br Data The data presented in

Data The data presented in this article is related to the research article entitled “Longitudinal multiple sclerosis lesion segmentation: Resource and challenge\" [1]. The data consists of magnetic resonance (MR) images (MRI) divided into two cohorts: 1) Training Set; and 2) Test Set. The Training Set consists of five subjects, four of which had four time-points, while the fifth subject had five time-points. The Test Set includes fourteen subjects, ten of which had four time-points, three had five time-points, and one had six time-points. Two consecutive time-points are separated by approximately one year for all subjects. Table 1 includes a demo-graphic breakdown for the training and test data sets. The data does not supply the multiple sclerosis (MS) subtype of the subjects for either the training or the test data. The data is available for download from the Challenge Evaluation Website: http://smart-stats-tools.org/lesion-challenge-2015.
Methods Each scan was imaged and preprocessed in the same manner, with data acquired on a 3.0T MRI scanner (Philips Medical Systems, Best, The Netherlands) using the following sequences: a T1-weighted (T1-w) magnetization prepared rapid gradient echo (MPRAGE) with TR=10.3ms, TE=6ms, flip angle=8°, & 0.82×0.82×1.17mm3 voxel size; a double spin echo (DSE) which produces the proton density weighted (PD-w) and T2-weighted (T2-w) images with TR=4177ms, TE1=12.31ms, TE2=80 ms, & 0.82×0.82×2.2mm3 voxel size; and a T2-w fluid attenuated order Nutlin 3 recovery (FLAIR) with TI=835ms, TE=68ms, & 0.82×0.82×2.2mm3 voxel size. The imaging protocols were approved by the local institutional review board. Each subject underwent the following preprocessing: the baseline (first time-point) MPRAGE was inhomogeneity-corrected using N4 [2], skull-stripped [3,4], dura stripped [5], followed by a second N4 inhomogeneity correction, and rigid registration to a 1mm isotropic MNI template [6]. We have found that running N4 a second time after skull and dura stripping is more effective (relative to a single correction) at reducing any inhomogeneity within the images (see Fig. 1 for an example image set after preprocessing). Once the baseline MPRAGE is in MNI space, it is used as a target for the remaining images. The remaining images include the baseline T2-w, PD-w, and FLAIR, as well as the scans from each of the follow-up time-points. These images are N4 corrected and then rigidly registered to the 1mm isotropic baseline MPRAGE in MNI space. Our registration steps are inverse consistent and thus any registration based biases are avoided [7] The skull & dura stripped mask from the baseline MPRAGE is applied to all the subsequent images, which are then N4 corrected again. For each time-point of every subject׳s scans in the Training Set and Test Set, the following data are provided: the original scan images consisting of T1-w MPRAGE, T2-w, PD-w, and FLAIR, as well as the preprocessed images (in MNI space) for each of the scan modalities. The Training Set also included manual delineations by two experts identifying and segmenting the white matter lesions on the MR images. To facilitate the dissemination of the data and promote the sharing of results we have created a website. Visitors to the site can see a list of the Top 25 submitted results. Groups interested in running their methods on the data need only register for an account, download the data, and upload their results. The uploader of the results will receive an e-mail within ten minutes detailing the results on a per subject and per time-point basis. The report includes the following computed metrics: Dice, Jaccard, positive predictive value (PPV), true positive rate (TPR), lesion false positive rate (LFPR), lesion true positive rate (LTPR), absolute volume difference (AVD), average symmetric surface distance (ASSD), algorithm and manual lesion volume. The website also provides an overall score; for a given algorithm A this is computed as follows,