MR-based brain Morphometry Improves Localization of Focal Cortical Dysplasia at Individual Level
Xin Chen1, Tianyi Qian2, Bénédicte Maréchal3,4,5, Nan Chen1, and Kuncheng Li1

1Radiology, Xuanwu Hospital, Capital Medical University, Beijing, China, People's Republic of, 2MR Collaborations NE Asia, Siemens Healthcare, Beijing, China, People's Republic of, 3Advanced Clinical Imaging Technology (HC CMEA SUI DI BM PI), Siemens Healthcare AG, Lausanne, Switzerland, 4LTS5, Ecole Polytechnique Fédérale de Lausanne, Lausanne, Swaziland, 5Radiology, University Hospital (CHUV), Lausanne, Switzerland


In order to explore the potential of volume-based morphometry for computer-aided diagnosis on an individual level, we evaluated a volume-based morphometric MRI analysis prototype for detection of cortical abnormalities in individual focal cortical dysplasia (FCD) patients for whom no lesion was reported after routine MR exam. Using intracranial EEG as the gold standard, the results of a performed ROC analysis show good detection performance with AUC=0.882, sensitivity =93.88%, and specificity 79.57% at the optimal cut-off point. These results suggest that such automated methods provide additional value for MR-based diagnostics.


Focal cortical dysplasia (FCD) is a common cause of medically intractable seizures in both children and adults. Detecting and locating lesions accurately is beneficial to pre-surgical planning and surgical outcomes. Therefore, many modalities have been used in combination to improve the accuracy, such as clinical assessments, EEG, MEG, intracranial EEG, SPECT, PET-CT/MRI and MRI 1. However, performing all of these methods is time-consuming and costly, especially the routine MR exam which sometimes may not provide any conclusive findings. T1-weighted 3D-MPRAGE MR imaging is commonly used in combination with MEG for clinical preoperative localization. In this study, we aimed at evaluating the potential of a volume-based morphometric MRI analysis based on 3D-T1w MPRAGE for detecting and locating FCD lesions at the individual level.


We retrospectively analyzed 16 pharmacoresistant epilepsy patients who had undergone surgery. The inclusion criteria were: (1) existence of a preoperative 1.5 or 3T MRI with conventional protocol (T2-w, axial and coronal FLAIR) and 3D T1w MPRAGE scan, (2) histologically proven FCD, (3) existence of a >1 month of postsurgical follow-up. The exclusion criteria were: (1) <16 years old because of the lack of age-matched healthy controls, (2) poor MRI image quality. We obtained the brain volume information about white and grey matter (defined as V) for each patient of several brain structures, and then compared them individually with an age-matched reference range ([Vmin,Vmax]) calibrated on a large Caucasian database2 with Vmin and Vmax corresponding to the 10th and 90th percentiles, respectively. The volume of each area was then transformed into z-score by using the mean and standard deviation of normal population. The surgical resection regions, mapped comprehensively from clinical symptoms, EEG and/or MEG, intracranial EEG, SPECT and/or PET-CT, and conventional MRI, were regarded as the duty focus, as long as there was no post-operative recurrence within 1 month follow-up. Each patient was scanned on a MAGNETOM Trio Tim 3T MR scanner (Siemens Healthcare, Erlangen, Germany) using a 3D MPRAGE sequence (TR=2300ms; TE=3.01ms; TI=900ms; flip angle=9°; matrix size=256x256x176; voxel size=1×1×1mm3). Prior to this acquisition, a routine MRI exam was performed to screen for obvious abnormalities, e.g., patchy cerebral infarcts or occupying lesions. All MPRAGE scans were then processed using the MorphoBox prototype, a volume-based morphometry package2. A total of 25 brain structure volumes normalized by the total intracranial volume (TIV) were obtained, namely: global grey matter, white matter and CSF, thalamus, putamen, caudate, pallidum, deep white matter, hippocampus, ventricles, cingulate gyrus, insula, cerebellum, mesencephalon, pons, medulla oblongata, corpus callosum, grey and white matter in frontal/parietal/occipital/temporal lobe (Table.1). Finally, a Receiver Operating Characteristics (ROC) curve was used to evaluate the performance of the z-score to predict the location of duty focus.


Clinical profiles of the patients are summarized in Table 1. All patients have structures with out-of-range volumes which are difficult to assess visually on conventional MRI. 14/16 patients (87.5%) have duty foci located in the structures with out-of-range volumes. Moreover, these areas showed atrophy in 12 out of 14 patients (85.7%). The results of the ROC analysis show that the morphometric MRI analysis exhibits good performance in detecting duty focus with AUC=0.882, sensitivity=93.88%, and specificity 79.57% providing a cut-off point value of -3.03.


Detecting FCD lesions using MRI is still challenging as lesions may be very subtle or appearing normal. Our findings suggest that atrophic regions, as revealed by the MorphoBox prototype, may be a more suitable aid for surgery than structures showing increased volume. The volume decrease detected by morphometric MRI analysis demonstrated focal thickening and also suggested that focal brain atrophy was the basic morphological feature of FCD.

One interesting finding in patient No.8 is that his surgical resection regions were located in areas with out-of-range volumes, and the surgical outcome was bad: there was recurrence on the second day after surgery. In our morphometric MRI analysis, white-matter volume in left temporal (was not cut in surgery) was decreased and grey matter in right parietal (was cut in surgery) was increased. A VEEG monitoring test on the twenty-third day after surgery showed many shape waves in left frontal, temporal and midline, and few shape waves in right posterior and central frontal, which implied abnormalities both in the left temporal and right frontal lobe. Furthermore, the regions including epilepsy focus were all shown degeneration than normal subjects.


The MorphoBox prototype shows good performance in guiding the clinical diagnosis in FCD patients without any positive findings in routine MR scan. This prototype provides a time-saving individualized quantitative analysis which could have a great potential in routine practice for detecting and localizing duty focus of focal cortical dysplasia.


No acknowledgement found.


1. Guerrini R, Duchowny M, Jayakar P, et al. Diagnostic methods and treatment options for focal cortical dysplasia.Epilepsia 2015;56:1669-1686.

2. Schmitter D, Roche A, Maréchal B, et al. An evaluation of volume-based morphometry for prediction of mild cognitive impairment and Alzheimer's disease. NeuroImage: Clinical, 2015, 7: 7-17.


Figure 1: An example of patients with FCD in the right temporal lobe (patient No.16). The right temporal angle is slightly larger than the left. Yellow circles in ECoG are the areas showing epileptic discharge. Different structures were labeled in different color as a mask after morphometric analysis.

Figure 2: The ROC analysis comparing all the patients’ abnormal degree for each brain area with duty focus. The AUC, the best cutoff value sensitivity and the specificity were 0.882, 93.88%, and 79.57%, respectively.

Table 1: The clinical data, the regions of surgical resection and the results of morphometric MRI analysis of the FCD patients.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)