201: Hyperspectral Imaging and Machine-Learning Approaches for Predicting Microbial Contamination on Fresh Produce

Introduction

Fresh produce is highly perishable in nature due to their susceptibility to microbial contamination and spoilage. Traditional methods for detecting microbial contamination are time-consuming and labor-intensive, whereas the food industries require rapid detection technologies to enhance food safety and reduce microbial spoilage. Hyperspectral imaging (HSI) is an emerging technology that integrates traditional imaging with spectroscopy, capturing both spatial and spectral information from fresh produce surfaces to predict microbial contamination. This study aims to develop predictive models that can estimate microbial contamination levels on fresh produce surfaces using a combination of HSI and machine learning.

Methods

Vegetable samples were collected and inoculated with known levels of microbial loads of E. coli, Listeria and yeast strains. Hyperspectral images were acquired in the visible to near-infrared (400–1000 nm) spectrum using a hyperspectral camera. Pre-processing techniques, including calibration, noise reduction, normalization, and identification of regions of interest, were applied to the spectral data. Furthermore, development of performance metrics has been developed to address the key wavelengths and machine learning algorithms, such as partial least squares regression (PLSR), support vector regression (SVR), artificial neural network (ANN), and convolutional neural network (CNN), were used to build predictive models correlating spectral features with microbial load. The model performance was evaluated using cross-validation and root mean square error (RMSE).

Results

The HSI-based models are expected to provide high accuracy in predicting microbial load, using machine learning models. Further validation of the models on independent samples could provide confirmation of their robustness and reliability across various vegetable types and contamination levels.

Significance

The findings will pave the way for developing a continuous, non-destructive, and rapid system for detecting microbial contamination. This innovative approach has the potential to revolutionize food safety practices by enabling real-time monitoring in food processing environments, ultimately reducing the risk of foodborne diseases. The application of this technology could enhance the efficiency of quality control processes while contributing to improved public health outcomes.

Authors: Rishabh Goyal, Kang Huang