Material & Methods – Patients
The study was performed in accord with the Helsinki Declaration of 1975 and approved by the local ethics committee.
Written informed consent was waived.
We conducted a retrospective study at our institution on MRI data achieved between September 1st,
2009 and April 30th,
2015.
Every patient who received a baseline MRI and at least one follow-up study under suspicion or for control of a chronic demyelinating central nervous disease was included.
Material & Methods – MR imaging
All MR examinations were performed at 3T (Magnetom Verio,
Siemens,
Erlangen; Syngo Vers.
VB 17A) using 32-channel head coils for signal detection.
Imaging protocol consisted of at least:
- sagittal 3D T1-weighted MPR dataset with 0.9 mm isotropic spatial resolution (TE 2.58 ms,
TR 1900 ms,
TI 900 ms,
flip angle 9°,
matrix 246 x 256,
FOV 220,
TA 4:18 min)
- axial T2-weighted TSE dataset with 0.6 x 0.5 x 3 mm spatial resolution (TE 103 ms,
TR 6100 ms,
TI 1900 ms,
flip angle 150°,
matrix 358 x 448,
FOV 220,
TA 4:12 min)
T1w datasets were used for spatial and signal intensity normalization.
T2w datasets were used for the assessment of T2ll.
For each patient the recent T2w dataset was compared with the T2w dataset of the immediate preceding examination.
Material & Methods – Image Processing
For this study an algorithm,
which was introduced by Huppertz in 2011 [16],
was modified.
For all patients and all available T2w datasets the following steps were performed:
- Conversion of DICOM images to NIFTI format using dcm2nii [17]
- The most recent T1w image was spatially normalized to Montreal Neurological Institute (MNI) space and segmented in gray matter (GM),
white matter (WM) and cerebrospinal fluid (CSF).
The tissue classes were corrected for signal intensity inhomogeneities.
- Correction of the T2w images for signal intensity inhomogeneities.
- Coregistration of T1w and T2w images in native space.
- Normalization of the coregistered and bias corrected T2w image to MNI space.
- For signal intensity normalization of the T2w dataset,
the previously obtained results of GM and WM segmentation (step 2) were binarized and then used as masks to determine average intensities of GM and WM voxels in the T1 image and in the T2 image,
respectively.
- An intensity average of the whole brain was calculated as the mean of both tissue class intensities.
- After that the global brain intensity was set to an arbitrary value of 1000 and the resulting modulating factor applied to each voxel to achieve the intensity normalization of the brain and make MRI scans comparable to each other.
- Calculating a difference image (SI) between baseline and follow-up T2w image after coregistration and intensity normalization of both images using SPM12-based algorithm allowed the visual analysis of cT2ll.
Material & Methods – Quality Check
To ensure high quality of the used MR images,
the resulting images and its intermediate steps after every step the images were checked and excluded in case of lacking quality.
Material & Methods – Determination of changes in T2 Lesion Load
Changes in T2 lesion load were determined by evaluation of the subtractions images (SI) with MRIcro (Version 1.37 build 4).
Each SI was evaluated using standardized w/l-settings and standardized zoom.
Each change in SI clearly determined in three dimensions was counted as new lesion.
Exceptions were made regarding changes which were clearly located within a previously existing lesion,
and hence not counted as a new lesion (i.e.
change within lesion).
Written radiological reports were used for assessment of visual evaluation of cT2ll.
Material & Methods - Comparison between sadcT2ll and radiological findings
Radiological assessment of changes in T2ll (racT2ll) and sadcT2ll were compared to each other in order to categorize the results as true positive (TP),
false positive (FP),
true negative (TN) and false negative (FN):
If both methods had the same results,
they were both counted as TP – in case of change in T2ll – or as TN in case of no change,
respectively.
If the results of both methods differed from each other,
T2w images were reviewed by an experienced neuroradiologist and the results were categorized by arbitration for each method:
- TP: correctly assessed number of cT2ll – in case a change took place
- FP: falsely assessed number of cT2ll – either too many lesion changes found or detected change in SI without cT2ll
- TN: correctly assed cT2ll – in case no change took place
- FN: missed lesion changes – either less or no detected changes in case a change took place
Material & Methods - Statistics
For statistical evaluation of the results,
the assessed data were used to calculate statistical parameters: mean,
range,
prevalence for the T2 lesions.
Sensitivity,
specificity,
positive prediction value (PPV),
negative prediction value (NPV),
false positive rate (FPR),
false negative rate (FNR),
positive (LH+) and negative Likelihood Ratio (LH-) were calculated for the results regarding the quantitative assessment on cT2ll on the one hand and for the results disregarding quantitatively correct assessment on the other.
Calculations were performed with Microsoft Office Excel 2013.