Development and Application of AI for Food Processing and Safety Regulations

Application Approaches and Methods

A number of research articles have been published that showcase real-world case studies of how machine learning is employed for the rapid detection of pathogens, preventing contamination incidents,¹⁶ or how the synergy between AI and blockchain technologies functions for enhancing traceability and transparency throughout the food supply chain.¹⁷ These studies provide great examples showing that the integration of these AI technologies ensures their accountability in assisting users to make quick and correct responses to both management and technical food safety issues.

From these successful AI application case studies,^{12–15,18,19} it can also be surmised that the successful application of AI technologies to FSIS work will depend on the successful development and deployment of specific AI application approaches and methods that meet the needs of specific operation procedures and steps, types of products, inspection requirements, monitoring and control systems, and management purposes.^14,15,18 Keeping in mind that the purposes of AI application to food safety and inspection are to solve the most challenging and complex food safety and quality problems across various domains, it is essential to correctly determine what AI application approaches and methods should be chosen.

General AI application systems can be classified into these groups if they will be used for any of the following example tasks:

• Analyze data
• Understand patterns
• Automate repetitive tasks
• Extract meaningful insights from mega-datasets
• Enable the creation of personalized experiences by understanding user preferences and behavior as a recommendation system
• Adapt and learn from new data for improving performance over time
• Optimize resource allocation and utilization
• Provide insights and recommendations
• Assist in decision-making
• Understand and generate human language
• Develop user-friendly interfaces and applications such as speech recognition, gesture control, and facial recognition
• Foster innovation by opening up new possibilities and avenues for research and development
• Optimize routes for delivery trucks
• Minimize fuel consumption and transportation costs.^9,14

Based on this information, current AI applications can be divided into ten major AI systems that will benefit USDA-FSIS and the FSIS-regulated food production industries in the future. The ten AI systems include:

1. Automated inspection and quality control
2. Predictive analytics for inspection planning
3. Food processing and supply chain monitoring and traceability
4. Natural language processing for compliance analysis
5. Early warning systems for food safety incidents
6. AI-assisted food safety and risk assessment
7. Dynamic regulatory compliance updates
8. Smart labeling and product information extraction
9. AI-enhanced training and skill development
10. IoT integration for real-time monitoring.

Each system is discussed in detail below.

AI System 1: Automated Inspection and Quality Control

AI-powered computer vision systems could be applied to automate safety inspection and quality control systems. AI algorithms can analyze visual data such as images or videos of the food items from cameras to identify defects, contamination, or irregularities along the entire process, from animal slaughtering to post-mortem processing, packaging, and delivery. AI systems have a high level of accuracy, which will help reduce the heavy reliance on manual inspection and ensure faster and better food safety and quality controls.²⁰ In this system, machine learning algorithms should be pre-trained on large datasets to recognize patterns associated with food safety and quality criteria.

The major components of this system may include:

1. Computer vision is used for visual inspection of food items, enabling the identification of product imperfections associated with both food safety and non-food safety issues. These imperfections can include fecal contamination, septicemia/toxemia, dead-on-arrival (DOA), cadaver, metastatic tumors, bruises, inflammatory process (IP), synovitis (hock joints), and other foreign objects or irregular shapes. Current computer vision tools encompass cameras with high-resolution imaging capabilities.
2. Machine learning algorithms could be integrated with AI-powered computer vision devices and pre-trained on diverse datasets to classify and predict food safety and quality attributes based on features extracted from images. Proprietary platforms provide frameworks for developing machine learning models for food inspection. Cloud-based machine learning services are available from major providers for scalable solutions.
3. Infrared or near-infrared spectroscopy is most often used in a laboratory setting to analyze the chemical composition of food and provide insight into freshness, ripeness, or contamination. Several companies offer spectroscopy solutions for food analysis. Portable spectrometers could be used by USDA employees and plant quality control personnel for onsite food safety inspections in industrial settings.
4. Smart sensor technologies can monitor parameters like temperature, humidity, foreign objects such as metal contaminants, and others to ensure food safety and quality. Temperature and humidity sensors are commercially available. IoT-based sensor systems could be provided for comprehensive environmental monitoring. Advanced food metal detectors could be used for automatic inspection of wet, hot, chilled, cooling, or metal film-packed products.

