Page Banner

United States Department of Agriculture

Agricultural Research Service

Research Project: SORTING AGRICULTURAL MATERIALS FOR DEFECTS USING IMAGING AND PHYSICAL METHODS Title: Automatic image analysis and spot classification for detection of pathogenic Escherichia coli on glass slide DNA microarrays

Authors
item Haff, Ronald
item Quiñones, Beatriz
item Swimley, Michelle
item Toyofuku, Natsuko

Submitted to: Computers and Electronics in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 25, 2010
Publication Date: February 26, 2010
Citation: Haff, R.P., Quinones, B., Swimley, M.S., Toyofuku, N. 2010. Automatic image analysis and spot classification for detection of pathogenic Escherichia coli on glass slide DNA microarrays. Computers and Electronics in Agriculture. 71(2):163-169.

Interpretive Summary: The present manuscript discusses a computer algorithm was created to inspect scanned images from DNA microarray slides. These slides are developed to rapidly detect and genotype E. Coli O157 virulent strains. The algorithm computes the locations of signal spots and background pixels and defines a plane that best separates the two as a decision boundary. The algorithm was tested on 1534 potential spot locations which were visually classified depending on the strength of the signal to the background noise (Signal to Noise Ratio or SNR). Three other standard measures of SNRs (SSR, SBR, and SSDR) were also performed for each potential spot location. The results were compared with visual classifications of each spot. The number of errors were computed for each of the four measures. SSR and SSDR, which depend on pixel intensity standard deviations, performed poorly with high false positive results, while the current algorithm and SBR, which do not use standard deviations, performed much better. Overall error rates were 1.4% for the reported algorithm, 2.0% for SBR, 14.2% for SSDR, and 16.8% for SSR.

Technical Abstract: A computer algorithm was created to inspect scanned images from DNA microarray slides developed to rapidly detect and genotype E. Coli O157 virulent strains. The algorithm computes centroid locations for signal and background pixels in RGB space and defines a plane perpendicular to the line connecting them as a decision boundary. The algorithm was tested on 1534 potential spot locations which were visually classified depending on the strength of the signal. Three other standard measures of SNR (SSR, SBR, and SSDR) were also performed for each potential spot location. The number of errors, when comparing results with visual classifications, were computed for each of the four measures. SSR and SSDR, which depend on pixel intensity standard deviations, performed poorly with high false positive results, while the current algorithm and SBR, which do not use standard deviations, performed much better. Overall error rates were 1.4% for the reported algorithm, 2.0% for SBR, 14.2% for SSDR, and 16.8% for SSR.

Last Modified: 10/22/2014