|Huang, J, - UNIVERSITY OF ARKANSAS|
|Li, Y. - UNIVERSITY OF ARKANSAS|
|Slavik, M. - UNIVERSITY OF ARKANSAS|
|Tao, Y. - UNIVERSITY OF ARKANSAS|
Submitted to: American Society of Agricultural Engineers Transactions
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 11, 1999
Publication Date: N/A
Interpretive Summary: Poultry products are frequently contaminated with salmonella bacteria. This public health problem has led to a requirement for rapid tests which can detect these bacteria within the time it takes to process and pack the poultry products. Standard testing methods involve growing and counting bacteria which takes several days and is labor intensive. This project describes the use of a computerized microscope and video camera to detect and count bacteria present in chicken carcass wash water. The results of this method were consistent with the standard testing methods only when more than 1000 bacteria were present in the sample. The new method takes less than 4 hours, and requires less labor, however, there is an initial investment for equipment and higher material costs. Further automation of this method, using an automatic moving and focusing stage, may improve the sensitivity.
Technical Abstract: Rapid detection of bacteria on poultry carcasses is desirable for the poultry industry to enhance food safety of poultry products. This research focused on image analysis with fluorescent microscopy for identification and enumeration of salmonella typhimurium on sample slides of poultry carcass wash water which were prepared using fluorescent antibodies and immuno-magnetic beads. The criteria of morphological and optical characteristics of S. typhimurium cells, including area, aspect ratio, diameter, major and minor axis, maximum and minimum radii, perimeter, radius ratio, length and width, and intensity, were developed. An algorithm that includes channel extracting, median filtering, image sharpening, image dilation and erosion, image flattening, and watershed filtering was selected using Image-Pro (Registered trademark) Plus software to analyze the acquired images. The algorithm was trained with 110 slide samples. The regression analysis showed that the relationship between the automatic image counting and the microbial plate counting was linear with a correlation coefficient of 0.88. The detection limit of this method was 1 x 10**4 cells/mL, and the time needed was less than 4 h.