK-SNAP, NeuroQLab, Xinapse JIM) in terms of their reproducibility, robustness, and the influence of lesions on the segmentations. MUCCA from all scanners was highly reproducible within-scanner, but not between-scanner or between-methods. The presence of lesions in the upper cervical cord did not affect the accuracy of MUCCA measurements in any of the methods tested. Background Multiple sclerosis (MS) is a demyelinating and neurodegenerative autoimmune disease of the central nervous system (CNS). Abnormalities in the spinal cord such as lesions and atrophy often manifest early in the disease course and have shown to be important indicators for disease progression and prognosis . To study these abnormalities, the mean upper cervical cord area (MUCCA) is measured using magnetic resonance imaging (MRI) . Because manual MUCCA measurements suffer from large intra- and inter-rater variability , several (semi-)automated methods have been developed, but there has been limited work comparing and validating these in clinical MR images. Aims This study aims first to assess the reproducibility and robustness of five (semi-)automated spinal cord segmentation methods and second to investigate whether the presence of lesions in the cervical spinal cord affected the accuracy of these MUCCA measures. Methods Five methods to measure MUCCA (SCT-PropSeg , SCT-DeepSeg , ITK-SNAP , NeuroQLab , and Xinapse JIM ) were investigated. First, within-scanner reproducibility and between-scanner robustness were assessed using intra-class correlation coefficient (ICC) and Dice’s similarity index (SI) in same-day scan-rescan 3DT1-weighted images (head coil, including cervical spine) performed on three 3T MR scanners (GE MR750, Philips Ingenuity, Toshiba Vantage Titan) in 21 relapsing-remitting (RR) MS patients and 6 healthy controls (HC) (dataset A, see Table 1). Second, sensitivity of MUCCA to lesions in the upper cervical cord was assessed with clinical cervical 3DT1-weighted images (3T GE HDxT head-neck-spine coil) in 7 MS patients without and 14 MS patients with cervical lesions (dataset B, see Table 1), using ICC and SI with manual segmentations as reference. Repeated measures ANOVA for MUCCA and for SI were used, and when appropriate, post-hoc Wilcoxon signed ranks tests (related samples) or Mann-Whitney U tests (unrelated samples) were performed. Statistical significance was obtained upon p < 0.05. Results In dataset A, repeated measures ANOVA for MUCCA showed main effects of vendor (p < 0.001) and method (p < 0.001), as well as an interaction between vendor and method (p = 0.025), but no effect between session (scan or rescan) or subject group (RRMS or HC) (Figure 1). Within-scanner reproducibility was highest for Xinapse JIM (ICC absolute agreement = 0.995) and lowest for ITK-SNAP (ICC absolute agreement = 0.825). Between-scanner robustness was highest for Xinapse JIM and SCT-PropSeg (ICC consistency = 0.981 and 0.976, respectively) and lowest for ITK-SNAP (ICC consistency = 0.871). Repeated measures ANOVA for SI (scan vs rescan) showed a main effect of method (p < 0.001), but not of vendor or subject group, with highest SI in Xinapse JIM and SCT-PropSeg (0.918±0.020) and lowest SI in SCT-DeepSeg (0.906±0.021) (Figure 2). In dataset B, repeated measures ANOVA for MUCCA showed a main effect of method (p < 0.001), but not of lesion group (Figure 3). However, repeated measures ANOVA for SI (automated vs manual) showed an effect of method (p < 0.001) and lesion group (p = 0.024), but no interaction between method and lesion group, with highest SI in Xinapse JIM (0.958±0.014) and lowest SI in SCT-PropSeg (0.806±0.031) (Figure 4). In post-hoc tests, no paired differences were found between the many and no lesion groups in any of the methods. Discussion and conclusion MUCCA measurements from all tested methods were highly reproducible within-scanner, but not between-scanner or between-methods. The presence of cervical lesions did not significantly affect MUCCA measurement accuracy. Results should be confirmed in multi-center cervical images at different vertebral levels. Acknowledgements This work was supported by the Dutch MS Research Foundation (grant numbers 14-876 and 09-358d). References 1. Kearney H, Miller DH, Ciccarelli O. Spinal cord MRI in multiple sclerosis--diagnostic, prognostic and clinical value. Nat Rev Neurol. 2015;11(6):327-38. 2. Liu Z, Yaldizli O, Pardini M, Sethi V, Kearney H, Muhlert N, et al. Cervical cord area measurement using volumetric brain magnetic resonance imaging in multiple sclerosis. Mult Scler Relat Disord. 2015;4(1):52-7. 1 1 2 1 1 1 3 3 3 4 4 1,5 1 1 1 2 3 4 5 1 2 3-6 6 7 8 9 10 3. Cadotte A, Cadotte DW, Livne M, Cohen-Adad J, Fleet D, Mikulis D, et al. Spinal Cord Segmentation by One Dimensional Normalized Template Matching: A Novel, Quantitative Technique to Analyze Advanced Magnetic Resonance Imaging Data. PLoS One. 2015;10(10):e0139323. 4. El Mendili MM, Chen R, Tiret B, Pelegrini-Issac M, Cohen-Adad J, Lehericy S, et al. Validation of a semiautomated spinal cord segmentation method. J Magn Reson Imaging. 2015;41(2):454-9. 5. Kearney H, Yiannakas MC, Abdel-Aziz K, Wheeler-Kingshott CA, Altmann DR, Ciccarelli O, et al. Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging. 2014;39(3):617-23. 6. De Leener B, Kadoury S, Cohen-Adad J. Robust, accurate and fast automatic segmentation of the spinal cord. Neuroimage. 2014;98:528-36. 7. De Leener B, Levy S, Dupont SM, Fonov VS, Stikov N, Louis Collins D, et al. SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. Neuroimage. 2017;145(Pt A):24-43. 8. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116-28. 9. Lukas C, Hahn HK, Bellenberg B, Rexilius J, Schmid G, Schimrigk SK, et al. Sensitivity and reproducibility of a new fast 3D segmentation technique for clinical MR-based brain volumetry in multiple sclerosis. Neuroradiology. 2004;46(11):906-15.
Original languageEnglish
Publication statusSubmitted - 6 Nov 2018
EventISMRM: Annual Meeting 2019 - Montréal, Canada
Duration: 11 May 201916 May 2019
https://www.ismrm.org/19m/

