MRI - data

We assembled four sets (A, B, C, D) of each six MRIs of people with epilepsy. In each set one lesion of the subsequently specified five typical EL-categories (according to large patient series at tertiary epilepsy centers [5, 6]) and one non-lesional MRI were given. No individual lesion was represented more than once in each data set. The six lesion categories are:

Hippocampal sclerosis or limbic encephalitis (HS/LE)

Focal cortical dysplasia (FCD)

Periventricular nodular or other heterotopia (PNH)

Long-term epilepsy associated tumors (LEAT; ganglioglioma, dysembryoplastic neuroepithelial tumor, pleomorphic xantroastrocytoma)

Gliotic scar / hemosiderin (e. g. due to cavernoma) (G/H/C)

Non - lesional MRI (NL)

The lesions shown in the MRIs are representative for the common spectrum of the respective category in terms of their volume and signal characteristics. We did not stratify the lesions with regards to volume or signal specifics, neither within nor between lesion categories since this wouldn’t mirror clinical reality and was not the scope of this study. To find out, however, if lesion size correlates with their recognition we estimated the volume of each lesion (see below).

MRIs in 18 patients (including all in the FCD and NL groups) were acquired according to an epilepsy specific MRI protocol at our centre at a field strength of 3 T including the following sequences: volumetric MPRAGE (magnetization prepared rapid acquisition with gradient echo) 1x1x1mm, volumetric FLAIR (fluid attenuation inversion recovery) 1x1x1mm, T2 coronal 0.56 × 0.56 mm, 2.2 mm slice gap, T2 STIR axial images 0.45 × 0.45 mm, slice gap 3.75 mm, SWI axial 0.72 × 0.72 mm, 1.2 mm slice gap [7]. This protocol meets the HARNESS (harmonized neuroimaging of epilepsy structural sequences) criteria suggested by the ILAE neuroimaging taskforce on epilepsy [4]. In six patients (three in the G/H/C group, two LE, one PNH), examinations acquired at 1.5 T in outside practices were used, some lacking volumetric sequences. The EL, if present, was clearly recognizable in at least one sequence on the laptop computer used to display the MRI scans for the study. Lesion volume was estimated by the product of the maximal diameter in the axial, coronal and sagittal plane divided by 2. The lesions of MRI set C are shown in Fig. 1.

Representative images of epileptogenic lesions in MRI sets C. A Left frontal cortical dysplasia, axial FLAIR. B left frontal long-term epilepsy associated tumor, axial T2. C Right occipital hemosiderin deposits after hemorrhage, axial T2*. D right hippocampal sclerosis, coronal T2. E left periventricular nodular heterotopia, coronal T2