Automatic detection of fast ripples 1

Highlights 1 1) We propose a novel method for automatically detecting fast ripples (FRs, 250-600 Hz) 2 2) The signal energy in low and high frequency bands is used to classify EEG events as FRs, interictal 3 epileptic spikes or artifacts. 4 3) The sensitivity and the specificity of this method is high enough to avoid " false ripples " caused by 5 sharp transients. 6 7 8 9 Abstract 10 Objective: We propose a new method for automatic detection of fast ripples (FRs) which have been 11 identified as a potential biomarker of epileptogenic processes. 12 Methods: This method is based on a two-stage procedure: i) global detection of events of interest 13 (EOIs, defined as transient signals accompanied with an energy increase in the frequency band of 14 interest 250-600 Hz), and ii) local energy vs. frequency analysis of detected EOIs for classification as 15 FRs, interictal epileptic spikes or artifacts. For this second stage, two variants were implemented based 16 either on Fourier or Wavelet Transform. The method was evaluated on simulated and real depth-EEG 17 signals (human, animal). The performance criterion was based on receiving operator characteristics. 18 Results: The proposed detector showed high performance in terms of sensitivity and specificity. 19 Conclusions: As designed to specifically detect FRs, the method outperforms any method simply 20 based on the detection of energy changes in high-pass filtered signals and avoids spurious detections 21 caused by sharp transient events often present in raw signals 22 Significance: In most of epilepsy surgery units, huge data sets are generated during pre-surgical 23 evaluation. We think that the proposed detection method can dramatically decrease the workload in 24 assessing the presence of FRs in intracranial EEGs. 25 26 27 1 Introduction 1 High frequency oscillations (HFOs) have been a topic of increasing interest in neuroscience over the 2 past decade. They now constitute a novel trend in neurophysiology (Jefferys et al., 2012) that has been 3 made possible with the development of digital EEG equipments allowing for high sampling rates and 4 with the identification of oscillations at up to 600 Hz in animals (Buzsaki and Lopes da Silva, 2012). 5 Among the wide diversity of HFOs which dominant frequency can vary from 30 Hz to 600 Hz, fast 6 ripples (FRs) are particular transient oscillations (a few tens of ms) occurring in the frequency band 7 ranging from 250 Hz …