In typical complementary DNA microarray experiments, two fluorescently labeled RNAs are hybridized to an array of cDNA probes on a glass slide, and their relative gene expression intensities, ratios or both are measured in order to quantify the gene expression level relative to its reference sample. Our early reports discussed a systematic data extraction algorithm in which a unique method of extracting gene expression intensities and ratios along with an adaptive ratio confidence interval, measurement qualities of gene expression ratios and intensities were presented. In many methods of gene expression data analysis, only expression ratios or normalized intensities are employed because of insufficient assessment at the individual data points (clones). Common practice dictates that data derived with poor measurement quality—such as expression ratios derived from weak reference expression levels or noise-corrupted measurements—shall not be used in the analysis. We present an automatic decision-making process for various algorithms, such as gene expression clustering and classification, in which gene expression ratios and intensities are chosen to participate in the analysis according to their measurement quality, expression signal-to-noise ratio relative to the reference channel and other parameters derived from cDNA microarray image analysis software.