Contributions
Abstract: P1007
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
MRI-based brain volume measurements may be affected by MS lesions which can distort the segmentation process of specific tissue classes. Lesion filling (LF) has been applied to improve automatic segmentation-based brain volume estimation. However, in clinical practice it is time-consuming and technically demanding. We therefore examined to what extent different amounts of white matter (WM) changes affect such volume estimations and thus distort group comparisons between healthy controls (HC) and MS patients or different MS subgroups in a large patient sample.
180 patients with early RR-MS (mean age= 35.9, SD=9.7; 63% female; EDSS median= 1) and 65 healthy controls (HC; mean age=31.0, SD=8.9; 51% female) underwent brain MRI at 3T. Hyperintense T2-FLAIR lesion load (T2-LL) was assessed by a semi-automated region growing algorithm subsequent to lesion identification by an experienced rater and normalized by intracranial volume. T2 lesion maps were then registered to 3D-T1-weighted sequences. Lesion filling according to T2 lesion maps was performed using the respective FSL tool. A segmentation-based algorithm (SIENAX) served to obtain brain volume measurements with and without LF using the T1 sequences. We stratified patients by the extent of WM changes into four groups defined by T2-LL quartiles.
T2-LL ranged from 1.45 cm³ to 127.61 cm³. Normalized brain volume, total grey matter (GM), and cortical GM volume were significantly lower in the total MS groups compared to HC with and without LF. Within-group comparisons further demonstrated that brain, GM and cortical GM volumes were significantly overestimated without LF, and these within-group differences were mainly driven by patients with high T2-LL (3rd / 4th quartiles). Within the 4th quartile, mean GM overestimation without LF reached 6 cm³(0.8%). LF did not change WM volume (WM underestimation without vs with LF 0.65 cm³ across all MS groups) and cerebrospinal fluid volume estimation to a significant extent.
For large patient groups brain volume estimation with SIENAX showed robust differences between MS patients and HC, even without LF. The misclassification of brain and GM volumes varied considerably with T2-LL, with higher deviation with larger T2-LL. While these results are not unexpected, they emphasise the need for LF, in particular when smaller cohorts of MS patients or individuals with high T2-LL are studied. However, LF for group comparisons in large clinical cohorts seems negligible.
Disclosure: Daniela Pinter declares no conflict of interest.
Michael Khalil declares no conflict of interest.
Paul Greiner declares no conflict of interest.
Daniel Moser declares no conflict of interest.
Eva Pirker declares no conflict of interest.
Alexander Pichler declares no conflict of interest.
Gerhard Bachmaier declares no conflict of interest.
Stefan Ropele declares no conflict of interest.
Siegrid Fuchs declares no conflict of interest.
Franz Fazekas declares no conflict of interest.
Christian Enzinger declares no conflict of interest.
Abstract: P1007
Type: Poster
Abstract Category: Pathology and pathogenesis of MS - Imaging
MRI-based brain volume measurements may be affected by MS lesions which can distort the segmentation process of specific tissue classes. Lesion filling (LF) has been applied to improve automatic segmentation-based brain volume estimation. However, in clinical practice it is time-consuming and technically demanding. We therefore examined to what extent different amounts of white matter (WM) changes affect such volume estimations and thus distort group comparisons between healthy controls (HC) and MS patients or different MS subgroups in a large patient sample.
180 patients with early RR-MS (mean age= 35.9, SD=9.7; 63% female; EDSS median= 1) and 65 healthy controls (HC; mean age=31.0, SD=8.9; 51% female) underwent brain MRI at 3T. Hyperintense T2-FLAIR lesion load (T2-LL) was assessed by a semi-automated region growing algorithm subsequent to lesion identification by an experienced rater and normalized by intracranial volume. T2 lesion maps were then registered to 3D-T1-weighted sequences. Lesion filling according to T2 lesion maps was performed using the respective FSL tool. A segmentation-based algorithm (SIENAX) served to obtain brain volume measurements with and without LF using the T1 sequences. We stratified patients by the extent of WM changes into four groups defined by T2-LL quartiles.
T2-LL ranged from 1.45 cm³ to 127.61 cm³. Normalized brain volume, total grey matter (GM), and cortical GM volume were significantly lower in the total MS groups compared to HC with and without LF. Within-group comparisons further demonstrated that brain, GM and cortical GM volumes were significantly overestimated without LF, and these within-group differences were mainly driven by patients with high T2-LL (3rd / 4th quartiles). Within the 4th quartile, mean GM overestimation without LF reached 6 cm³(0.8%). LF did not change WM volume (WM underestimation without vs with LF 0.65 cm³ across all MS groups) and cerebrospinal fluid volume estimation to a significant extent.
For large patient groups brain volume estimation with SIENAX showed robust differences between MS patients and HC, even without LF. The misclassification of brain and GM volumes varied considerably with T2-LL, with higher deviation with larger T2-LL. While these results are not unexpected, they emphasise the need for LF, in particular when smaller cohorts of MS patients or individuals with high T2-LL are studied. However, LF for group comparisons in large clinical cohorts seems negligible.
Disclosure: Daniela Pinter declares no conflict of interest.
Michael Khalil declares no conflict of interest.
Paul Greiner declares no conflict of interest.
Daniel Moser declares no conflict of interest.
Eva Pirker declares no conflict of interest.
Alexander Pichler declares no conflict of interest.
Gerhard Bachmaier declares no conflict of interest.
Stefan Ropele declares no conflict of interest.
Siegrid Fuchs declares no conflict of interest.
Franz Fazekas declares no conflict of interest.
Christian Enzinger declares no conflict of interest.