Pattern Recognition of Neurotransmitters: Complexity Reduction for Serotonin and Dopamine.

Area of Science:

• Electrochemistry
• Neuroscience
• Machine Learning

Background:

• Simultaneous detection of neurotransmitters serotonin (SE) and dopamine (DA) is crucial for understanding neurological functions.
• Conventional methods often face challenges in deconvolving complex signals from mixtures.

Purpose of the Study:

• To develop and validate a method for simultaneous detection and prediction of SE and DA concentrations in vitro.
• To explore feature reduction techniques to simplify and improve the accuracy of neurotransmitter estimation.

Main Methods:

• Utilized conventional glassy carbon electrodes (CGCEs) and differential pulse voltammetry (DPV) for signal acquisition.
• Employed Principal Component Analysis with Gaussian Process Regression (PCA-GPR) and Partial Least Squares with Gaussian Process Regression (PLS-GPR) for signal deconvolution.
• Investigated reduced feature subsets within oxidation potential windows to optimize estimation.

Main Results:

• PCA-GPR achieved average testing accuracies of 87.6% (DA), 88.1% (SE), and 96.7% (DA-SE).
• PLS-GPR yielded average testing accuracies of 87.3% (DA), 83.8% (SE), and 95.1% (DA-SE).
• Feature reduction improved DA-SE estimation accuracy to 97.4%.

Conclusions:

• The developed methods, particularly with feature reduction, offer high accuracy for simultaneous SE and DA detection and prediction.
• Reduced complexity in feature selection enhances prediction accuracy, paving the way for practical clinical applications.
• This approach holds promise for improving neurotransmitter measurements in clinical settings, such as deep brain stimulation monitoring.