We propose a new nonlinear matching measure for automatic analysis of the on-off type DNA microarray images in which the hybridized spots are detected by a template-matching method. The proposed measure is obtained by binary thresholding over the entire template region and taking the number of white pixels inside the spotted area. This measure is compared with the normalized covariance method in terms of classifying the ability to successfully locate markers. The proposed measure was evaluated for scanned images of human papillomavirus (HPV) DNA microarrays where locating markers is a critical issue because of the small number of spots. The targeting spots of HPV DNA chips are designed for genotyping twenty-two types of the human papillomavirus. The proposed measure is proven to give a more discriminative response, reducing the missed cases of successful marker location. The locating accuracy of the proposed method is also shown to have the same performance as that of the normalized covariance.
(c) 2004 Society of Photo-Optical Instrumentation Engineers.