In the ever-evolving landscape of agricultural technology, a groundbreaking approach pioneered at the University of Illinois Urbana-Champaign promises to revolutionize poultry production. Employing advanced hyperspectral imaging (HSI) coupled with state-of-the-art machine learning algorithms, this research tackles longstanding challenges in hatchery operations—most notably, the early prediction of chick embryo mortality, sex determination, and evaluation of eggshell quality. Such innovations could significantly enhance productivity, economic viability, and animal welfare across the poultry industry.
Chicken eggs and poultry stand as central pillars in the global protein supply, underpinning a multibillion-dollar industry with implications reaching far beyond food production. However, hatchery operations grapple with critical inefficiencies, especially embryo mortality rates exceeding 10%, the indiscriminate culling of male chicks, and the destructive testing of shell quality. These issues not only compromise output and profitability but also raise ethical and biosecurity concerns. The University of Illinois research team’s novel application of near-infrared (NIR) and hyperspectral imaging technologies heralds a new era of precision, efficiency, and humane practices in poultry farming.
At the core of this research is the use of hyperspectral cameras that capture detailed spectral information across hundreds of narrow wavelength bands, spanning beyond the visible spectrum into near-infrared regions. Unlike conventional photography, which records images in just three color channels, hyperspectral imaging detects subtle molecular signatures in eggs, revealing internal qualities that are otherwise inaccessible through visual inspection. This capability allows for the non-destructive analysis of embryos and eggshells, capturing critical data about the developing chick and the egg’s structural integrity.
In their pioneering study, the researchers scanned 300 chicken eggs sourced from the university’s poultry farm, applying hyperspectral imaging prior to incubation and again on the fourth day of incubation. By analyzing the spectral data, they identified distinct wavelength patterns associated with viable and non-viable embryos. Leveraging these datasets, machine learning models were trained to predict embryo mortality with remarkable accuracy—up to 97% by day four of incubation. This level of precision offers the potential to remove non-viable eggs early, reducing biosecurity risks posed by decaying embryos harboring harmful bacteria.
The implications for animal welfare and farm biosecurity are profound. Embryo death during incubation is not only a loss in productivity but also creates a vector for bacterial contamination that can jeopardize entire hatchery batches. Early detection via hyperspectral imaging provides an opportunity to intercept and isolate affected eggs swiftly. Such interventions could alleviate economic losses and enhance sanitation protocols without additional labor-intensive manual inspections, augmenting operational sustainability at scale.
Beyond mortality prediction, the team has made strides in non-destructive sex determination of chicken embryos pre-incubation. Presently, hatcheries routinely cull approximately 6 billion male chicks annually in the U.S. alone, due to their limited economic value for egg production or meat. This practice has drawn increasing ethical criticisms and regulatory scrutiny worldwide. By employing hyperspectral imaging combined with machine learning and explainable artificial intelligence, the researchers achieved approximately 75% accuracy in predicting the sex of embryos right at day zero, before incubation even begins.
This advancement opens the door to dramatically reducing male chick culling by enabling hatcheries to segregate eggs based on predicted sex. Male-designated eggs could then be redirected for food production or table egg use, transforming a problematic byproduct into valuable resources. Across Europe, some nations have already legislated bans on chick culling, and the U.S. industry is actively seeking technological solutions to adhere to emerging animal welfare standards. This research thus situates itself at the vanguard of ethical poultry farming innovation.
Additional investigations explored eggshell characteristics, including shell strength, shell thickness, and yolk composition, employing near-infrared spectroscopy—a more cost-effective but less intricate alternative to hyperspectral imaging. Traditional methods to assess these qualities are destructive, requiring eggshell breakage, which wastes valuable products and complicates quality control. The combination of NIR and machine learning enables accurate, non-invasive measures of these parameters, supporting quality assurance and breeding improvements without compromising eggs.
Crucially, translating these laboratory-scale methodologies into practical hatchery applications necessitates automation. The research team is actively developing robotic systems coordinated with machine learning outputs to automate sorting and handling. For example, once the system identifies an egg as male or non-viable, a robotic arm could precisely remove that egg from the production line. This integration of hardware and AI streamlines processes and minimizes human error or labor costs while boosting throughput.
The potential of hyperspectral imaging and explainable AI extends well beyond poultry. Both technologies hold transformative promise across agricultural sectors, from crop health monitoring to food safety evaluations. Their application in poultry production is nonetheless pioneering, providing unprecedented insight into molecular-level egg characteristics and enabling decisions that were once unfeasible without destructive testing or extensive incubation periods.
To facilitate broader research and development, the team has openly shared their near-infrared spectral datasets related to shell strength, thickness, and yolk ratio, fostering collaboration in this emergent field. Plans to publish comprehensive hyperspectral image datasets are underway, offering a valuable resource for researchers aiming to refine models or explore new applications of spectral analysis in animal agriculture.
Funded in part by the USDA’s National Institute of Food and Agriculture under Hatch funding and supported through award #2023-67015-39154, this research exemplifies how interdisciplinary engineering, computer science, and agricultural science can converge to deliver solutions that resonate across economic, ethical, and practical dimensions. As these technologies mature and integrate into commercial hatcheries, the poultry industry may witness a paradigm shift towards more sustainable, humane, and data-driven production systems.
In summary, the fusion of hyperspectral imaging and machine learning sets a new standard in non-destructive egg analysis. From accurately forecasting embryo viability to discerning sex and shell integrity, this technology streamlines hatchery operations, mitigates welfare dilemmas, and bolsters biosecurity. The ongoing development of automated systems further amplifies its transformative potential, ushering in an era where precision agriculture meets artificial intelligence in the service of global food security and ethical animal husbandry.
Subject of Research: Non-destructive analysis of chicken eggs using hyperspectral imaging and machine learning for embryo mortality prediction, sex determination, and eggshell quality measurement.
Article Title: Non-destructive chick embryo mortality prediction at pre-incubation and early incubation using hyperspectral imaging and explainable artificial intelligence
News Publication Date: 20-Feb-2026
Web References:
- British Poultry Science DOI
- Food Control Study on Sex Determination
- Journal of the Science of Food and Agriculture Article
- ACS Food Science and Technology Paper
- Smart Agricultural Technology Article
- Scientific Data Publication
References:
- Ahmed, Md. Wadud, et al. “Non-destructive chick embryo mortality prediction at pre-incubation and early incubation using hyperspectral imaging and explainable artificial intelligence.” British Poultry Science, 2026.
- Additional peer-reviewed articles listed above.
Image Credits: College of ACES, University of Illinois Urbana-Champaign
Keywords: hyperspectral imaging, near infrared spectroscopy, machine learning, explainable AI, chick embryo mortality, sex determination, eggshell strength, poultry industry, non-destructive testing, agricultural engineering, food production, biosecurity.
