Comparison between different similarity measure functions for optimal clustering AEPs Independent Components

Area of Science:

• Neuroscience
• Signal Processing
• Auditory Neuroscience

Background:

• Independent Component Analysis (ICA) is vital for neurophysiological data analysis, but selecting relevant independent components (ICs) with neurological meaning remains challenging.
• Standard ICA typically yields as many ICs as EEG channels, necessitating objective methods for identifying components related to specific stimuli.
• Auditory Evoked Potentials (AEPs) are key signals in response to auditory stimuli, characterized by repetition and time-locking with the stimulus.

Purpose of the Study:

• To present an updated, objective procedure for selecting relevant ICs from neurophysiological data, specifically focusing on Auditory Evoked Potentials (AEPs).
• To evaluate the effectiveness of Mutual Information (MI) and cluster analysis in identifying neurologically meaningful ICs.
• To determine optimal clustering parameters, including similarity functions and cluster numbers, for accurate AEPs IC identification.

Main Methods:

• An updated procedure combining Mutual Information (MI) and cluster analysis for objective IC selection was developed.
• Four different similarity functions and three inter/intra-cluster quality criteria were evaluated to find optimal cluster numbers.
• The methods were tested on both synthetic AEPs and real-world data from normal hearing children.

Main Results:

• The study identified optimal parameters for clustering AEPs ICs, yielding the best results across both synthetic and real datasets.
• The optimal number of clusters was determined to be 8.
• The Euclidean link-clustering average similarity function proved most effective.

Conclusions:

• The refined MI and cluster analysis approach provides an objective method for selecting AEPs-related ICs from neurophysiological data.
• The identified optimal clustering parameters (8 clusters, Euclidean link-clustering average) enhance the accuracy of auditory response identification.
• This method contributes to a more reliable interpretation of neurophysiological signals in auditory research.