Conference

ConferenceISMRM
CountryCanada
CityMontréal
Period11/05/201916/05/2019
Internet address

Cite this

@conference{1df1d4d0b8e348abb1eb87ca4885caa7,
title = "High reproducibility and robustness to lesions, but large software and scanner effects for mean upper cervical cord area (MUCCA) measurement in MS",
abstract = "Synopsis Atrophy of the spinal cord is known to occur in multiple sclerosis (MS). To measure such atrophy, the mean upper cervical cord area (MUCCA) can be assessed. We tested five different (semi-)automated spinal cord segmentation methods (SCT-PropSeg, SCT-DeepSeg, ITK-SNAP, NeuroQLab, Xinapse JIM) in terms of their reproducibility, robustness, and the influence of lesions on the segmentations. MUCCA from all scanners was highly reproducible within-scanner, but not between-scanner or between-methods. The presence of lesions in the upper cervical cord did not affect the accuracy of MUCCA measurements in any of the methods tested. Background Multiple sclerosis (MS) is a demyelinating and neurodegenerative autoimmune disease of the central nervous system (CNS). Abnormalities in the spinal cord such as lesions and atrophy often manifest early in the disease course and have shown to be important indicators for disease progression and prognosis . To study these abnormalities, the mean upper cervical cord area (MUCCA) is measured using magnetic resonance imaging (MRI) . Because manual MUCCA measurements suffer from large intra- and inter-rater variability , several (semi-)automated methods have been developed, but there has been limited work comparing and validating these in clinical MR images. Aims This study aims first to assess the reproducibility and robustness of five (semi-)automated spinal cord segmentation methods and second to investigate whether the presence of lesions in the cervical spinal cord affected the accuracy of these MUCCA measures. Methods Five methods to measure MUCCA (SCT-PropSeg , SCT-DeepSeg , ITK-SNAP , NeuroQLab , and Xinapse JIM ) were investigated. First, within-scanner reproducibility and between-scanner robustness were assessed using intra-class correlation coefficient (ICC) and Dice’s similarity index (SI) in same-day scan-rescan 3DT1-weighted images (head coil, including cervical spine) performed on three 3T MR scanners (GE MR750, Philips Ingenuity, Toshiba Vantage Titan) in 21 relapsing-remitting (RR) MS patients and 6 healthy controls (HC) (dataset A, see Table 1). Second, sensitivity of MUCCA to lesions in the upper cervical cord was assessed with clinical cervical 3DT1-weighted images (3T GE HDxT head-neck-spine coil) in 7 MS patients without and 14 MS patients with cervical lesions (dataset B, see Table 1), using ICC and SI with manual segmentations as reference. Repeated measures ANOVA for MUCCA and for SI were used, and when appropriate, post-hoc Wilcoxon signed ranks tests (related samples) or Mann-Whitney U tests (unrelated samples) were performed. Statistical significance was obtained upon p < 0.05. Results In dataset A, repeated measures ANOVA for MUCCA showed main effects of vendor (p < 0.001) and method (p < 0.001), as well as an interaction between vendor and method (p = 0.025), but no effect between session (scan or rescan) or subject group (RRMS or HC) (Figure 1). Within-scanner reproducibility was highest for Xinapse JIM (ICC absolute agreement = 0.995) and lowest for ITK-SNAP (ICC absolute agreement = 0.825). Between-scanner robustness