Study tests AI spectroscopy for earlier in ovo chick sexing

Version 2 — Full analysis

A newly published Animals study puts fresh attention on one of poultry production’s hardest technical problems: how to identify the sex of a chicken embryo very early, non-destructively, and at scale. The paper describes a dual-stage difference-enhanced spectral fusion network built on visible–near-infrared transmission spectroscopy, using longitudinal data from 1,600 fertilized pink-shelled Bashang Long-tailed chicken eggs collected over incubation days 1 to 7. The goal is straightforward, even if the biology and engineering are not: detect sex-linked differences early enough to improve hatchery efficiency and reduce reliance on post-hatch male chick culling. (pmc.ncbi.nlm.nih.gov)

That goal sits within a broader welfare and industry push that has accelerated over the past decade. In ovo sexing has been pursued as an alternative to culling newly hatched male chicks in layer production, with research spanning spectroscopy, imaging, endocrine assays, DNA-based methods, odor analysis, and MRI. Reviews and policy documents have emphasized that timing matters as much as accuracy, because the ethical and regulatory value of sexing declines as incubation advances. Several reviews cite day 7 as an important precautionary threshold in public debate, even though scientific discussions around embryonic nociception and pain perception are more nuanced. (sciencedirect.com)

The technical challenge is that sex-related optical differences are faint in the earliest days of incubation and change with developmental stage. That’s why the new paper’s use of longitudinal spectra and a staged fusion model is notable. Recent reviews of the field describe this exact problem as a central bottleneck: earlier detection offers the biggest welfare and economic upside, but signal robustness is weakest early on and can be confounded by shell color, embryo orientation, batch effects, and acquisition conditions. In that context, a model designed to enhance weak day-to-day spectral differences is a logical step, even if it remains at the research stage. (sciencedirect.com)

Visible–near-infrared spectroscopy is already a familiar tool in the in ovo sexing literature. Prior reviews describe it as one of the most widely studied optical methods, with published work across different wavelength bands, incubation windows, and model architectures. At the same time, the literature also makes clear that headline accuracy numbers can be hard to compare across studies because datasets, breeds, shell types, and validation methods vary widely. That means any promising result still has to answer practical questions about repeatability, speed, contamination risk, hatchability, and transferability outside the original dataset. (pmc.ncbi.nlm.nih.gov)

Commercial activity in the sector shows why that distinction matters. Companies including Seleggt and Orbem already market in ovo sexing platforms, while SANOVO, Canadian Egg Technologies, and MatrixSpec announced a partnership in June 2026 to commercialize a day-4 non-invasive in ovo sexing system. Those developments suggest the field is moving beyond proof-of-concept and into operational competition over speed, accuracy, cost, and hatchery integration. Against that backdrop, new academic models are likely to be judged less on novelty alone and more on whether they can outperform or complement systems that are already being installed in commercial hatcheries. (seleggt.com)

Why it matters: For veterinarians working in poultry medicine, hatchery health, animal welfare, or technical services, this study is another sign that embryo-stage decision-making is becoming more data-driven. If early, non-destructive sexing becomes more accurate and scalable, it could change how hatcheries allocate incubator space, manage biosecurity workflows, and communicate welfare practices to regulators, retailers, and pet parents who increasingly scrutinize food-animal production systems. But veterinary professionals should read these studies with a practical lens: the key questions are not just whether a model works in one paper, but whether it works across strains, shell colors, and real hatchery environments without compromising embryo viability. (sciencedirect.com)

There’s also a policy and market angle. Government and industry discussions in Europe and the UK have continued to frame in ovo sexing as a likely route away from routine chick culling, and U.S. commercial adoption appears to be growing, though from a smaller base. That creates pressure for methods that are both scientifically defensible and operationally credible. A spectroscopy-based system that can work earlier, non-invasively, and across diverse egg types would be valuable, but that bar remains high. (gov.uk)

What to watch: Watch for external validation studies, side-by-side comparisons with existing commercial platforms, and any data on hatchability, false-classification costs, and performance in breeds beyond pink-shelled Bashang Long-tailed chickens. Those are the milestones that will determine whether this remains an interesting model paper or becomes relevant to hatchery practice. (sciencedirect.com)