Visually validated semi-automatic high-frequency oscillation detection aides the delineation of epileptogenic regions during intra-operative electrocorticography

Objective: To test the utility of a novel semi-automated method for detecting, validating, and quantifying high-frequency oscillations (HFOs): ripples (80–200 Hz) and fast ripples (200–600 Hz) in intra-operative electrocorticography (ECoG) recordings. Methods: Sixteen adult patients with temporal lobe epilepsy (TLE) had intra-operative ECoG recordings at the time of resection. The computer-annotated ECoG recordings were visually inspected and false positive detections were removed. We retrospectively determined the sensitivity, specificity, positive and negative predictive value (PPV/NPV) of HFO detections in unresected regions for determining post-operative seizure outcome. Results: Visual validation revealed that 2.81% of ripple and 43.68% of fast ripple detections were false positive. Inter-reader agreement for false positive fast ripple on spike classification was good (ICC = 0.713, 95% CI: 0.632–0.779). After removing false positive detections, the PPV of a single fast ripple on spike in an unresected electrode site for post-operative non-seizure free outcome was 85.7 [50–100%]. Including false positive detections reduced the PPV to 64.2 [57.8–69.83%]. Conclusions: Applying automated HFO methods to intraoperative electrocorticography recordings results in false positive fast ripple detections. True fast ripples on spikes are rare, but predict non-seizure free post-operative outcome if found in an unresected site. Significance: Semi-automated HFO detection methods are required to accurately identify fast ripple events in intra-operative ECoG recordings.