was highest for Xinapse JIM and SCT-PropSeg (ICC consistency = 0.981 and 0.976, respectively) and lowest for ITK-SNAP (ICC consistency = 0.871). Repeated measures ANOVA for SI (scan vs rescan) showed a main effect of method (p < 0.001), but not of vendor or subject group, with highest SI in Xinapse JIM and SCT-PropSeg (0.918±0.020) and lowest SI in SCT-DeepSeg (0.906±0.021) (Figure 2). In dataset B, repeated measures ANOVA for MUCCA showed a main effect of method (p < 0.001), but not of lesion group (Figure 3). However, repeated measures ANOVA for SI (automated vs manual) showed an effect of method (p < 0.001) and lesion group (p = 0.024), but no interaction between method and lesion group, with highest SI in Xinapse JIM (0.958±0.014) and lowest SI in SCT-PropSeg (0.806±0.031) (Figure 4). In post-hoc tests, no paired differences were found between the many and no lesion groups in any of the methods. Discussion and conclusion MUCCA measurements from all tested methods were highly reproducible within-scanner, but not between-scanner or between-methods. The presence of cervical lesions did not significantly affect MUCCA measurement accuracy. Results should be confirmed in multi-center cervical images at different vertebral levels. Acknowledgements This work was supported by the Dutch MS Research Foundation (grant numbers 14-876 and 09-358d). References 1. Kearney H, Miller DH, Ciccarelli O. Spinal cord MRI in multiple sclerosis--diagnostic, prognostic and clinical value. Nat Rev Neurol. 2015;11(6):327-38. 2. Liu Z, Yaldizli O, Pardini M, Sethi V, Kearney H, Muhlert N, et al. Cervical cord area measurement using volumetric brain magnetic resonance imaging in multiple sclerosis. Mult Scler Relat Disord. 2015;4(1):52-7. 1 1 2 1 1 1 3 3 3 4 4 1,5 1 1 1 2 3 4 5 1 2 3-6 6 7 8 9 10 3. Cadotte A, Cadotte DW, Livne M, Cohen-Adad J, Fleet D, Mikulis D, et al. Spinal Cord Segmentation by One Dimensional Normalized Template Matching: A Novel, Quantitative Technique to Analyze Advanced Magnetic Resonance Imaging Data. PLoS One. 2015;10(10):e0139323. 4. El Mendili MM, Chen R, Tiret B, Pelegrini-Issac M, Cohen-Adad J, Lehericy S, et al. Validation of a semiautomated spinal cord segmentation method. J Magn Reson Imaging. 2015;41(2):454-9. 5. Kearney H, Yiannakas MC, Abdel-Aziz K, Wheeler-Kingshott CA, Altmann DR, Ciccarelli O, et al. Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging. 2014;39(3):617-23. 6. De Leener B, Kadoury S, Cohen-Adad J. Robust, accurate and fast automatic segmentation of the spinal cord. Neuroimage. 2014;98:528-36. 7. De Leener B, Levy S, Dupont SM, Fonov VS, Stikov N, Louis Collins D, et al. SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. Neuroimage. 2017;145(Pt A):24-43. 8. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116-28. 9. Lukas C, Hahn HK, Bellenberg B, Rexilius J, Schmid G, Schimrigk SK, et al. Sensitivity and reproducibility of a new fast 3D segmentation technique for clinical MR-based brain volumetry in multiple sclerosis. Neuroradiology. 2004;46(11):906-15.",
author = "MM Weeda and SM Middelkoop and MD Steenwijk and M Daams and {Amiri Doumari}, H and I Brouwer and J Killestein and BMJ Uitdehaag and I Dekker and C. Lukas and B. Bellenberg and F Barkhof and PJW Pouwels and H Vrenken",
year = "2018",
month = "11",
day = "6",
language = "English",
note = "ISMRM : Annual Meeting 2019 ; Conference date: 11-05-2019 Through 16-05-2019",
url = "https://www.ismrm.org/19m/",

