EVALUATION OF PH FIELD-EFFECT TRANSISTOR MEASUREMENT SIGNALS BY NEURAL NETWORKS

A method is described which allows to evaluate analyte concentrations from flow-injection analysis signals of a pH field effect transistor (pH-FET) detector independently of the buffer capacity of the probe. The enzyme penicillin-G amidase was employed as the biocompound to monitor penicillin concentrations. Measurements have been carried out by using samples with various penicillin concentrations (penicillin-G: 1.5, 2.5, 5.0, 7.5 and 10.0 g/l) as well as various buffer ion concentrations (phosphate: 5, 15, 20, 25, 35, 45 and 50 mM). Depending on the penicillin concentration as well as the buffer ion concentration the peak shaped measurement signals did not only change their height but also their contour. Using the standard evaluation methods (peak height, integral or width) no unambiguous relation to the considered concentration could be found. In contrast to this, neural networks have been found to be able to evaluate the different contours. The relative errors were 4.7% for the determination of the penicillin concentration and 4.9% for the buffer ion concentration. The possibility of broadening the scope of this method to other pH-FET systems is discussed.