New metric for optimizing Continuous Loop Averaging Deconvolution (CLAD) sequences under the 1/f noise model

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Continuous loop averaging deconvolution (CLAD) is crucial for analyzing auditory evoked potentials (AEPs) in rapid stimulation paradigms.
• Effective CLAD sequence design is essential to minimize noise interference in electroencephalogram (EEG) data.
• Existing noise gain factor (NGF) metrics show limitations, particularly with pink (1/f) noise.

Purpose of the Study:

• To develop and validate a novel metric for evaluating CLAD sequences, specifically addressing limitations with pink noise.
• To introduce a generalized NGF measurement for more accurate quantification of CLAD sequence performance.
• To demonstrate the utility of the new metric in optimizing CLAD sequence generation using genetic algorithms.

Main Methods:

• Derived a new metric by incorporating a 1/f noise model into a time-continuous sequence framework.
• Tested the new metric on simulated EEG data and real CLAD EEG recordings to assess middle latency responses.
• Employed a genetic algorithm to generate and quantify optimized CLAD sequences using the new metric.

Main Results:

• The new metric demonstrated improved accuracy in measuring actual noise gains across different frequencies compared to the original NGF.
• The metric showed superior performance in various aspects, particularly in the presence of pink noise.
• Optimized CLAD sequences generated using the new metric and genetic algorithm exhibited enhanced performance.

Conclusions:

• The developed metric offers a generalized and more effective approach to quantifying CLAD sequence performance.
• This advancement facilitates the creation of optimized CLAD sequences for improved AEP analysis.
• The study supports the enhanced clinical application of the CLAD paradigm with tailored sequences.