Integration of these technologies should allow for real-time monitoring and automated decision-making, streamlining the inspection and control processes. However, it is crucial to choose equipment based on specific application needs, scale, and integration requirements. Importantly, users should always consider factors like data privacy and regulatory compliance when implementing these technologies for food safety.

AI System 2: Predictive Analytics for Inspection Planning

This system uses machine learning algorithms to analyze historical data records on inspection outcomes, food safety incidents, and compliance trends or patterns. It could enable predictive analytics of the likelihood of key food safety issues for optimizing resource allocation and making strategic plans.^20,21

Some device and tools that could work with this AI system and are available on the current market may include:

1. Predictive analytics platforms.
2. Business chain visibility solutions could be used for business chain optimization, including predicting potential food safety issues in the food processing and supply chains.
3. Custom machine learning models could be built using popular frameworks and tailored to meet specific food safety challenges. Some models excel in deep learning and large-scale data processing, while others are better suited for traditional machine learning applications with smaller datasets.
4. IoT-based monitoring systems could be used to monitor environmental conditions, track product movement, and predict potential risks with the help of sensors. They may be suitable for monitoring good commercial practices, sanitation process standards, sanitation standard operation procedures (SSOPs), and food defense verification processes at food companies.
5. Data analytics software tools provide interactive visualizations, business intelligence capabilities, and options for chart types and visualizations. Data analytics software is designed for the collection, manipulation, integration, analysis, and presentation of business data and is not AI-powered, at present. Integration of data collection from other AI-powered devices will be required to build a better data analysis platform for automation of large food safety and quality data analysis in the future. Other analytics platforms could integrate AI technology and be directly used for analyzing large datasets, enhancing traceability, and monitoring capabilities.

These tools analyze data from various sources, including production records, supply chain information, and quality control data, and will be powered by AI technology for better food safety outcomes in the future.

AI System 3: Food Processing and Supply Chain Monitoring and Traceability

Current AI technologies can be applied to track and monitor the processing and movement of food products, providing real-time visibility into the food processing and supply chains to enhance traceability for food safety.²² This could facilitate a quicker response in the case of food contamination or recall. These AI-powered, automatic processes could be realized by utilizing technologies like radio frequency identification (RFID), sensors, and data analytics, as shown above. Some new devices and tools available for this purpose include:

1. Blockchain platforms are tamper-resistant, shared digital ledger systems used to record transactions and share information. Walmart's Food Traceability Initiative leverage blockchain is an example of such a system and is used for secure and transparent traceability across the corporation's food processing and supply chains.
2. Track-and-trace solutions include commercially available solutions such as a cloud-based supply chain planning solution that scales to accommodate business growth and integrates with other proprietary and third-party systems. It could provide end-to-end visibility and traceability in the food supply chain, as well as give executives and decision-makers insights to make data-driven decisions and drive strategic initiatives.
3. IoT-based monitoring systems include platforms for integrating sensors and devices throughout the supply chain to monitor conditions like temperature and humidity.
4. AI-powered data analytics integrate AI to analyze large datasets, enhancing traceability and monitoring capabilities.
5. RFID technology solutions help track and trace products throughout the supply chain.

Integration of AI technology with these tools allows for real-time monitoring, data analysis, and proactive identification of potential issues in the food processing chains, thereby enhancing the traceability of food safety compliance.

AI System 4: Natural Language Processing for Compliance Analysis

This system uses natural language processing (NLP) algorithms to analyze text-based data, including regulations, reports, and communications.²³ This can streamline the process of extracting insights from regulatory documents, thereby facilitating compliance analysis and ensuring adherence to food safety standards. This technology has been successfully applied for food safety inspection.

Several tools are available on the market:

1. NLP platforms.
2. Regulatory compliance solutions include proprietary platforms that use AI and NLP to analyze regulations, ensuring compliance in the food industry.
3. Custom NLP models are tailored to interpret food safety regulations and compliance documents.
4. Document management systems integrate AI-driven NLP for efficient analysis of large volumes of compliance-related documents.
5. Semantic analysis tools use semantic analysis, a subset of NLP, to understand the context and meaning within compliance documents.

