Authors :
Presenting Author: Masamune Kimura, MD – Kobe University Graduate School of Medicine
Katsuya Kobayashi, MD, PhD – Kyoto University Graduate School of Medicine
Kazuma Koda, MD – Kobe University Graduate School of Medicine
Yosuke Fujimoto, MD – Kobe University Graduate School of Medicine
Mayumi Otani, MD – Kobe University Graduate School of Medicine
Masaya Togo, MD, PhD – Kobe University Graduate School of Medicine
Masahiro Sawada, MD, PhD – Kyoto University Graduate School of Medicine
Akihiro Shimotake, MD, PhD – Kyoto University Graduate School of Medicine
Takayuki Kikuchi, MD, PhD – Kyoto University Graduate School of Medicine
Masao Matsuhashi, MD, PhD – Kyoto University Graduate School of Medicine
Akio Ikeda, MD, PhD – Kyoto University Graduate School of Medicine
Riki Matsumoto, MD, PhD – Kyoto University Graduate School of Medicine.
Rationale:
Rationale: The spectral slope, defined as the decay rate of power across frequencies in intracranial EEG (iEEG), is considered to reflect cortical excitation/inhibition (E/I) balance. We investigated whether spectral slope can serve as a biomarker for E/I dynamics and help localize the seizure onset zones (SOZ) in focal epilepsy.Methods:
Methods: We analyzed spike-free, wakeful 10-second iEEG segments from 13 patients with drug-resistant focal epilepsy (870 analyzable electrodes). The spectral slope was calculated from the power spectral density in the 30-45 Hz range for each electrode using multitaper methods, based on prior work implicating this range in the local E/I balance (Gao, et al. Neuroimage. 2017). Slope values were compared between SOZ and non-SOZ electrodes using a linear mixed-effects model accounting for anatomical location and patient-level variability. Correlations were further evaluated between the slope values and interictal epileptiform discharge characteristics, including spike and post-spike slow (PSS) amplitudes, as well as associated high-frequency activities (HFAs): low gamma (LG, 30-50 Hz), ripple (R, 80-200 Hz), and fast ripple (FR, 200-300 Hz) bands. Diagnostic performance was evaluated using leave-one-patient-out cross-validation.
Results:
Results: SOZ electrodes demonstrated significantly flatter (i.e., higher) spectral slopes compared with non-SOZ electrodes (median -2.58 vs. -2.90; p< 0.001). This association remained significant after adjusting for anatomical location and patient-level variability (β=0.27, 95% CI 0.03–0.51, p=0.029). The spectral slope positively correlated with PSS amplitude (r=0.25, p=0.046) and showed a trend-level positive correlation with spike amplitude (r=0.24, p=0.051). It also positively correlated with spike-related HFAs across all bands (LG: r=0.48, R: r=0.39, FR: r=0.47; p< 0.001), whereas showing a negative correlation with PSS-related FR power (r=0.29, p=0.028), further supporting its role as an E/I balance marker. As a biomarker for SOZ localization, the spectral slope achieved an area under the curve of 0.67 overall, which improved to 0.81 in patients with mesial temporal lobe epilepsy (MTLE). In the MTLE subgroup, the slope yielded a high negative predictive value (NPV) of 98% for excluding non-SOZ electrodes.