Seizure onset detection based on a Uni- or Multi-modal Intelligent Seizure Acquisition (UISA/MISA) system

Abstract

Abstract-An automatic Uni-or Multi-modal Inteligent Seizure Acquisition (UISA/MISA) system is highly applicable for onset detection of epileptic seizures based on motion data. The modalities used are surface electromyography (sEMG), acceleration (ACC) and angular velocity (ANG). The new proposed automatic algorithm on motion data is extracting features as "log-sum" measures of discrete wavelet components. Classification into the two groups "seizure" versus "nonseizure" is made based on the support vector machine (SVM) algorithm.The algorithm performs with a sensitivity of 91-100%, a median latency of 1 second and a specificity of 100% on multi-modal data from five healthy subjects simulating seizures. The uni-modal algorithm based on sEMG data from the subjects and patients performs satisfactorily in some cases. As expected, our results clearly show superiority of the multimodal approach, as compared with the uni-modal one.

Automated seizure detection systems and their effectiveness for each type of seizure

“…On another study the MISA system had the best results followed by the combination ACM and sEMG [125]. Some patients found the suit uncomfortable so the prototype is under modification, probably involving smaller and fewer electrodes [125,128].…”

“…Seizure detection is more accurate if it combines more than one modality, as multimodal systems have shown increased sensitivity and lower FDR [125][126][127][128].…”

Evaluation of novel algorithm embedded in a wearable sEMG device for seizure detection

“…Several groups (including ourselves) have attempted to develop an effective alarm system based on accelerometer data [11], [12], [13], [14], [15], but with a performance which could be improved. Our previous results [11] on conventional sEMG data (measured with standard sEMG electrodes) were promising and we expect to achieve even better results with the wireless sEMG data, due to the avoidance of artifacts from wire-pulls.…”

Automatic multi-modal intelligent seizure acquisition (MISA) system for detection of motor seizures from electromyographic data and motion data

“…The discrete wavelet transformation (DWT) seems to be a good choice as a feature extraction method, since it provides a good frequency resolution at low frequencies and furthermore a good time resolution at the high frequencies. Based on this we used DWT for feature extraction and support vector machines (SVM) as a classifier in a pilot study [12], including both sEMG, ACM and ANG data, with very promising results on distinguishing between seizures/simulated seizures and normal activities.…”

“…Besides DWT we have therefore also tested the WPT as a method for extracting features for all modalities in this automatic multi-modal intelligent seizure acquisition (MISA) system. To classify our data into the two groups, seizures and normal activities, we used SVM [13] (as in our pilot study [12]) as a binary classifier trained on feature vectors from both classes, since it is well known to function better than other classifiers when the data classes are of unequal sizes. We used data from healthy subjects who simulated seizures (as instructed by a physician) to develop our algorithm upon.…”