Degree of Separability of IED and NonIED Waveforms
Abstract number :
3.153
Submission category :
Year :
2000
Submission ID :
1104
Source :
www.aesnet.org
Presentation date :
12/2/2000 12:00:00 AM
Published date :
Dec 1, 2000, 06:00 AM
Authors :
Mark D Bej, Richard C Burgess, Yoh-Han Pao, Cleveland Clin Fdn, Cleveland, OH; Case Western Reserve Univ, Cleveland, OH.
RATIONALE: To determine the degree of separability of interictal epileptiform discharges (IEDs) and non-IED waveforms when using dimension reduction via artificial neural network (ANN) pattern matching. METHODS: A 5-layer ANN was trained on raw EEG as previously described, with desired output values set to be identical with the input values. Three hidden layers were employed, with the middle layer containing very few (n=3) neurons. The network was trained to mean standard error (MSE) stability, ca. 10,000 iterations. RESULTS: Separation of IEDs and nonIEDs was achieved. IEDs were centered around (0.668, 0.554, 0.323), whereas the centroid of the nonIEDs was (0.426, 0.608, 0.592). The Euclidian distance between the centroids is 0.365. Standard deviations of distances of IED points from their centroid was 0.178, whereas that for nonIEDs was 0.164. CONCLUSIONS: This method reliably separates IEDs from nonIEDs. Moreover, the degree of separation can easily be manipulated. it is useful, therefore, as a prefilter component of a larger detection algorithm.