bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2025, Volume and Issue: unknown
Published: May 4, 2025
Abstract Sensory neural coding, the brain’s process of transforming inputs into informative patterns activity, generates complex and multiplexed codes which are hard to interpret. Although decoding methods have facilitated interpretation these codes, specific features activity that generalize across individuals, along with their precise timing, largely remain elusive. To address this gap, we investigated potential interpretable time-series in magnetoencephalography (MEG) for visual stimulus attributes (spatial frequency orientation) generalizability individuals. We extracted a comprehensive set highly from 18 subjects engaged task performed time-resolved analysis. Our findings revealed particular features, especially those capturing rapid changes within first 200 milliseconds presentation, yielded high accuracy. Notably, early, transient exhibited robust cross-subject generalizability, suggesting shared coding mechanism during initial processing inputs. Furthermore, outperformed previously successful electroencephalography (EEG) wavelet when applied MEG data. While within-subject demonstrated sustained above-chance performance, generalization diminished after milliseconds, indicating more individualized at later stages. results underscore importance systematic, data-driven evaluation signals elucidating developing transparent generalizable Brain-Computer Interface (BCI) systems capitalize on reliable signatures.
Language: Английский