Appendix E-1: Image Post-Processing

Appendix E-1: Image Post-Processing

1

NEUROLOGY/2011/396911R2Campbell_e-text

Appendix e-1: Image Post-processing

CEL number, total volume per month, individual CEL size, MTR and aBH number were computed from the M2-scan and M3-scan. T2 lesion volume (LV) and T2 lesion number (LN), along with chronic BH (cBH) LN and brain parenchyma fraction (BPF) were computed from the M1-scan. While CEL number, total volume per month, individual CEL size, T2-LV, T2-LN, cBH-LN and brain parenchyma fraction (BPF) were computed in all the 88 patients, CEL MTR and likelihood to convert into aBHs were analyzed only in the 26 patients presenting both n-CELs and re-CELs during the study period.

  1. Image acquisition

The MRI protocol for all subjects included the following sequences obtained with 3 mm thick contiguous slices and a 24 cm field of view: (1) dual echo PD-w and T2-w SE image with variable echo time (TE) 20/100 msec, repetition time (TR) 2000 msec, 2 excitations and a 128x256 matrix; (2) FLAIR image with TE 140 msec, TR 10,000 msec, inversion time 2200 msec, 1 excitation, 192x256 matrix; (3) T1-w SE image with a TE 16 msec, TR 600 msec, two excitations, 192x256 matrix, performed before contrast injection; and (4) post-contrast SE T1-w image within 15 minutes after the injection of Gd-DTPA (Magnevist, Berlex Labs, Cedar Knolls, NJ, USA) at 0.1 mmol/kg.

  1. Image Registration

Image analyses were performed using a semi-automated procedure utilizing the scripting language ImageScript of an interactive visualization and analysis software package (i.e., MEDx 3.42for PC LINUX platform). Details on the above procedure have been previously described.e1 A linear image registration algorithm (as developed at the Functional Magnetic Resonance Imaging of the Brain Centre, University of Oxford, UK,e2was used to perform image registrations. The M1-PD-w image was used as a reference image, on top of which the following images were registered: M1-pre-contrast T1-w SE image and M1-FLAIR image, M1-, M2-, and M3-post-contrast MRIs.

  1. CEL Identification and Computation of LN and LV

An experienced physician (NR) began the study analyses by identifying and marking CELs present in M1-scan, M2-scan, and M3-scan. Hard copies (i.e. actual films) were used for this CEL identification. These manually generated masks were used as a reference for outlining CELs on the electronic image copies and computing their number, size and total volume.

Two observers (DS and KD) traced the lesions on the M2- and M3-post-contrast images. Once generated, these CEL templates were overlaid above the M1-pre- and M1-post-contrast T1-w SE image, M1-PD-w image and, M1-FLAIR image. These CELs were classified into re-CELs and n-CELs based on their relation with pre-existing lesions in PD-w and FLAIR images. LL and LV of n-CELs and re-CELs were finally computed using Medex as previously described.e1The threshold for CEL segmentation was fixed in each individual case. For the assessment of ring CELs, central, non-enhancing portions within the enhancing rims were not included in the LV measures.

  1. aBH LL Computation

Areas of hypointensity on T1-w SE image obtained before the contrast injection and corresponding to a CEL have been classically referred to as aBHs.e3 For the identification of aBHs present at M2- and M3-MRI for each patient, n-CELs and re-CELs were overlaid on the M1- and M2-T1-w pre-contrast MRI image. A limited number of lesions (data not shown) were found to arise above a preexisting cBH. These lesions were not classified as aBHs in the present analysis. One investigator (CP) identified the presence of aBHs, which was subsequently inspected by a second experienced observer (FB).

  1. T2-LL and LV Computation

An observer (MD), blinded to the clinical features of the patients, marked the hyperintense lesions on the M1-FLAIR images according to the strategy previously described.e4 The generated lesion masks were then inspected from a second experienced investigator (FB) and LV computed using Medex as previously described.e5

  1. cBH LL Computation

In the M1-scan cBHs were defined as areas of hypointensity on T1-w SE image obtained before the contrast injection corresponding to hyperintense areas in the FLAIR images and in the absence of enhancement in the post-contrast T1-w SE image. cBHs were counted by one investigator (JJ) and inspected by a second one (FB).

