Classification of bacterial contamination using image processing and distributed computing.

Area of Science:

• Microbiology
• Computer Science
• Food Science

Background:

• Foodborne illness outbreaks pose significant risks to public health and the food industry.
• Automated microbial detection and classification are crucial for preventing outbreaks and ensuring food supply safety.
• Label-free identification of foodborne pathogens using colony scatter patterns is a promising but computationally intensive technique.

Purpose of the Study:

• To accelerate the feature extraction process for colony scatter patterns using computer clusters.
• To analyze the contribution of different scatter-based features to bacterial classification accuracy.
• To identify an optimal subset of features for rapid and accurate bacterial classification.

Main Methods:

• Utilized computer clusters to expedite feature extraction from 1000 scatter patterns of ten bacterial strains.
• Computed Zernike moments, Chebyshev moments, and Haralick texture features from light-scatter patterns.
• Employed Fisher's discriminant analysis for feature selection, followed by a support-vector machine (SVM) classifier with a linear kernel.

Main Results:

• Identified a small subset of scatter-based features that significantly improved classification accuracy and execution speed.
• Demonstrated the effectiveness of distributed computing in analyzing scatter patterns and extracting relevant features.
• Achieved desired results in classification accuracy and speed through extensive feature testing and selection.

Conclusions:

• Distributed computing offers a feasible approach for large-scale deployment of light scatter-based bacterial classification.
• Optimized feature selection is critical for balancing classification accuracy and computational efficiency.
• This methodology enhances the potential for real-time monitoring and control in food safety applications.