110 Chapter 5 handedness, and WHO grade. For the stroke patients characteristics were obtained either from the semi-structured interview before the neuropsychological assessment and/or by reviewing the electronic patient file. This data included sex, age, level of education, handedness, stroke location based on MRI, date of stroke onset and medical history. Image processing Image Acquisition A T2 FLAIR sequence was used for lesion delineation in the present study. For the glioma patients this was acquired as part of standard clinical care and the pulsesequence details of the FLAIR MRI varied between glioma patients. T2 FLAIR scans with a slice thickness >5 mm were excluded in order to maintain adequate quality for lesion segmentation for all included glioma patients. The MRI scan that was closest to the pre-operative neurocognitive testing was chosen. Depending on the medical center, stroke patients underwent a 3T MRI on a Philips Ingenia R5 (Amsterdam UMC and UMCU) or on the Siemens Magnetom Prisma (Radboudumc and UMCG), using a 32-channel head coil. For the Philips scanner the pulse-sequence details were: 3D T2 FLAIR (TI = 1650ms, TR = 4800ms, TE = 253ms, [FOV] = 250mm, voxel size 1.12×1.12×0.56mm). For the Siemens scanner they were 3D T2 FLAIR (TI = 1650ms, TR = 4800ms, TE = 484ms, [FOV] = 280mm, voxel size 0.9×0.9×0.9mm). Lesion delineation Both tumor and stroke lesions were segmented on individual T2 FLAIR images. Both tumor and stroke lesion were first drawn in the axial plane and adjusted accordingly in the sagittal and coronal plane. Tumor lesions were delineated using the Smartbrush implemented in the iPlan v3.0 software (BrainLab AG, Feldkirchen, Germany) and represent the total lesion volume, including both tumor and edema. For the tumor patients, a training set (N = 22) was completed in which all tumor regions were drawn by two researchers (EG & VR) under the supervision of an experienced neurologist (TS). After completing the training set, tumor regions were drawn by one of the two researchers. Through consensus meetings with the neurologist (TS), definitive lesion maps were created. The interrater reliability was calculated as the number of voxels included by both raters, in reference to the mean number of voxels selected per rater.33 Based on eight different tumor lesions, the interrater reliability was 93.0% (range 88.8-96.7%). Stroke lesions were delineated semi-automatically or manually with the ITK-snap software.34 Stroke lesions were delineated by three researchers and in case of doubt for specific scans, a neurologist
RkJQdWJsaXNoZXIy MTk4NDMw