Tetrahedral feature map for analyzing meditative states in prenatal EEG

Purpose: Electroencephalography (EEG) is a widely used non-invasive method to explore brain activity and cognitive states, including the effects of meditation. The potential of prenatal EEG to understand the neural mechanisms underlying maternal brain activity makes it an exciting field of study. However, research on prenatal EEG during different stages of meditation is currently lacking. Additionally, feature map generation and selection remain tedious and time-intensive, requiring careful identification of relevant features to capture both linear and non-linear insights from EEG data.

Methods: This paper proposes a Tetra Feature Map (TFM) for analyzing prenatal EEG signals, enabling effective differentiation of meditative states. TFM integrates four key aspects of the data: frequency bands, time domains, statistical features, and entropy features along with geometrical features facilitating a comprehensive analysis of dynamic EEG behavior. The PEM-43 dataset was used to evaluate TFM, tested across 33 machine learning classifiers to assess its performance in categorizing different meditative states.

Results: The study revealed that TFM achieved a test accuracy of 92%. Furthermore, TFM was compared with two other feature sets, the Statistical Feature Map (SFM) and the New Feature Map (NFM), which achieved classification accuracies of 88% and 89%, respectively. TFM demonstrated superior performance in classification accuracy.

Conclusion: These findings indicate that TFM can effectively analyze prenatal EEG data with a high classification success rate. This marks a significant advancement in prenatal EEG research and its application in brain state classification, particularly in the context of meditation.