Challenges for multivariate and multimodality analyses in "real life" projects: Epilepsy Susanne Mueller M.D. Center for Imaging of Neurodegenerative Diseases
Background: Epilepsy What is epilepsy? Recurrent seizures unprovoked by any identifiable cause. How common is epilepsy? About 1.3 3.1 % of the population will develop epilepsy during their lifetime. What types of epilepsy are there? Two types: a. Generalized Epilepsy. b. Partial Epilepsy with and without lesion.
Treatment of Epilepsy a. Pharmacological Treatment Suppression of epileptogenic activity by increasing inhibition or decreasing excitation. HOWEVER, more than 30% of patients continue to have seizures despite adequate pharmacological treatment. b. Surgical Treatment Resection of the brain area responsible for seizure generation.
Goals of Presurgical Evaluation Identification of the epileptogenic brain region. Assessment of risk of epilepsy surgery. Prediction of surgical outcome regarding seizure freedom or at least significant reduction of seizure frequency.
Standard Presurgical Evaluation Where is the epileptogenic focus? Non-invasive exams: Clinical and electrophysiological (EEG) information. Neuroimaging: MRI, optional PET, ictal SPECT. Invasive exams: Intracranial EEG recordings.
Presurgical Evaluation: New Imaging Approaches Structural Imaging: Quantitative analysis, additional modalities, e.g. DTI, high resolution MRI Metabolic Imaging: Perfusion MRI MR spectroscopy Functional Imaging: EEG fmri
Multimodality Definition of the Epileptic Focus: A Simple Cooking Receipt? 1. Use control population to establish the physiological range of each imaging modality. 2. Determine region of greatest deviation from physiological range in patients. 3. Apply weight to each modality to account for differences in sensitivity and specificity. 4. Concatenate modalities and identify region with the most consistent and most severe abnormalities = epileptic focus.
Challenges for Mulitvariate/Multimodality Analyses Imaging features of the focus. Extrafocal abnormalities. Findings dependent on analysis technique even within same imaging modality. Technical challenges/missing data. Small study population/single case.
Challenge: Imaging Characteristics of Epileptic Focus Requirement: Imaging characteristics of focus stable within modality for epilepsy type of interest. Structural imaging/volumetry: Volume loss but also volume gain. Perfusion Imaging: Regional hypoperfusion but also hyperperfusion. MR spectroscopy: NAA decreases but also Cho increases.
Challenge: Extrafocal Abnormalities Cortical Thinning in TLE-MTS NAA/Cr+Cho reductions in TLE-MTS
Challenge: Single Subject Extrafocal Abnormalities Ipsi Ipsi FEP006-1 Contra FEP006-1 Contra FEP120-1 Contra FEP120-1
Challenge: Influence of Analysis Cortical Thickness: Freesurfer Technique: TLE-MTS IPSI CONTRA Gray Matter Volumes: Voxel-Based Morphometry CONTRA IPSI Brain Volumes: Deformation Based Morphometry CONTRA IPSI Dr. C. Scanlon, CIND
Challenge: Influence of Analysis Technique: TLE-no Cortical Thickness: Freesurfer IPSI CONTRA Gray Matter Volumes: Voxel-Based Morphometry CONTRA IPSI Brain Volumes: Deformation Based Morphometry CONTRA IPSI Dr. C. Scanlon, CIND
Challenge: Technical Challenges/Missing Data T1 Gray matter map CBF map NAA/(Cr+Cho) map
Challenge: Small Population/Single Case Currently available multivariate/multimodality approaches require large data sets. Partial epilepsy is a heterogeneous disease (different hemisphere, lobes, lesional/nonlesional) and thus study populations usually do not exceed 20 subjects. Ultimate goal is the delineation of the epileptogenic brain region in a single subject.
Conclusions Multivariate/multimodality approaches are potentially very powerful tools for image analysis. Requirements for conventional multivariate/multimodality approaches (study population, correction for multiple comparisons) limit their applicability. Solution unconventional approaches? For example: - Voxel the optimal unit? - Probabilistic multivariate/multimodality disease atlases?