A Machine Learning Approach for Simultaneous Electrochemical Detection of Dopamine and Serotonin in an Optimized Carbon Thread-Based Miniaturized Device

To further optimize output from electrochemical sensing technology and minimize human intervention, machine learning (ML) models are capable of imparting data-driven predictions. This work focuses on developing a miniaturized EC sensing platform for the simultaneous detection of neurotransmitters, such as dopamine (DA) and serotonin or 5-hydroxytryptamine (5-HT). A modified carbon thread-based miniaturized device (CTMD) was developed using a CO2 laser scriber to detect dopamine and serotonin. The devices showed a linear range for DA and 5-HT as 0.5–150
$\mu $
M and 0.5–200
$\mu $
M, respectively. The limit of detection (LoD) and limit of quantification (LoQ) for DA and 5-HT were 0.25
$\mu $
M and 0.76
$\mu $
M (
${R}^{{2}} =0.99$
and
${N} =3$
), and 0.22
$\mu $
M and 0.78
$\mu $
M (
${R}^{{2}} =0.98$
and N =3), respectively. Furthermore, a real sample analysis in blood serum was performed, demonstrating good recovery and selectivity. Finally, ML prediction was performed over 100% of the generated data through analytical methods, whereas 80% of the data were used for training purposes, and 20% of the data were used for testing purposes. Various ML regression models, such as linear regression, decision tree, k-NN, support vector regression, gradient, adaptive boosting, and random forest, were used to obtain the best accurate prediction, low error values, and increased
${R}^{{2}}$
scores. Apart from support vector and linear regression, all other techniques provided the best
${R}^{{2}}$
scores of over 0.98 with low error values. Based on the obtained results, the fabricated device, including the ML approach, can effectively be leveraged in diagnostics devices.