Image analysis for multianalyte xerogel based sensor arrays

This paper describes an image analysis neural network that is used to determine analyte concentrations in xerogel-based sensor arrays. Xerogel-based sensor elements developed using sol-gel processing techniques have been used for recognition elements in many areas including the biomedical, industrial and environmental fields. These arrays are composed of many individual pin-printed xerogel "spots" each which can be selective to a different analyte or be formed so that multiple spots correspond to a single analyte, but with different sensitivities. We analyze an array of spots printed on a microscopic glass slide with diameters ranging from 10mum to 500mum. We describe a multistage neural network that correctly identifies the analyte concentration for a prototype 10 times 10 array of sensor spots, and discuss how this can be scaled to handle arrays for multiple analytes