}

High reproducibility and robustness to lesions, but large software and scanner effects for mean upper cervical cord area (MUCCA) measurement in MS. / Weeda, MM; Middelkoop, SM; Steenwijk, MD; Daams, M; Amiri Doumari, H; Brouwer, I; Killestein, J; Uitdehaag, BMJ; Dekker, I; Lukas, C.; Bellenberg, B.; Barkhof, F; Pouwels, PJW; Vrenken, H.

2018. Abstract from ISMRM, Montréal, Canada.

Research output: Contribution to conferenceAbstractAcademic

TY - CONF

T1 - High reproducibility and robustness to lesions, but large software and scanner effects for mean upper cervical cord area (MUCCA) measurement in MS

AU - Weeda, MM

AU - Middelkoop, SM

AU - Steenwijk, MD

AU - Daams, M

AU - Amiri Doumari, H

AU - Brouwer, I

AU - Killestein, J

AU - Uitdehaag, BMJ

AU - Dekker, I

AU - Lukas, C.

AU - Bellenberg, B.

AU - Barkhof, F

AU - Pouwels, PJW

AU - Vrenken, H

PY - 2018/11/6

Y1 - 2018/11/6

N2 - Synopsis Atrophy of the spinal cord is known to occur in multiple sclerosis (MS). To measure such atrophy, the mean upper cervical cord area (MUCCA) can be assessed. We tested five different (semi-)automated spinal cord segmentation methods (SCT-PropSeg, SCT-DeepSeg, ITK-SNAP, NeuroQLab, Xinapse JIM) in terms of their reproducibility, robustness, and the influence of lesions on the segmentations. MUCCA from all scanners was highly reproducible within-scanner, but not between-scanner or between-methods. The presence of lesions in the upper cervical cord did not affect the accuracy of MUCCA measurements in any of the methods tested. Background Multiple sclerosis (MS) is a demyelinating and neurodegenerative autoimmune disease of the central nervous system (CNS). Abnormalities in the spinal cord such as lesions and atrophy often manifest early in the disease course and have shown to be important indicators for disease progression and prognosis . To study these abnormalities, the mean upper cervical cord area (MUCCA) is measured using magnetic resonance imaging (MRI) . Because manual MUCCA measurements suffer from large intra- and inter-rater variability , several (semi-)automated methods have been developed, but there has been limited work comparing and validating these in clinical MR images. Aims This study aims first to assess the reproducibility and robustness of five (semi-)automated spinal cord segmentation methods and second to investigate whether the presence of lesions in the cervical spinal cord affected the accuracy of these MUCCA measures. Methods Five methods to measure MUCCA (SCT-PropSeg , SCT-DeepSeg , ITK-SNAP , NeuroQLab , and Xinapse JIM ) were investigated. First, within-scanner reproducibility and between-scanner robustness were assessed using intra-class correlation coefficient (ICC) and Dice’s similarity index (SI) in same-day scan-rescan 3DT1-weighted images (head coil, including cervical spine) performed on three 3T MR scanners (GE MR750, Philips Ingenuity, Toshiba Vantage Titan) in 21 relapsing-remitting (RR) MS patients and 6 healthy controls (HC) (dataset A, see Table 1). Second, sensitivity of MUCCA to lesions in the upper cervical cord was assessed with clinical cervical 3DT1-weighted images (3T GE HDxT head-neck-spine coil) in 7 MS patients without and 14 MS patients with cervical lesions (dataset B, see Table 1), using ICC and SI with manual segmentations as reference. Repeated measures ANOVA for MUCCA and for SI were used, and when appropriate, post-hoc Wilcoxon signed ranks tests (related samples) or Mann-Whitney U tests (unrelated samples) were performed. Statistical