These tools enhance the speed and accuracy of compliance analysis, helping businesses stay informed about regulatory requirements and ensuring adherence to food safety standards.

AI System 5: Early Warning Systems for Food Safety Incidents

Developing an AI-based early warning system for analyzing diverse data sources has been a scientific dream for food safety inspectors.²⁴ Such AI-powered early warning systems leverage diverse data sources to detect potential food safety incidents or outbreaks in time. Some equipment and tools used for this purpose include:

1. IoT sensors can generate real-time environmental data automatically.
2. Predictive analytics platforms use machine learning algorithms to predict potential food safety issues in food processing and supply plants.
3. Blockchain platforms enhance traceability, allowing for rapid identification of the sources of safety incidents.
4. Machine learning models can be trained to recognize patterns indicative of potential safety risks.
5. Data integration platforms integrate and analyze data from diverse sources, enabling the instant creation of comprehensive early warnings.

All of these tools enable real-time monitoring, analysis, and alerts, helping businesses respond swiftly to emerging food safety incidents and mitigating potential risks.

AI System 6: AI-Assisted Food Safety and Risk Assessment

Food safety risk assessment is a time-consuming process. AI-powered food safety risk assessment models may provide optimal supports for this purpose. These models could be used to analyze various risk factors, such as production methods, transportation conditions, and historical safety data, to prioritize inspections and interventions based on potential risks.^12,25 Some equipment and tools used for this purpose are available in the market and include:

1. Risk assessment platforms leverage AI to assess risks in the food supply chain, considering factors like contamination, transportation conditions, and regulatory compliance.
2. Predictive analytics tools utilize machine learning algorithms for predictive modeling, aiding in risk assessment based on historical data.
3. Custom machine learning models use frameworks tailored to specific risk factors that are relevant to food safety inspections.
4. IoT-based monitoring systems and data analytics platforms.

AI System 7: Dynamic Regulatory Compliance Updates

Current AI technologies can help users stay updated on changing regulations and compliance requirements through monitoring regulatory changes, interpreting updates, and automatically ensuring real-time compliance.²⁶ These technologies could be used in conjunction with evolving USDA-FSIS food safety policies, regulations, and standards to adjust internal processes and take correct corrective actions. The devices, tools, and approaches for these purposes include:

1. Regulatory compliance solutions leverage AI to track and interpret regulatory changes, providing real-time updates for compliance.
2. NLP and text analytics platforms can be applied to analyze and extract information from regulatory texts for automatic updates.
3. Customized compliance management systems use AI-driven platforms to track and adapt to regulatory changes.
4. Document management systems use AI-enhanced tools to facilitate the tracking and updating of compliance documentation.
5. Automated compliance monitoring platforms automate compliance monitoring, ensuring that organizations are promptly informed about changes in regulations.

These tools enable businesses to stay agile in response to evolving regulatory landscapes, ensuring ongoing compliance with food safety standards.

AI System 8: Smart Labeling and Product Information Extraction

AI technologies could be leveraged to automate the extraction of product information from labels and packaging,^20,23,27 which would enhance the accuracy in capturing details such as eight label features, nine food allergens and allergen warnings, production and expiration dates, nutritional statements, inspection legends, manufacturer's information, and so on. AI performs smart labeling and food product information extraction by utilizing technologies like computer vision and NLP. The required devices and tools for this purpose include:

1. Computer vision inspection systems are equipped with AI to read and interpret labels, ensuring accurate product information.
2. Optical character recognition tools can extract text from images, enabling the extraction of product information from labels.
3. NLP platforms can process textual information to extract relevant details from product descriptions.
4. Custom machine learning models recognize and extract specific product information.
5. Smart labeling solutions provide smart labeling technologies that embed digital information into product packaging, which can be extracted using AI.

These tools help automate the extraction of product information, ensuring accurate labeling and providing consumers with transparent and detailed information about food products.

