Aperiodic 1/f noise drives ripple activity in humans

Area of Science:

• Neuroscience
• Computational Neuroscience
• Signal Processing

Background:

• Sharp-wave ripples (SWRs) are crucial for memory consolidation and are observed in rodent and human brains during sleep and wakefulness.
• Detecting SWRs across different brain states and regions, particularly in humans, presents significant challenges.
• Existing ripple detection methods may be susceptible to artifacts from background neural activity.

Purpose of the Study:

• To investigate the reliability of common sharp-wave ripple detection algorithms.
• To assess the contribution of background cortical activity (1/f^χ noise) to putative ripple events.
• To develop a method for estimating false positive rates in ripple detection.

Main Methods:

• Analysis of intracranial EEG data from three studies involving human participants during sleep and cognitive tasks.
• Evaluation of five common ripple detection algorithms for their sensitivity to noise.
• Simulation-based approach to quantify false positive rates and identify noise-resilient detection scenarios.

Main Results:

• An average of 77% of detected awake ripples in the medial temporal lobe, including the hippocampus, were identified as false positives.
• Putative ripples often represent noise modulated by region, brain state, and cognitive demand.
• Task-related modulations of 1/f^χ background activity can lead to spurious ripple detections.

Conclusions:

• Current ripple detection methods may overestimate ripple occurrence due to high false positive rates.
• 1/f^χ noise significantly impacts ripple detection, especially during cognitive engagement and across different brain states.
• A simulation-based approach is valuable for assessing detection algorithm performance and understanding cortical processing.