significance was obtained upon p < 0.05. Results In dataset A, repeated measures ANOVA for MUCCA showed main effects of vendor (p < 0.001) and method (p < 0.001), as well as an interaction between vendor and method (p = 0.025), but no effect between session (scan or rescan) or subject group (RRMS or HC) (Figure 1). Within-scanner reproducibility was highest for Xinapse JIM (ICC absolute agreement = 0.995) and lowest for ITK-SNAP (ICC absolute agreement = 0.825). Between-scanner robustness was highest for Xinapse JIM and SCT-PropSeg (ICC consistency = 0.981 and 0.976, respectively) and lowest for ITK-SNAP (ICC consistency = 0.871). Repeated measures ANOVA for SI (scan vs rescan) showed a main effect of method (p < 0.001), but not of vendor or subject group, with highest SI in Xinapse JIM and SCT-PropSeg (0.918±0.020) and lowest SI in SCT-DeepSeg (0.906±0.021) (Figure 2). In dataset B, repeated measures ANOVA for MUCCA showed a main effect of method (p < 0.001), but not of lesion group (Figure 3). However, repeated measures ANOVA for SI (automated vs manual) showed an effect of method (p < 0.001) and lesion group (p = 0.024), but no interaction between method and lesion group, with highest SI in Xinapse JIM (0.958±0.014) and lowest SI in SCT-PropSeg (0.806±0.031) (Figure 4). In post-hoc tests, no paired differences were found between the many and no lesion groups in any of the methods. Discussion and conclusion MUCCA measurements from all tested methods were highly reproducible within-scanner, but not between-scanner or between-methods. The presence of cervical lesions did not significantly affect MUCCA measurement accuracy. Results should be confirmed in multi-center cervical images at different vertebral levels. Acknowledgements This work was supported by the Dutch MS Research Foundation (grant numbers 14-876 and 09-358d). References 1. Kearney H, Miller DH, Ciccarelli O. Spinal cord MRI in multiple sclerosis--diagnostic, prognostic and clinical value. Nat Rev Neurol. 2015;11(6):327-38. 2. Liu Z, Yaldizli O, Pardini M, Sethi V, Kearney H, Muhlert N, et al. Cervical cord area measurement using volumetric brain magnetic resonance imaging in multiple sclerosis. Mult Scler Relat Disord. 2015;4(1):52-7. 1 1 2 1 1 1 3 3 3 4 4 1,5 1 1 1 2 3 4 5 1 2 3-6 6 7 8 9 10 3. Cadotte A, Cadotte DW, Livne M, Cohen-Adad J, Fleet D, Mikulis D, et al. Spinal Cord Segmentation by One Dimensional Normalized Template Matching: A Novel, Quantitative Technique to Analyze Advanced Magnetic Resonance Imaging Data. PLoS One. 2015;10(10):e0139323. 4. El Mendili MM, Chen R, Tiret B, Pelegrini-Issac M, Cohen-Adad J, Lehericy S, et al. Validation of a semiautomated spinal cord segmentation method. J Magn Reson Imaging. 2015;41(2):454-9. 5. Kearney H, Yiannakas MC, Abdel-Aziz K, Wheeler-Kingshott CA, Altmann DR, Ciccarelli O, et al. Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging. 2014;39(3):617-23. 6. De Leener B, Kadoury S, Cohen-Adad J. Robust, accurate and fast automatic segmentation of the spinal cord. Neuroimage. 2014;98:528-36. 7. De Leener B, Levy S, Dupont SM, Fonov VS, Stikov N, Louis Collins D, et al. SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. Neuroimage. 2017;145(Pt A):24-43. 8. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116-28. 9. Lukas C, Hahn HK, Bellenberg B, Rexilius J, Schmid G, Schimrigk SK, et al. Sensitivity and reproducibility of a new fast 3D segmentation technique for clinical MR-based brain volumetry in multiple sclerosis. Neuroradiology. 2004;46(11):906-15.

