Lead Inventors:
Shih-Fu Chang, Ph. D.; Namwon Kim; Zhenguo Li
Automated Pattern Recognition Techniques in Stain Analysis Overcomes Manual Inspection Risks
Stains formed by the drying of a fluid on a solid substrate exhibit a variety of reproducible complex patterns and features determined by the fluid composition and substrate chemistry. These features can be exploited to identify the composition of a stain formed by an unknown substance. Identification of stains based on manual inspection is tedious, time-consuming, and subject to human bias. The use of automated pattern recognition techniques in stain analysis affords the potential to both accelerate and increase the accuracy of stain composition identification.
Automated Analysis for Identifying Liquid and Substrate Chemistry of Unknown Stains Improves Speed and Accuracy
The technology is an automated method for accurately identifying the liquid and substrate chemistry of an unknown stain using machine learning techniques to analyze features such as color and local binary patterns (LBPs) obtained from photographic images of the stain. Supervised and unsupervised pattern recognition algorithms are employed to respectively classify stains based upon existing training data and cluster stains into groups with similar characteristics and components. These algorithms were trained and tested using two comprehensive collections of microscopic stain images from a range of consumable and biological fluids.
Applications:
• The technology can be employed to improve the speed and accuracy of stain composition identification in biological testing, medical diagnosis, and forensic analysis contexts.
Advantages:
• The technology can perform stain identification rapidly and without lengthy chemical analysis.
• Specific chemical characteristics can be accurately identified using the technology, for example, the presence or absence of specific compounds, salt composition, and pH.
• Multiple substances can be analyzed in parallel with the technology by increasing processing throughput with a micro-arrayer.
• The technology is fully automated and requires no manual human inspection of the stains being analyzed.
Patent Status: Patent Pending
Licensing Status: Available for Licensing or Sponsored Research Support
Publications:
Identification of fluid and substrate checmistry based on automatic pattern recognition of stains, N. Kim, Z. Li, C. Hirth, F. Zenhausern, S.F. Chang, D. Attinger, Anal. Methods, 2012.