Elsevier

Epilepsy Research

Volume 97, Issues 1–2, November 2011, Pages 146-156
Epilepsy Research

Automated quantitative FLAIR analysis in hippocampal sclerosis

https://doi.org/10.1016/j.eplepsyres.2011.08.001Get rights and content

Summary

Purpose

To describe and evaluate a novel MRI post-processing technique for automated quantitative hippocampal FLAIR analysis in patients with hippocampal sclerosis (HS).

Patients and methods

Based on a method for FLAIR analysis presented by Focke et al. (2009), T1 and coregistered FLAIR scans of individual subjects were processed together in SPM5 to conduct both a spatial and an intensity normalization of the FLAIR scans. In a further development described here, the resulting normalized FLAIR images were thresholded and weighted by a probabilistic hippocampal mask to determine the average FLAIR intensities of left and right hippocampus. The method was applied to the MRI data of 103 HS patients and 131 controls. Using a 95% confidence region calculated from the FLAIR intensities of controls as threshold, the performance in discriminating both groups was assessed.

Results

One hundred of 103 patients and among those all 23 patients with histologically confirmed HS fell outside the 95% confidence region, amounting to 97.1% sensitivity. All but 6 controls (=95.4%) were found within the confidence region, corresponding to the expected specificity. The method could also distinguish bilateral HS and visualize signal changes after status epilepticus.

Conclusion

Automated FLAIR analysis is a promising tool to quantify hippocampal signal alterations, to support the detection of HS, and to monitor the temporal evolution of the disease.

Introduction

Hippocampal sclerosis (HS) is the most frequent cause of pharmaocoresistant temporal lobe epilepsy (TLE) and the most common histopathologic diagnosis in patients with TLE undergoing epilepsy surgery (Urbach et al., 2004, Wiebe et al., 2001). The primary features of HS in magnetic resonance imaging (MRI) are atrophy of the hippocampus and hyperintense signal in fluid-attenuated inversion-recovery (FLAIR) and T2-weighted sequences (Jack et al., 1996, Jackson et al., 1990, Jackson et al., 1993a). Visual detection of clear unilateral HS in good-quality MR images by experienced radiologists is considered unproblematic (Van Paesschen, 2004), but the recognition can be difficult for subtle sclerosis or cases of bilateral HS where a side comparison is hampered. In addition, subtle changes of T2 signal intensity over time, e.g. during potential progression of limbic encephalitis to HS, are difficult to assess by pure visual analysis. Focke et al., 2008, Focke et al., 2009 have recently described a new method for quantitative analysis of FLAIR scans, i.e. images with T2-weighted contrast but complete suppression of high signal intensity of cerebrospinal fluid (CSF). Their approach avoids difficulties due to partial volume effects with CSF which are to consider for other quantitative methods like T2 relaxometry. Compared to T2 mapping with FLAIR CSF suppression (Rugg-Gunn et al., 2005) it does not require a special FLAIR T2 map which has a long acquisition time and is often not available but takes a standard clinical FLAIR spin echo sequence as input. Essentially, the voxel-based method performs both a spatial and intensity normalization of FLAIR images by using internal reference regions and spatial normalization parameters derived from combined normalization and segmentation of a coregistered T1-weighted image (Focke et al., 2008, Focke et al., 2009). The normalized and rescaled FLAIR images are the starting point for a whole brain FLAIR analysis which has been shown to be successful in the detection of focal cortical dysplasia. In the present study, we describe a further development of this method for regional quantitative FLAIR analysis of the hippocampus and present first results in patients with HS compared to controls.

Section snippets

Patients and controls

At the Swiss Epilepsy Centre in Zürich, a register has been maintained for all patients receiving an MRI since January 2006. The majority of these patients had a high-resolution MRI at a Philips Achieva 3T scanner (Philips, Amsterdam, The Netherlands) according to a dedicated epilepsy protocol which comprised the following sequences: an unenhanced T1-weighted volume data set (3D Turbo Field Echo sequence with 1 mm3 voxel; TR 8.1 ms; TE 3.7 ms; flip angle 8°; field of view 256 mm × 256 mm, slab

Results

Between January 2006 and April 2011, 1640 patients of the Swiss Epilepsy Centre received an MRI at the local Philips 3T scanner according to the protocol described above. Based on clinical history, seizure semiology, and EEG findings, 998 of these patients (=61%) were classified as suffering from focal epilepsy (i.e. having seizures which originate from and at least at the beginning are limited to a part of one hemisphere). HS was diagnosed in 115 of these patients. One patient had to be

Interpretation of results

In the present study, a novel method for automated quantitative FLAIR analysis of the hippocampus has been evaluated in patients with HS and controls, with the following main findings:

  • -

    Separation of HS patients and controls by this method was in high agreement with prior classification according to visual assessment. This was especially true for the subgroup of histologically confirmed HS patients (i.e. sensitivity of 100%).

  • -

    The method allowed to separate right and left HS without overlap.

  • -

    The

Conclusion

The method of automated quantitative FLAIR analysis of the hippocampus presented here appears to be a valuable additional diagnostic tool in the evaluation of epilepsy patients which may assist in the detection of HS and in monitoring the temporal evolution of the disease. Due to spatial and intensity normalization of the FLAIR images, signal changes in the hippocampus can be quantified and directly compared to the results of healthy controls and other HS patients. The automated image

Acknowledgements

The development of the MRI post-processing technique presented here was kindly supported by the Swiss National Foundation. Dr. Bernd Weber was supported by a Heisenberg Grant of the DFG (WE 4427/3-1) and by the SFB TR3 projects A1 and A8.

The authors are grateful to Dr. Dominik Huber, MRI Institute, Schulthess Clinic, Zürich, for providing and conducting the MRI measurements.

References (39)

  • L. Bonilha et al.

    Automated MRI analysis for identification of hippocampal atrophy in temporal lobe epilepsy

    Epilepsia

    (2009)
  • R.S. Briellmann et al.

    Hippocampal pathology in refractory temporal lobe epilepsy: T2-weighted signal change reflects dentate gliosis

    Neurology

    (2002)
  • N.K. Focke et al.

    Voxel-based analysis of whole brain FLAIR at 3T detects focal cortical dysplasia

    Epilepsia

    (2008)
  • N.K. Focke et al.

    Automated normalized FLAIR imaging in MRI-negative patients with refractory focal epilepsy

    Epilepsia

    (2009)
  • A. Hammers et al.

    Three-dimensional maximum probability atlas of the human brain, with particular reference to the temporal lobe

    Hum. Brain Mapp.

    (2003)
  • H.J. Huppertz et al.

    Automatic striatal volumetry allows for identification of patients with chorea-acanthocytosis at single subject level

    J. Neural. Transm.

    (2008)
  • H.J. Huppertz et al.

    Voxel-based 3D MRI analysis helps to detect subtle forms of subcortical band heterotopia

    Epilepsia

    (2008)
  • C.R. Jack et al.

    Mesial temporal sclerosis: diagnosis with fluid-attenuated inversion-recovery versus spin-echo MR imaging

    Radiology

    (1996)
  • G.D. Jackson et al.

    Optimizing the diagnosis of hippocampal sclerosis using MR imaging

    AJNR Am. J. Neuroradiol.

    (1993)
  • Cited by (36)

    View all citing articles on Scopus
    1

    Tel.: +49 228 6885 262; fax: +49 228 6885 261.

    2

    Tel.: +49 40 5077 3345.

    3

    Tel.: +49 228 287 16507; fax: +49 228 287 16093.

    View full text