AB - Synopsis Atrophy of the spinal cord is known to occur in multiple sclerosis (MS). To measure such atrophy, the mean upper cervical cord area (MUCCA) can be assessed. We tested five different (semi-)automated spinal cord segmentation methods (SCT-PropSeg, SCT-DeepSeg, ITK-SNAP, NeuroQLab, Xinapse JIM) in terms of their reproducibility, robustness, and the influence of lesions on the segmentations. MUCCA from all scanners was highly reproducible within-scanner, but not between-scanner or between-methods. The presence of lesions in the upper cervical cord did not affect the accuracy of MUCCA measurements in any of the methods tested. Background Multiple sclerosis (MS) is a demyelinating and neurodegenerative autoimmune disease of the central nervous system (CNS). Abnormalities in the spinal cord such as lesions and atrophy often manifest early in the disease course and have shown to be important indicators for disease progression and prognosis . To study these abnormalities, the mean upper cervical cord area (MUCCA) is measured using magnetic resonance imaging (MRI) . Because manual MUCCA measurements suffer from large intra- and inter-rater variability , several (semi-)automated methods have been developed, but there has been limited work comparing and validating these in clinical MR images. Aims This study aims first to assess the reproducibility and robustness of five (semi-)automated spinal cord segmentation methods and second to investigate whether the presence of lesions in the cervical spinal cord affected the accuracy of these MUCCA measures. Methods Five methods to measure MUCCA (SCT-PropSeg , SCT-DeepSeg , ITK-SNAP , NeuroQLab , and Xinapse JIM ) were investigated. First, within-scanner reproducibility and between-scanner robustness were assessed using intra-class correlation coefficient (ICC) and Dice’s similarity index (SI) in same-day scan-rescan 3DT1-weighted images (head coil, including cervical spine) performed on three 3T MR scanners (GE MR750, Philips Ingenuity, Toshiba Vantage Titan) in 21 relapsing-remitting (RR) MS patients and 6 healthy controls (HC) (dataset A, see Table 1). Second, sensitivity of MUCCA to lesions in the upper cervical cord was assessed with clinical cervical 3DT1-weighted images (3T GE HDxT head-neck-spine coil) in 7 MS patients without and 14 MS patients with cervical lesions (dataset B, see Table 1), using ICC and SI with manual segmentations as reference. Repeated measures ANOVA for MUCCA and for SI were used, and when appropriate, post-hoc Wilcoxon signed ranks tests (related samples) or Mann-Whitney U tests (unrelated samples) were performed. Statistical significance was obtained upon p < 0.05. Results In dataset A, repeated measures ANOVA for MUCCA showed main effects of vendor (p < 0.001) and method (p < 0.001), as well as an interaction between vendor and method (p = 0.025), but no effect between session (scan or rescan) or subject group (RRMS or HC) (Figure 1). Within-scanner reproducibility was highest for Xinapse JIM (ICC absolute agreement = 0.995) and lowest for ITK-SNAP (ICC absolute agreement = 0.825). Between-scanner robustness was highest for Xinapse JIM and SCT-PropSeg (ICC consistency = 0.981 and 