Abstract
AbstractObjectiveDeep learning methods have shown potential in automating interictal epileptiform discharge (IED) detection in electroencephalograms (EEGs). While it is known that these algorithms are dependent on the type of data used for training, this has not been explored in EEG analysis applications. We study the difference in performance of deep learning algorithms on routine and ambulatory EEG data.MethodsWe trained the same neural network on three datasets: 166 routine EEGs (VGGC–R), 75 ambulatory EEGs (VGGC–A) and a combination of the two data types (VGGC-C, 241 EEGs total). Networks were tested on 34 routine EEGs and 33 ambulatory recordings, where all 2 s non-overlapping epochs were labeled with a probability that expressed the likelihood of containing an epileptiform discharge. Performance was quantified as sensitivity, specificity and the rate of false detections (FPR).ResultsThe VGGC-R led to 84% sensitivity at 99% specificity on the routine EEGs, but its sensitivity was only 53% on ambulatory EEGs, with a FPR > 3 FP/min. The VGGC-C and VGGC-A yielded sensitivities of 79% and 60%, respectively, at 99% specificity on ambulatory data, but their sensitivity was less than 60% for routine EEGs.ConclusionWe show that performance of deep nets for IED detection depends critically on the type of recording. The VGGC-R should be used for routine recordings and the VGGC-C should be used for ambulatory recordings for IED detection.SignificanceThe type of data used to train algorithms should be optimized according to their application, as this has a significant impact on algorithm performance.
Publisher
Cold Spring Harbor Laboratory