NON-INVASIVE HMM-BASED DETECTION OF INTERICTAL ACTIVITY PROPAGATION BETWEEN SEVERAL EPILEPTOGENIC FOCI IN PATIENTS WITH MULTI-FOCAL EPILEPSY
Abstract number :
2.166
Submission category :
Year :
2003
Submission ID :
3996
Source :
www.aesnet.org
Presentation date :
12/6/2003 12:00:00 AM
Published date :
Dec 1, 2003, 06:00 AM
Authors :
Alexei Y. Ossadtchi, John C. Mosher, William W. Sutherling, Richard M. Leahy Electrical Engineering, University of Southern California, Los Angeles, CA; Los Alamos National Laboratory, Los Alamos, NM; Epilepsy and Brain Mapping program, Huntington Medical
Automatic spike detection techniques applied to interictal MEG data usually discover several potential epileptogenic regions in the brains of patients with partial epilepsy. For treatment planning it is important to determine which of the detected regions are the primary generators of abnormal interictal activity. Analysis of the patterns of interictal activity propagation between the detected regions may allow for detection of the primary epileptic foci. We describe use of a Hidden Markov Models (HMM) for estimation of the propagation patterns between several epileptogenic regions from interictal MEG data.
Five minutes of interictal MEG activity was recordered using a 68-sensor whole cortex neuromagnetometer (CTF Systems) from an adult male with multifocal epilepsy. We use the automatic spike detection technique described in [1] in order to find several regions of abnormal activity. We then use a matched adaptive subspace detector to find time markers for the spikes coming from each of the regions. Then we use the Expectation-Maximization (EM) algorithm in order to identify parameters of a HMM structure that models propagation of interictal activity between the detected foci. Analysis of the estimated transition probability matrix allows us to make conclusions regarding propagation patterns of abnormal activity and to identify potential epileptogenic foci.
We applied the method to three independent interictal datasets collected from a patient with multi-focal epilepsy. Several epileptogenic regions were automatically detected and our propagation detection technique was applied. In all three datasets propagation from the temporal lobe to the frontal lobe was reliably detected. A subsequent surgery in which this portion of the temporal lobe (as well as the portion of the frontal lobe) was removed has left the patient seizure free for already more than 36 months.
The procedure described is a promising approach to non-invasive detection of the propagation patterns in patients with multi-focal epilepsy. The HMM framework allows explicit modeling of interactions between foci and also allows for direct incorporation of the spike detector specificity and sensitivity characteristics. Preliminary results show that the model is capable of capturing patterns of interaction in interictal MEG data in patients with partial focal epilepsy.
[1] A. Ossadtchi, J. C. Mosher, S. Baillet, N. Lopez, W.S. Sutherling, R. M. Leahy, [italic][quot]Automated Detection of Dipole Clusters in Interictal MEG Data[quot][/italic] , 13th International Conference on Biomagnetism, August 8-14 2002, Jena, Germany
[Supported by: National Institutes of Health grants EB002010, Huntington Medical Research Institutes, Huntington Hospital, Epilepsy and Brain Mapping Program, and Robert S. and Denise Zeilstra. ]