Sazonova Nadezhda, Njagi John I, Marchese Zachary S, Ball Michael S, Andreescu Silvana, Schuckers Stephanie
Department of Electrical and Computer Engineering, Clarkson University, Potsdam, NY 13676, USA.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:3493-6. doi: 10.1109/IEMBS.2009.5334575.
Neurotransmitters (NTs) are substances in the brain which are responsible for the transmission of neurological impulses. Changes in their concentrations are associated with numerous behavioral and physiological processes and neurological disorders. As opposed to the traditional chromatographic and capillary electrophoresis, using electrochemical sensors is a fast and inexpensive way to determine concentrations of NTs. In this study we measure the combination of dopamine (DA) and serotonin (SE) with glassy carbon electrodes and differential pulse voltammetry. The major challenge using this method is to differentiate between different NTs, since the signal obtained from the electrode represents the interactive effect of both NTs present. We address this problem through methods of pattern recognition which relate the voltammetric measurements provided by the sensor to the concentration of individual NTs. Two methods of pattern recognition were applied (PCR and PLS-regression). The best rates of correct classification for the validation sets ranged at 42-62% (DA) and 33-50% (SE). When the ranges for correct prediction were extended to include one level above and below the true concentration level, the rates values ranged at 81-91% (DA) and 91-100%(SE). These findings suggest that pattern recognition can be used to model the interaction between different neurotransmitters to predict actual concentrations of neurotransmitters using voltammetry.
神经递质是大脑中的物质，负责神经冲动的传递。它们浓度的变化与众多行为、生理过程及神经紊乱相关。与传统色谱法和毛细管电泳法不同，使用电化学传感器是测定神经递质浓度的一种快速且经济的方法。在本研究中，我们使用玻碳电极和差分脉冲伏安法测量多巴胺（DA）和血清素（SE）的组合。使用这种方法的主要挑战在于区分不同的神经递质，因为从电极获得的信号代表了存在的两种神经递质的相互作用效果。我们通过模式识别方法解决这个问题，该方法将传感器提供的伏安测量与单个神经递质的浓度相关联。应用了两种模式识别方法（主成分回归和偏最小二乘回归）。验证集的最佳正确分类率在42 - 62%（DA）和33 - 50%（SE）范围内。当将正确预测范围扩展到包括真实浓度水平上下各一个水平时，率值在81 - 91%（DA）和91 - 100%（SE）范围内。这些发现表明，模式识别可用于对不同神经递质之间的相互作用进行建模，以使用伏安法预测神经递质的实际浓度。