  1. BPF Computation

BPF computation was done based on Structural Image Evaluation, including Normalization, of Atrophy (X-sectional) (Sienax).e2 In order to account for misclassification of lesions, intermediate segmentation masks of grey matter, WM, and cerebrospinal fluid were saved and manually edited. Subsequently, the slice containing the velum interpositumwas identified in all volumes; BPF calculation was restricted to seven slices above and including the slice containing the velum interpositum as previously reported.e6

  1. Computation of CEL MTR

For only patients who presented with both n-CELs and re-CELs, CEL masks from M2-scan and M3-scan were overlaid on the M2- and M3-MTR map image previously registered to the M1-PD-w image. The analysis was performed by one investigator (CP) who computed the MTRs using the methodology previously described. e7,e8 To avoid biases due to partial volume effect in the MTR measurements, regions of interest (ROIs) containing ≤10 voxels were not included in the MTR computations. e8,e9

  1. Reliability Measures

Prior to the analyses, five patients were randomly selected, and three investigators (KD, DS, and FB) calculated CEL LL and LV. CEL volumes for these patients were then compared among all three observers to yield an inter-observer variability of 4.4%±2.2 (1.1-6.8%). KD and DS then repeated the calculations at a separate time, each of which resulted in an intra-observer coefficient of variation (COV) of <5% [KD=3.8%±3.6 (0-8.5), DS=1.6%±1.0 (0-2.4)].

The COV was given by the standard deviation (SD) divided by the mean of the observations’ values.

e-References

e1.Gupta S, Solomon JM, Tasciyan TA, et al. Interferon-b-1b effects on re-enhancing lesions in patients with multiple sclerosis. MultScler 2005; 11: 658-668.

e2.Smith SM, Zhang Y, Jenkinson M, Chen J, Matthews PM, Federico A et al. Accurate, robust and automated longitudinal and cross-sectional brain change analysis. NeuroImage 2002; 17: 479-489.

e3.Bagnato F, Jeffries N, Richert ND, et al. Evolution of T1 black holes in patients with multiple sclerosis imaged monthly for 4 years. Brain 2003; 126: 1782-1789.

e4.Filippi M, Gawne-Cain ML, Gasperini C et al. Effect of training and different measurement strategies on the reproducibility of brain MRI lesion load measurements in multiple sclerosis. Neurology 1998; 50: 238-244.

e5.Bagnato F, Butman JA, Mora CA et al. Conventional magnetic resonance imaging features in patients with tropical spastic paraparesis. J Neurovirol 2005; 11: 525-534.

e6.Stevenson VL, Smith SM, Matthews PM, Miller DH, Thompson AJ. Monitoring disease activity and progression in primary progressive multiple sclerosis using MRI: sub-voxel registration to identify lesion changes and to detect cerebral atrophy. J Neurol 2002; 249: 171-177.

e7.Richert ND, Ostuni JL, Bash CN, Duyn JH, McFarland HF, Frank JA. Serial whole-brain magnetization transfer imaging in patients with relapsing-remitting multiple sclerosis at baseline and during treatment with interferon beta-1b. Am J Neuroradiol 1998; 19:1705-1713.

e8.Riva M, Ikonomidou VN, Ostuni JJ, et al. Tissue-specific imaging is a robust methodology to differentiate in vivo T1 black holes with advanced multiple sclerosis-induced damage. Am J Neuroradiol 2009; 30: 1394-1401.

e9.Richert ND, Ostuni JL, Bash CN, Leist TP, McFarland HF, Frank JA. Interferon beta-1b and intravenous methylprednisolone promote lesion recovery in multiple sclerosis. MultScler 2001; 7: 49-59.