Novel image processing method study for a label-free optical biosensor

Abstract

ABSTRACT Optical biosensor is generally divided into labeled type and label-free type, the former mainly contains fluorescence-labeled method and radioactive-labeled method, while fluorescence-labeled method is more mature in the application.The m ainly image processing methods of fluorescent-labeled biosensor includes smooth filtering, artificial gridding andconstant threshold. Since some fluorescent molecules may influence the biological reaction, label-free methods havebeen the main developing direction of optical biosensors nowadays. The using of wider field of view and larger angle of incidence light path which could effectively improve the sensitivity of the label-free biosensor also brought moredifficulties in image processing, comparing with the fluorescent-labeled biosensor. Otsu’s method is widely applied inmachine vision, etc, which choo se the threshold to minimize the intraclass variance of the thresholded black and white pixels. It’s capacity-constrained with the asymmetrical distribution of images as a global threshold segmentation. Inorder to solve the irregularity of light intensity on the transducer, we improved the algorithm. In this paper, we present anew image processing algorithm based on a reflectance modulation biosensor platform, which mainly comprises the design of sliding normalization algorithm for image rectification and utilizing the improved otsu’s method for imagesegmentation, in order to implement automatic recognition of target areas. Finally we used adaptive gridding methodextracting the target parameters for analysis. Those methods could improve the efficiency of image processing, reduce human intervention, enhance the reliability of experiments and laid the foundation for the realization of high throughput of label-free optical biosensors. Keywords: label-free optical biosensor; image processing; otsu’s method; reflectance modulation