Researchers at the University of South Australia are using machine learning in hyperspectral imaging to detect toxic contamination that can cause food poisoning.
One of the first applications of artificial intelligence was created in 1951. It is a program written for the games of checkers and chess. In recent years, AI has taken the world by storm with platforms that can be applied across industries to reduce labor and simplify processes.
When it comes to the food and beverage industry, many people associate AI with data analysis and machine process automation. But in a world where food safety and quality assurance are becoming increasingly important, the use of AI and machine learning has the potential to go far beyond saving lives.
New research led by the University of South Australia (UniSA) is paving the way to smarter and more efficient pollution detection. Collaborating with an international team of researchers, the study aims to remove harmful toxins from food, especially nuts, before they reach consumers.
The paper, published in the journal “Food Chemistry, a correlation-aware evolutionary sparse hybrid spectral band selection algorithm to detect almonds contaminated with aflatoxin B1 using hyperspectral images,” describes the integration of machine learning and advanced hyperspectral imaging (HSI) to identify toxins that can contaminate food.
The paper is written by first author and UniSA PhD candidate Ahasan Kabir, Associate Professor Ivan Lee and Professor Sang-Heon Lee (UniSA). Professor Chandra Singh (University of Lethbridge, Canada). Assistant Professors Gayatri Mishra and Brajesh Kumar Panda (Indian Institute of Technology Kharagpur);
4 million people died from food poisoning
So what exactly is being detected? From growing to storing grains and nuts, many types of fungi can grow and produce numerous mycotoxins. Mycotoxins are toxic, mutagenic, and carcinogenic compounds that cause a variety of health problems when ingested by humans and animals.
According to the World Health Organization (WHO), food contamination has made nearly 600 million people sick and killed 4.2 million. Mycotoxin contamination poses a threat to public health as a common cause of foodborne illness, and economic and health losses erect trade barriers year after year.
Dr. Kabir's doctoral research focused on the detection of aflatoxin B1 in almonds, a mycotoxin classified as a Group 1 carcinogen by the International Agency for Research on Cancer. According to him, aflatoxin B1 is the most dangerous of the four major aflatoxins (B1, B2, G1, and G2). These naturally occurring pollutants can grow in warm, humid environments and cause liver damage and long-term health problems. It also poses a threat to both public health and trade.
Contamination can occur at almost any point in the supply chain, from growing, harvesting, drying, packaging, and transportation. International regulations place strict limits on acceptable levels of toxins, and even small violations can result in the rejection of the entire export shipment.
“Australia is the second largest producer of almonds,” Kabir said. “When exporting almonds to countries such as Europe and Japan, the amount of aflatoxin in the almonds is regulated.”
Kabir explained that the aim of the research is to develop technology that can identify highly contaminated almonds before they are exported and protect both consumers and producers.
A thorough review of traditional approaches
Traditional aflatoxin detection relies on chemical testing using high-performance liquid chromatography (HPLC). This method provides high precision and selectivity with detection levels as low as 0.01 ppb. This involves grinding large quantities of almonds into a powder, mixing it with a methanol solution, and performing chromatographic analysis in the laboratory.
Although it has been proven to be effective, it also has drawbacks.
“The current process is destructive and time-consuming, and random sampling poses problems,” Professor Lee said.
Typically, only a uniformly ground 100-gram portion of a 20-kilogram sample is used for testing. This means that if the results show a higher contamination level than the standard, the entire lot will be rejected. HPLC tests are performed in two different laboratories, and from start to finish, one sample can take nearly an hour to prepare and analyze. Since only a small number of nuts are tested from each batch, there is always a risk that contaminated almonds may remain undetected.
UniSA's research takes a new approach. Rather than relying on chemical destruction, it uses a combination of HSI and machine learning to detect toxins based on the spectrum of light reflected from each almond. This method allows the entire batch to be tested quickly and non-destructively in real time.
Where AI comes into play
Most people are familiar with the standard red, green, blue (RGB) imaging used in everyday cameras. Hyperspectral imaging extends this principle by capturing data across hundreds of narrow optical bands beyond the visible spectrum from 280 to 750 nanometers, including near-infrared wavelengths, which Kabir describes as 900 to 1,700 nanometers.
In this study, a hyperspectral camera that can scan wavelengths invisible to the human eye generates 224 different channels. Each channel represents a specific wavelength of light and provides detailed information about the chemical composition and structure of the substance being scanned. By analyzing these light patterns, the research team can determine which spectral regions are most sensitive to the presence of aflatoxin B1.
The strength of this technology is that it can accurately identify spectral features that are invisible to the human eye. Almonds contaminated with aflatoxin may look the same as safe almonds under normal light, but hyperspectral imaging reveals subtle chemical differences. These differences are classified using a trained artificial intelligence model, allowing for rapid screening of almonds on conveyor belts without physical contact or chemical treatment.
Next is machine learning. Using algorithms to process data and identify contaminated almonds within milliseconds (a process called “dimensionality reduction” and “classification”), machine learning helps distinguish between toxic nuts and safe nuts with over 93% accuracy. Without machine learning, the amount of data in hyperspectral images would be too large to be processed by traditional statistical classifiers. The ability of algorithms built to learn complex data allows HSI to be performed accurately.
Beyond nuts
According to the Food and Agriculture Organization (FAO), a specialized agency of the United Nations, about 25 percent of the world's crops are contaminated by fungi that produce mycotoxins. To combat this, we need fast and reliable frameworks such as HSI with machine learning to address threats such as aflatoxin B1. This enables the detection and quantification of maximum acceptable levels of mycotoxins from complex food materials.
Although the research began with almonds, the potential extends beyond the nut industry. The same hyperspectral imaging principles can be applied to pistachios, grains, rice, and barley, all of which can be affected by similar toxins. Furthermore, this technology extends beyond food.
“This technology has been around for more than a decade and has applications not only in the food and beverage sector, but also in agriculture, medicine and satellite applications,” Kabir said.
In agriculture, it helps determine soil composition, crop health, and potential contamination. In medicine, this technology is used to detect conditions such as skin cancer and breast tumors by analyzing tissue composition without invasive procedures. Satellite image processing and mining performs chemical analysis of soil to detect mineral deposits.
Powering this technology through machine learning provides a scalable, non-invasive solution across multiple industries. One of the features of this approach is that you can work in real time. With further development, HSI and machine learning could be deployed on processing lines and handheld devices, potentially reducing health risks and trade losses by ensuring only safe and uncontaminated produce reaches consumers.
What's next?
While this technology is still primarily used in laboratory research, the next challenge lies in applying it to industrial environments. The UniSA team is currently working with industry partners to move from a research prototype to a scalable commercial system. With funding from the Federal Government's Research and Training Program and Australian nut producer SureNut Australia, the team is refining the technology to improve accuracy and reliability through deep learning and AI.
SureNut Australia is currently piloting a prototype imaging system using cameras mounted above the conveyor line. The system aims to automatically detect and eliminate aflatoxin-contaminated almonds in real time, reducing waste and ensuring only safe products are packaged.
“Our future goal is to hopefully apply this technology to other areas where other toxins can be detected,” Professor Lee said.
For food and beverage manufacturers, this is an opportunity to integrate smarter detection systems that meet global expectations for traceability, sustainability, and consumer protection. The combination of hyperspectral imaging and machine learning could soon define new standards for food safety. This doesn't start in the lab, but on the production line itself.
