Performing Department
(N/A)
Non Technical Summary
Foodborne illnesses, especially those caused by Salmonella bacteria, continue to be a major public health issue, impacting millions of Americans annually. These infections not only cause personal suffering, but they also have a significant impact on the economy through medical expenses and reduced productivity. Ensuring the safety of our food supply, particularly in widely consumed products such as onions, is of utmost importance for the health and well-being of the public. Conventional techniques for identifying harmful bacteria in food can be quite time-consuming, often taking several days to produce results. This delay increases the risk of contaminated products reaching consumers before the issue is detected. This project is focused on tackling a pressing problem by creating a quicker and more effective method for identifying Salmonella contamination in onions. It utilizes state-of-the-art technology that combines microscopic imaging with artificial intelligence (AI). In order to accomplish these objectives, we will create and deploy an advanced AI system that can analyze images from phase-contrast microscopes. This system will be able to detect the presence of bacteria in onion samples. Through the utilization of deep learning models and a vast dataset of labeled bacterial images, our objective is to develop a system capable of swiftly and accurately detecting Salmonella. This entails nurturing bacterial samples, capturing intricate microscopic images, and utilizing these images to train our AI models in identifying Salmonella microcolonies. We will also participate in a wide range of educational activities, such as workshops and training sessions, to share this technology with food safety professionals, researchers, and industry stakeholders. This project has a significant potential impact as it aims to improve food safety protocols, decrease the occurrence of foodborne illnesses, and establish a model for incorporating AI into food safety practices. This will lead to a greater safety in food consumption, decreased healthcare expenses, and a stronger public health infrastructure. The long-term goal, this approach has the potential to be applied to various food types and bacteria, which could greatly transform the way we guarantee food safety throughout the entire food supply chain.
Animal Health Component
50%
Research Effort Categories
Basic
20%
Applied
50%
Developmental
30%
Goals / Objectives
The major goal of this project is to develop and implement an innovative artificial intelligence (AI) system for rapid detection of Salmonella contamination in onions using microscopic imaging. This project aims to enhance food safety measures by enabling quick, accurate, and cost-effective identification of foodborne pathogens, specifically targeting Salmonella in onion samples. The overarching purpose is to significantly reduce the time required for pathogen detection compared to traditional culture-based methods, thereby improving food safety protocols and potentially preventing foodborne illness outbreaks.To achieve this goal, the project has the following specific objectives:Construct a large-scale image dataset with labels to capture Salmonella Thompson cells in a model food system. This objective involves: a) Cultivating Salmonella Thompson and E. coli K-12 strains under various conditions. b) Preparing and imaging onion samples inoculated with these bacterial strains. c) Capturing high-resolution microscopic images of bacterial microcolonies at different growth stages. d) Annotating and categorizing the images to create a comprehensive dataset for AI model training. Attainability: This objective is achievable within the first year of the project, utilizing 0.7 FTE of the post-doctoral researcher (1.0 FTE in Year 1) and 0.022 FTE of the faculty member. The Olympus IX70 Inverted Phase Contrast DIC Fluorescence Microscope and Seward Stomacher 400 Circulator Blender will be crucial for sample preparation and imaging. The team aims to collect and annotate approximately 1,800 images (1,400 Salmonella, 400 E. coli) within this timeframe. Administrative support (0.05 FTE) will assist with procurement and scheduling. Part of PI's effort 0.022 FTE includes overseeing dataset creation and team coordination. Departmental support (0.05 FTE) will assist with procurement of supplies and scheduling of imaging sessions.Develop real-time, efficient, and automated deep learning models for early detection ofSalmonella in foods. This objective includes: a) Designing and implementing a CNNarchitecture, such as YOLOv4, for bacterial microcolony detection and classification. b) Training the AI model using the curated image dataset to recognize and differentiate Salmonella from other bacterial species. c) Optimizing the model for high accuracy, sensitivity, and specificity in detecting Salmonella contamination. d) Validating the model's performance against standard laboratory methods for Salmonella detection.Attainability: This objective will be primarily addressed in Year 2, utilizing 0.025FTE of the faculty member and 0.3 FTE of the graduate student. The high-performance computing cluster at the university will support the intensive computational requirements for AI model development and training. The team plans to develop and optimize the CNN model within approximately 6-8 months.Administrative support (.025 out of0.05 FTE) will assist with data management and reporting.Part of PI's effort 0.025 FTE in Year 2 includes supervising model development and validation. Departmental support (0.025 FTE (half of the 0.05 FTE allocated)) will assist with data management and progress reporting.Conduct educational transfer activities to promote the incorporation of AI in food science materials