Abstracts

Rationale: High-frequency oscillations (HFOs, 80 - 500 Hz) in intracranial EEG are reliable markers for epileptogenic cortex. Also, in the field of human intracranial cognitive electrophysiology, it has become increasingly evident that physiological responses of a cortical tissue can be measured with high-frequency broadband (HFB) activity that is time-locked to the appearance of relevant stimuli. To date, it has remained unknown whether cortical tissue with abundant intrinsic HFOs is also capable of generating normal physiological HFB responses during a cognitive task. A systematic comparison of signal properties of epileptic HFOs compared to functional HFBs is also lacking. Methods: We recruited three patients with anatomically similar electrode coverage in the occipital and ventral temporal cortices who participated in the same visual recognition task. HFOs were identified and mapped using an unsupervised method previously validated for the detection of spontaneous HFOs specific to seizure onset zones. We then measured the functional responses in the pathological sites by computing stimulus-locked HFB activations. We compared the temporal and spectral profiles and calculated the probability of temporal discordance of HFOs and HFBs. In addition, we introduced metrics that are optimal for the discrimination between HFOs and HFBs, and used these features in the classification of HFOs and HFBs in a supervised SVM classifier. Results: In all three subjects, the brain sites with abundant pathological HFOs were capable of generating physiological responses during the presentation of relevant stimuli. The pathological HFOs and stimulus-locked HFBs in the same sites fell within the similar frequency range (centering frequency 125 Hz vs. 120 Hz, p = 0.17) but showed group-wise difference in the signal duration (82 ms vs. 501 ms, p < 0.05) and spectral width (32 Hz vs. 92 Hz, p < 0.05). A significant change in the slope of power spectral density was found only in HFOs (p < 0.05) but not in HFBs. More importantly, epileptic HFOs were temporally discordant with stimulus-induced HFBs. The SVM classifier successfully separated HFOs from HFBs using our proposed features, achieving an AUC value of 0.98. Conclusions: Epileptic brain structures show normal physiological responses to relevant cognitive stimuli as the epileptic and physiological activities do not coincide in time. Investigating the connection and distinctions between HFOs and HFBs has practical implications, and should shed light on the cognitive reserve function of epileptic neuronal populations. Funding: Not applicable