0.976, respectively) and lowest for ITK-SNAP (ICC consistency = 0.871). Repeated measures ANOVA for SI (scan vs rescan) showed a main effect of method (p < 0.001), but not of vendor or subject group, with highest SI in Xinapse JIM and SCT-PropSeg (0.918±0.020) and lowest SI in SCT-DeepSeg (0.906±0.021) (Figure 2). In dataset B, repeated measures ANOVA for MUCCA showed a main effect of method (p < 0.001), but not of lesion group (Figure 3). However, repeated measures ANOVA for SI (automated vs manual) showed an effect of method (p < 0.001) and lesion group (p = 0.024), but no interaction between method and lesion group, with highest SI in Xinapse JIM (0.958±0.014) and lowest SI in SCT-PropSeg (0.806±0.031) (Figure 4). In post-hoc tests, no paired differences were found between the many and no lesion groups in any of the methods. Discussion and conclusion MUCCA measurements from all tested methods were highly reproducible within-scanner, but not between-scanner or between-methods. The presence of cervical lesions did not significantly affect MUCCA measurement accuracy. Results should be confirmed in multi-center cervical images at different vertebral levels. Acknowledgements This work was supported by the Dutch MS Research Foundation (grant numbers 14-876 and 09-358d). References 1. Kearney H, Miller DH, Ciccarelli O. Spinal cord MRI in multiple sclerosis--diagnostic, prognostic and clinical value. Nat Rev Neurol. 2015;11(6):327-38. 2. Liu Z, Yaldizli O, Pardini M, Sethi V, Kearney H, Muhlert N, et al. Cervical cord area measurement using volumetric brain magnetic resonance imaging in multiple sclerosis. Mult Scler Relat Disord. 2015;4(1):52-7. 1 1 2 1 1 1 3 3 3 4 4 1,5 1 1 1 2 3 4 5 1 2 3-6 6 7 8 9 10 3. Cadotte A, Cadotte DW, Livne M, Cohen-Adad J, Fleet D, Mikulis D, et al. Spinal Cord Segmentation by One Dimensional Normalized Template Matching: A Novel, Quantitative Technique to Analyze Advanced Magnetic Resonance Imaging Data. PLoS One. 2015;10(10):e0139323. 4. El Mendili MM, Chen R, Tiret B, Pelegrini-Issac M, Cohen-Adad J, Lehericy S, et al. Validation of a semiautomated spinal cord segmentation method. J Magn Reson Imaging. 2015;41(2):454-9. 5. Kearney H, Yiannakas MC, Abdel-Aziz K, Wheeler-Kingshott CA, Altmann DR, Ciccarelli O, et al. Improved MRI quantification of spinal cord atrophy in multiple sclerosis. J Magn Reson Imaging. 2014;39(3):617-23. 6. De Leener B, Kadoury S, Cohen-Adad J. Robust, accurate and fast automatic segmentation of the spinal cord. Neuroimage. 2014;98:528-36. 7. De Leener B, Levy S, Dupont SM, Fonov VS, Stikov N, Louis Collins D, et al. SCT: Spinal Cord Toolbox, an open-source software for processing spinal cord MRI data. Neuroimage. 2017;145(Pt A):24-43. 8. Yushkevich PA, Piven J, Hazlett HC, Smith RG, Ho S, Gee JC, et al. User-guided 3D active contour segmentation of anatomical structures: significantly improved efficiency and reliability. Neuroimage. 2006;31(3):1116-28. 9. Lukas C, Hahn HK, Bellenberg B, Rexilius J, Schmid G, Schimrigk SK, et al. Sensitivity and reproducibility of a new fast 3D segmentation technique for clinical MR-based brain volumetry in multiple sclerosis. Neuroradiology. 2004;46(11):906-15.

M3 - Abstract

ER -