and extension programs. This objective involves: a) Organizing and delivering workshops on AI applications in food safety for students, faculty, and industry professionals. b) Developing training materials on microscopic imaging techniques, data curation, and AI model development for food safety applications. c) Providing hands-on experience with the developed AI system to participants. d) Assessing the effectiveness of the educational programs through pre- and post-workshop evaluations.Attainability: This objective will be spread across both years, using approximately 0.0125 FTE out of 0.05 FTE of the faculty member each year. Two workshops are planned: one in the latter part of Year 1 and another in Year 2. The post-doctoral researcher (0.1 FTE) will assist in developing materials in Year 1, while the graduate student (0.1 FTE) will help with the Year 2 workshop. The team aims to train 5-10 participants over the training sessions.Departmental Administrative support (.025 out of0.05 FTE) will handle logistics and participant coordination.Disseminate these technologies to relevant food industry stakeholders and agencies to encourage the integration of AI-enabled imaging sensors for predictive food safety monitoring and smart decision making. This objective includes: a) Presenting research findings at relevant scientific conferences and industry events. b) Publishing results in peer-reviewed journals and open-access platforms. c) Creating a dedicated webpage to showcase the project's progress and outcomes. d) Engaging with food safety regulators and industry leaders to promote adoption of the developed technology.Attainability: This objective will be ongoing throughout the project, intensifying in the second year. It will utilize approximately 0.0125 FTE out of 0.05FTE of the faculty member each year. The team plans to submit at least one peer-reviewed publication and present at one major conference. The post-doc and graduate students (0.1 FTE) will assist in publications.DepartmentalAdministrative supportwill assist with travel arrangements, publication submissions, and stakeholder communication.
Project Methods
The project will be conducted through the following approaches:1. Preparation of Bacterial InoculumandDevelopment of Image Dataset:Cultivation of Salmonella Thompson and E. coli K-12 strains under controlled laboratory conditions.Preparation of onion samples spiked with known concentrations of bacterial cultures.Utilization of the Olympus IX70 Inverted Phase Contrast DIC Fluorescence Microscope to capture high-resolution images of bacterial microcolonies at various growth stages.Collection of approximately 1,800 images (1,400 Salmonella, 400 E. coli) over multiple experimental runs.Annotation and labeling of images to create a comprehensive dataset for AI model training.2. Deep Learning Model Development:Implementation of CNNarchitecture, specifically YOLOv4, using Python and the PyTorch framework.Division of the image dataset into training (60%), validation (10%), and test (30%) sets.Training of the CNN model using the annotated image dataset, with iterative refinement based on validation set performance.Optimization of model hyperparameters to maximize detection accuracy and minimize false positives.Validation of the final model using the held-out test set.3. Experimental Validation:Application of the optimized AI model to detect Salmonella in mixed culture samples containing both Salmonella and E. coli.Comparison of AI model predictions with results from standard selective agar plating methods.Statistical analysis of model performance, including calculation of sensitivity, specificity, and overall accuracy.4. Technology Transfer and Educational Activities:Development of curriculum materials on AI applications in food safety.Organization of two hands-on workshops (one per year) for students, faculty, and industry professionals.Incorporation of practical demonstrations using the developed AI system.Dissemination of Results:Preparation and submission of research findings to peer-reviewed journals.Presentation of results at relevant scientific conferences.Efforts to Cause Change:Presentations at conferences, publications in journals, and webinars to share research findings.Development of training materials and workshops to educate food safety professionals and students.Training workshops to demonstrate the use of AI-based detection methods in real-world settings.Pilot programs to test the technology in food industry operations.Implementation of AI detection methods in food safety protocols to reduce foodborne illnesses.The performance of the AI model will be evaluated using standard metrics in machine learning and food safety:Accuracy: Percentage of correct predictions (both positive and negative) among the total number of cases examined.Sensitivity: Ability of the model to correctly identify positive samples (true positive rate).Specificity: Ability of the model to correctly identify negative samples (true negative rate).F1 Score: Harmonic mean of precision and recall, providing a single score that balances both metrics.Evaluation Plan:Continuous Monitoring Milestones:Reduction in time required for Salmonella detection in participating facilities.Development of AI model with >90% accuracy in Salmonella detection.Conduct two workshopsImplementation of AI system in at least one food processing facility for trial use.Publication of at least one peer-reviewed article and presentation at one conference.Summative EvaluationsEvaluation of dataset diversity, model performance, and educational workshop effectiveness.Model performance metrics (accuracy, sensitivity, specificity, F1 score).Pre- and post-workshop surveys to assess knowledge gain.Publication metrics (citations, downloads) and conference feedback.