AI System 9: AI-Enhanced Training and Skill Development

AI-driven training programs could be developed to enhance the skills of inspectors and staff by offering virtual simulations and personalized, interactive, and adaptive learning paths and experiences to them.^20,23,28 AI-assisted training modules can contribute to continuous improvement in food safety knowledge and inspection capabilities. For example:

1. Personalized learning paths could incorporate AI algorithms that analyze individual learner data to create customized training paths, addressing specific knowledge gaps and focusing on relevant topics in food safety.
2. Interactive simulations, such as virtual reality (VR) or augmented reality (AR) applications powered by AI, simulate real-world scenarios and allow trainees to practice food safety protocols in a risk-free environment.
3. Adaptive learning systems include AI-driven platforms that adapt the difficulty and content of training modules based on individual progress, ensuring that learners are appropriately challenged and engaged.
4. NLP can be integrated to facilitate interactive conversations, quizzes, and assessments, enhancing the overall learning experience.
5. Data analytics for performance monitoring could involve AI analytics tracking and evaluating learners' progress, providing insight into areas that may require additional focus or improvement.

AI-enhanced training not only improves the effectiveness of food safety education, but also allows for continuous monitoring and adaptation to changing needs, contributing to a more skilled and knowledgeable workforce in the food industry.

AI System 10: IoT Integration for Real-Time Monitoring

As discussed above, AI algorithms could be used to analyze data from IoT sensors to ensure optimal conditions for food safety throughout the food processing and supply chains. This technology has already been applied for environmental monitoring.¹⁴ Such synergy enhances food safety by ensuring optimal conditions and enabling timely interventions. Some equipment and tools used for this purpose and available in the market include:

1. IoT platforms help connect, manage, and analyze data from IoT devices used in food safety monitoring.
2. Wireless sensor networks monitor parameters like temperature, humidity, and more, in real time.
3. Edge computing devices enable local processing of data at the edge, reducing latency and enhancing real-time monitoring capabilities.
4. Data analytics platforms analyze and derive insights from the vast amount of data generated by IoT devices.
5. Blockchain technology platforms leverage IoT data to enhance traceability and transparency in real time.

These tools and equipment enable a seamless integration of IoT with AI, ensuring that food safety monitoring is not only real time but also data-driven, allowing for proactive decision-making and risk mitigation.

Benefits, Challenges, and Perspectives of AI Applications

It is predicted that AI technologies will excel human intelligence in many areas,^12,19 such as processing large datasets, calculating complex statistical models, evaluating both historical and instant trends, and identifying potential risk factors with unparalleled speed and accuracy. These capabilities will empower users to implement targeted food safety assessment; design risk mitigation strategies; and minimize the likelihood of foodborne illness, contamination incidents, and potential burst events. In addition, AI technologies may contribute to the documentation and determination of data relevance, as well as more accurate decision-making processes.

The benefits of AI application in food safety and inspection services may be summarized as:

1. Real-time visibility. IoT sensors collect and transmit data in real time, offering continuous visibility into the environmental conditions of food products. This ensures that any deviations from optimal conditions are promptly identified.
2. Proactive maintenance. AI algorithms can analyze data from IoT devices to predict potential equipment failures or deviations from standard conditions. Proactive maintenance measures can be implemented to prevent issues that might compromise food safety.
3. Data-driven decision-making. The combination of AI and IoT generates actionable insights from the vast amounts of data collected. This data-driven approach supports informed decision-making in managing the supply chain, ensuring compliance, and enhancing overall food safety.
4. Enhanced traceability. IoT devices, coupled with AI, contribute to an enhanced traceability system. Real-time monitoring allows for accurate tracking of the movement and storage conditions of food products, aiding in the rapid identification and containment of potential safety issues.

In the same way, the challenges of AI application in food safety and inspection services may be summarized as:

1. Data security and privacy. The increased connectivity of IoT devices raises concerns about data security and privacy. Implementing robust security measures is crucial to protect sensitive information related to food safety and supply chain operations.
2. Interoperability. Ensuring compatibility and seamless communication among diverse IoT devices can be challenging. Standardization efforts are essential to promote interoperability and facilitate the integration of AI with various IoT sensors.
3. Cost of implementation. The initial investment required for deploying IoT devices and integrating AI can be significant. However, the long-term benefits, including improved food safety and operational efficiency, often outweigh the initial costs.
4. Scalability. As the volume of data generated by IoT devices increases, greater scalability is ensured.

