Interrater reliability of visually evaluated high frequency oscillations
Aaron M Spring1, Daniel J Pittman2, Yahya Aghakhani3
1Department of Clinical Neurosciences, University of Calgary, Calgary, AB, Canada; Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada; Seaman Family MR Research Centre, Foothills Medical Centre, Calgary, AB, Canada.
Objective:
High frequency oscillations (HFOs) and interictal epileptiform discharges (IEDs) have been shown to be markers of epileptogenic regions. However, there is currently no 'gold standard' for identifying HFOs. Accordingly, we aimed to formally characterize the interrater reliability of HFO markings to validate the current practices.
Methods:
A morphology detector was implemented to detect events (candidate HFOs, lower-threshold events, and distractors) from the intracranial EEG (iEEG) of ten patients. Six electroencephalographers visually evaluated these events for the presence of HFOs and IEDs. Interrater reliability was calculated using pairwise Cohen's Kappa (κ) and intraclass correlation coefficients (ICC).
Results:
The HFO evaluation distributions were significantly different for most pairs of reviewers (p<0.05; 11/15 pairs). Interrater reliability was poor for HFOs alone (κmean=0.403; ICC=0.401) and HFO+IEDs (κmean=0.568; ICC=0.570).
Conclusions:
The current practice of using two visual reviewers to identify HFOs is prone to bias arising from the poor agreement between reviewers, limiting the extrinsic validity of studies using these markers.
Significance:
The poor interrater reliability underlines the need for a framework to reconcile the important findings of existing studies. The present epoched design is an ideal candidate for the implementation of such a framework.
