Contributions
Abstract: EP1531
Type: ePoster
Abstract Category: Pathology and pathogenesis of MS - 21 Imaging
Background: Atrophy of the thalamus in multiple sclerosis (MS) is associated with cognitive malfunctioning and disability progression. An accurate and reliable volumetric analysis of the thalamus is hampered by low contrast between the thalamus and surrounding tissue on T1 MR images.
Methods: We analysed 3D T1 images from two independent datasets. Dataset 1 contained 30 MS patients, scanned on a 3 Tesla Philips scanner. Dataset 2 was publicly available (Biberacher et al. 2016) and contained a single MS patient, scanned 5-6 times on 3 different MR scanners (GE, Philips, and Siemens) within a few weeks. Thalamus volumes were determined manually using the annotation tool itk-SNAP (www.itksnap.org), along with anatomical atlases and under the supervision of an expert neurologist. For automated thalamus segmentation, we used a previously published (Opfer et al. 2016) atlas based volumetry approach built on SPM12 (Statistical Parametric Mapping), and volumetry with FSL FIRST (FMRIB Software Library).
Results: In dataset 1 (mean age 33 years, 70% females) the mean thalamus volume was 11.5 ml, 14.3 ml, and 9.9 ml for SPM, FSL, and manual segmentation, respectively. Mean volumetric off-set was 1.6 ml between manual and SPM derived volumes, and 4.5 ml between manual and FSL. Manually derived volumes correlated slightly higher with SPM volumes (r=0.77) than with FSL volumes (r=0.71, difference not significant).
For all repeated scans of the 29-year-old female in dataset 2, mean thalamus volumes were 9.2 ml (SPM), 11.2 ml (FSL) and 6.5 ml (manual). Coefficient of variance was highest for manual segmentations (8.1%, 3.2% and 3.3% for GE, Philips and Siemens), and similar for SPM (0.98%, 0.81% and 1.0% for GE, Philips and Siemens) and FSL volumes (1.3%, 0.84% and 1.2% for GE, Philips and Siemens). The mean off-set between manual and SPM derived volumes was 3.4 ml, 2.6 ml and 2.2 ml for GE, Philips, and Siemens data, respectively. The off-set between manual and FSL derived volumes was 5.2 ml, 4.5 ml, and 4.3 ml (for GE, Philips, and Siemens).
Conclusion: In our study, SPM and FSL thalamus volumes were larger than manually derived volumes. For FSL, this has been shown previously in healthy controls. The off-set was similar in dataset 1 and 2, and largely independent of the scanner. SPM derived volumes were closer to manual volumes than volumes assessed by FSL. Manual segmentations had a higher variability in the test-retest data than automatically derived volumes.
Disclosure:
AC. Ostwaldt, R. Opfer, L. Spies and C. Egger report no conflicts of interest.
S. Schippling reports receiving research grants from Novartis and Sanofi Genzyme; and consulting and speaking fees from Biogen, Merck Serono, Novartis, Sanofi Genzyme, and Teva.
P. Manogaran has received travel support from Sanofi Genzyme.
Abstract: EP1531
Type: ePoster
Abstract Category: Pathology and pathogenesis of MS - 21 Imaging
Background: Atrophy of the thalamus in multiple sclerosis (MS) is associated with cognitive malfunctioning and disability progression. An accurate and reliable volumetric analysis of the thalamus is hampered by low contrast between the thalamus and surrounding tissue on T1 MR images.
Methods: We analysed 3D T1 images from two independent datasets. Dataset 1 contained 30 MS patients, scanned on a 3 Tesla Philips scanner. Dataset 2 was publicly available (Biberacher et al. 2016) and contained a single MS patient, scanned 5-6 times on 3 different MR scanners (GE, Philips, and Siemens) within a few weeks. Thalamus volumes were determined manually using the annotation tool itk-SNAP (www.itksnap.org), along with anatomical atlases and under the supervision of an expert neurologist. For automated thalamus segmentation, we used a previously published (Opfer et al. 2016) atlas based volumetry approach built on SPM12 (Statistical Parametric Mapping), and volumetry with FSL FIRST (FMRIB Software Library).
Results: In dataset 1 (mean age 33 years, 70% females) the mean thalamus volume was 11.5 ml, 14.3 ml, and 9.9 ml for SPM, FSL, and manual segmentation, respectively. Mean volumetric off-set was 1.6 ml between manual and SPM derived volumes, and 4.5 ml between manual and FSL. Manually derived volumes correlated slightly higher with SPM volumes (r=0.77) than with FSL volumes (r=0.71, difference not significant).
For all repeated scans of the 29-year-old female in dataset 2, mean thalamus volumes were 9.2 ml (SPM), 11.2 ml (FSL) and 6.5 ml (manual). Coefficient of variance was highest for manual segmentations (8.1%, 3.2% and 3.3% for GE, Philips and Siemens), and similar for SPM (0.98%, 0.81% and 1.0% for GE, Philips and Siemens) and FSL volumes (1.3%, 0.84% and 1.2% for GE, Philips and Siemens). The mean off-set between manual and SPM derived volumes was 3.4 ml, 2.6 ml and 2.2 ml for GE, Philips, and Siemens data, respectively. The off-set between manual and FSL derived volumes was 5.2 ml, 4.5 ml, and 4.3 ml (for GE, Philips, and Siemens).
Conclusion: In our study, SPM and FSL thalamus volumes were larger than manually derived volumes. For FSL, this has been shown previously in healthy controls. The off-set was similar in dataset 1 and 2, and largely independent of the scanner. SPM derived volumes were closer to manual volumes than volumes assessed by FSL. Manual segmentations had a higher variability in the test-retest data than automatically derived volumes.
Disclosure:
AC. Ostwaldt, R. Opfer, L. Spies and C. Egger report no conflicts of interest.
S. Schippling reports receiving research grants from Novartis and Sanofi Genzyme; and consulting and speaking fees from Biogen, Merck Serono, Novartis, Sanofi Genzyme, and Teva.
P. Manogaran has received travel support from Sanofi Genzyme.