In conclusion, the application of AI in food safety and inspection services, particularly those offered by USDA-FSIS, has emerged as a game-changer for food safety and public health. This movement underscores the complexity and transformative potential of AI technologies in this field. As researchers continue to explore new frontiers and industry stakeholders embrace innovative solutions, the symbiotic relationship between AI and food safety management is poised to shape the future of a safer and more resilient food system. We expect that AI technologies will play a pivotal role in solving food safety problems, drive innovation across the federal government and food industry, address more real-world challenges, and improve our daily lives.

References

1. U.S. Department of Agriculture Food Safety and Inspection Service (USDA-FSIS). "Food Safety and Inspection Service Strategic Plan 2023–2026." January 2023. https://www.fsis.usda.gov/sites/default/files/media_file/documents/Strategic%20Plan%202023-2026.pdf.
2. Mosallam, A. and A.E. Hassanien. "A Comprehensive Review on Food Safety and Detection of Pathogenic Bacteria in Food." Journal of King Saud University—Computer and Information Sciences. 2020.
3. Buntain, B.J. "Emerging Challenges in Public Health Protection, Food Safety and Security: Veterinary Needs in the USDA's Food Safety and Inspection Service." Journal of Veterinary Medical Education 31, no. 4 (2004): 334–340.
4. Fung, F., H.-S. Wang, and S. Menon. "Food Safety in the 21st Century." Journal of Biomedical Science 41, no. 2 (2018): 88–95.
5. Keener, L., S.M. Nicholson-Keener, and T. Koutchma. "Harmonization of Legislation and Regulations to Achieve Food Safety: US and Canada Perspective." Journal of the Science of Food and Agriculture 94, no. 10 (2014): 1947–1953.
6. Benson, S.M. "Guidance for Improving the Federal Response to Foodborne Illness Outbreaks Associated with Fresh Produce." Food and Drug Law Journal 65, no. 3 (2010): 503–524.
7. Federal News Network. Building a Foundation for your AI Future. 2023. https://federalnewsnetwork.com/cme-event/federal-insights/building-a-foundation-for-your-ai-future/.
8. Lenniy, D. "AI in Agriculture—The Future of Farming." December 12, 2023. Intellias. https://intellias.com/artificial-intelligence-in-agriculture/.
9. Ding, H., J. Tian, W. Yu, D.I. Wilson, B.R. Young, X. Cui, X. Xin, Z. Wang, and W. Li. "The Application of Artificial Intelligence and Big Data in the Food Industry." Foods 12, no. 24 (2023): 4511.
10. Acuff, J.C., J.S. Dickson, J.M. Farber, E.M. Grasso-Kelley, C. Hedberg, A. Lee, and M. Zhu. "Practice and Progress: Updates on Outbreaks, Advances in Research, and Processing Technologies for Low-Moisture Food Safety." Journal of Food Protection 86, no. 1 (2023): 100018.
11. Qian, C., S.I. Murphy, R.H. Orsi, and M. Wiedmann. "How can AI Help Improve Food Safety?" Annual Review of Food Science and Technology 14 (2023): 517–538.
12. Singh, A.V., G. Bansod, M. Mahajan, P. Dietrich, S.P. Singh, K. Rav, A. Thissen, A.M. Bharde, D. Rothenstein, S. Kulkarni, and J. Bill. "Digital Transformation in Toxicology: Improving Communication and Efficiency in Risk Assessment." ACS Omega 8, no. 24 (2023): 21377–21390.
13. Mavegam Tango Assoumou, B.O., C. Coughenour, A. Godbole, and I. McDonough. "Senior Food Insecurity in the USA: A Systematic Literature Review." Public Health Nutrition 26, no. 1 (2023): 229–245.
14. Benammar, M., A. Abdaoui, S.H.M. Ahmad, F. Touati, and A. Kadri. "A Modular IoT Platform for Real-Time Indoor Air Quality Monitoring." Sensors 18, no. 2 (2018): 581.
15. Deng, X., C. Shuhao, and A.L. Horn. "Abstract Emerging Applications of Machine Learning in Food Safety." Annual Review of Food Science and Technology 25, no. 12 (March 2021): 513–538.
16. Harris, M. "Machine Learning and Artificial Intelligence for Pathogen Identification and Antibiotic Resistance Detection: Advancing Diagnostics for Urinary Tract Infections." BioMed 3, no. 2 (2023) 246–255.
17. Patela, D., A. Sinhaa, T. Bhansalia, U. Ga, and V. Sa. "Blockchain in Food Supply Chain." Procedia Computer Science 215 (2022): 321–330.
18. Kapoor, N., R. Lacson, and R. Khorasani. "Workflow Applications of Artificial Intelligence in Radiology and an Overview of Available Tools." Journal of the American College of Radiology 17, no. 11 (2022): 1363–1370.
19. Holzinger, A., K. Keiblinger, P. Holub, K. Zatloukal, and H.N. Müller. "AI for Life: Trends in Artificial Intelligence for Biotechnology." Biotechnology 74 (2023): 16–24.
20. Tian, J. "Computer Vision and Machine Learning for Intelligent Sensing Systems." Sensors 23, no. 9 (2023): 4214.
21. Badal, Y.T. and R.K. Sungkur. "Predictive Modelling and Analytics of Students' Grades Using Machine Learning Algorithms. Education and Information Technologies 28, no. 3 (2023): 3027–3057.
22. Bosona, T. and G. Gebresenbet. "The Role of Blockchain Technology in Promoting Traceability Systems in Agri-Food Production and Supply Chain." Sensors 23, no. 11 (2023): 5342.
23. Rocca, R., N. Tamagnone, S. Fekih, X. Contla, and N. Rekabsaz. "Natural Language Processing for Humanitarian Action: Opportunities, Challenges, and the Path Toward Humanitarian NLP." Frontiers in Big Data 6 (2023): 1082787. https://www.frontiersin.org/articles/10.3389/fdata.2023.1082787/full.
24. MacIntyre, C.R., X. Chen, M. Kunasekaran, A. Quigley, S. Lim, H. Stone, H.Y. Paik, L. Yao, D. Heslop, W. Wei, I. Sarmiento, and D. Gurdasani. "Artificial Intelligence in Public Health: The Potential of Epidemic Early Warning Systems." Journal of International Medical Research 3 (2023): 3000605231159335.
25. Eisenbrand, G. "Current Issues and Perspectives in Food Safety and Risk Assessment." Human and Experimental Toxicology 34, no. 12 (2015): 1286–1290.
26. Parimbelli, E., S. Wilk, R. Cornet, P. Sniatala, K. Sniatala, S.L.C. Glaser, I. Fraterman, A.H. Boekhout, M. Ottaviano, and M. Peleg. "A Review of AI and Data Science Support for Cancer Management." Artificial Intelligence in Medicine 117 (2021): 102111.
27. White-Dzuro, C.G., J.D. Schultz, C. Ye, J.R. Coco, J.M. Myers, C. Shackelford, S.T. Rosenbloom, D. Fabbri. "Extracting Medical Information from Paper COVID-19 Assessment Forms." Applied Clinical Informatics Journal 12, no. 1 (2021): 170–178.
28. Truong, H., D. Qi, A. Ryason, A.M. Sullivan, J. Cudmore, S. Alfred, S.B. Jones, J.M. Parra, S. De, and D.B. Jones. "Does Your Team Know How to Respond Safely to an Operating Room Fire? Outcomes of a Virtual Reality, AI-Enhanced Simulation Training." Surgical Endoscopy 36, no. 5 (2022): 3059–3067.

Davis W. Cheng, Ph.D. is a Supervisory Consumer Safety Inspector contributing his leadership and expertise in poultry, meat, egg product, and ready-to-eat food safety to the U.S. Department of Agriculture's Food Safety and Inspection Service (USDA-FSIS) Office of Field Operations, Philadelphia District. Dr. Cheng's academic journey reflects a profound dedication to the sciences, having majored in Agriculture and Biological Sciences. He holds a Ph.D. in Animal Genetics and Genetic Engineering. Previous to his current position, Dr. Cheng held other roles at USDA including Scientist, Assistant/Associate Research Professor, Senior Scientist, and Biologist, showcasing a multifaceted commitment to teaching, research, student supervision, and business leadership within the food safety industry.