Fluorescence microscopy has long been a cornerstone in biological and medical research due to its ability to selectively visualize specific molecular components within cells and tissues. However, the technique inherently suffers from a restricted depth of field, making it difficult to capture sharp images across thicker biological samples without mechanical sectioning or complex optical adjustments. By integrating engineered meta-optical components with sophisticated computational imagery, the researchers have dramatically extended the depth of focus while preserving spectral sensitivity. This feat addresses a critical bottleneck, potentially enabling volumetric imaging with unprecedented ease and fidelity.

At the heart of this innovation lies the co-design principle, whereby the hardware optics—particularly meta-surfaces composed of subwavelength nanostructures—are crafted in synergy with deep learning-based neural reconstruction models. Unlike traditional lenses, meta-optics provide flexible phase manipulation capabilities on an ultra-thin platform, allowing customized point spread functions that encode rich spatial and spectral information. The researchers exploited this characteristic to create a bespoke optical system that intentionally produces complex but decipherable light patterns, which are then computationally inverted by deep neural networks to reconstruct sharp, multispectral images with extended focus.

This co-engineering approach stands in stark contrast to conventional microscopy, where optical hardware and computational post-processing workflows are developed independently. By jointly optimizing the meta-optical design and neural reconstruction algorithms, the team achieved a holistic system with synergistic performance enhancements. The neural networks were trained to decode the tailored optical signatures into high-fidelity fluorescence images, effectively compensating for aberrations and chromatic dispersion while maintaining multispectral accuracy. The result is a robust imaging modality capable of capturing spatially resolved fluorescence signals over a larger axial range without mechanical scanning.

The meta-optics deployed in this system consist of arrays of nanostructured elements meticulously designed to modulate phase shifts for multiple wavelengths simultaneously. These meta-lenses are fabricated using advanced nanofabrication techniques, enabling precise control over light propagation that surpasses classical diffractive or refractive optics. By carefully tuning the geometric parameters of these nanostructures, the researchers engineered a spectral point spread function that encodes depth information intrinsically, facilitating the extended focal depth. This innovation allows for diffraction-limited resolution across a sample thickness previously unattainable with standard objectives.

To realize the full potential of these meta-optical components, the researchers developed a deep convolutional neural network tailored for inverse problem solving in fluorescence image formation. The network architecture is designed to process raw hyperspectral data and output high-resolution, depth-compensated images. It learns to disentangle overlapping signals and mitigate optical aberrations through extensive supervised training on simulated and experimentally acquired datasets. The neural reconstruction not only enhances image sharpness and contrast but also reconstructs spectral variations critical for multichannel fluorescence assays, paving the way for multiplexed biological analyses.

One of the profound implications of this methodology is its applicability to live-cell and tissue imaging. Traditional extended depth-of-field techniques often require mechanical focus stacking or suffer from significant image quality trade-offs. The approach demonstrated here circumvents these limitations by offering a snapshot imaging capability that captures volumetric information instantaneously. This can accelerate dynamic studies of complex biological processes such as intracellular trafficking, neuronal activity, and developmental morphogenesis, providing researchers with richer spatiotemporal datasets without compromising sample viability.

Further enhancing the utility of the system is its multispectral imaging capability, which is essential for simultaneous visualization of multiple fluorophores. The carefully designed meta-optics and neural reconstruction pipeline maintains spectral fidelity across a broad wavelength range, ensuring accurate identification and quantification of biomolecules tagged with different fluorescent markers. This multispectral capacity extends the range of experiments this technique can support, from mapping gene expression patterns to monitoring multiplexed immunostaining with higher accuracy and efficiency.

The integration of meta-optics with neural networks also opens avenues for translational biomedical applications. For instance, in pathological diagnosis, rapid and comprehensive fluorescence imaging could yield more insightful morphological and molecular profiles of tissues, enhancing early disease detection and personalized treatment planning. Moreover, the compact nature of the meta-optical elements allows miniaturization of imaging devices, facilitating point-of-care diagnostics and portable microscopy solutions.

Critically, the study also addresses the engineering challenges associated with hardware-software co-design. The researchers implemented an iterative optimization pipeline where optical design parameters and neural network training were continuously refined to maximize imaging performance. This closed-loop development harnesses simulation-driven feedback and experimental validation, ensuring the system operates optimally under real-world conditions. The resulting microscope prototype demonstrates reliable operation with minimal calibration overhead, emphasizing the practical readiness of this technology.

By transcending traditional microscopy constraints, the co-designed meta-optics and neural reconstruction framework herald a new era for fluorescence imaging. Its ability to achieve high-resolution, multispectral visualization across extended depth spans challenges existing paradigms and stimulates exciting research directions in photonics, machine learning, and life sciences. Researchers anticipate that further refinement and integration of these technologies will accelerate discoveries in cell biology, neuroscience, and beyond.

This breakthrough exemplifies the profound impact of converging physics-based optical engineering with data-driven computational methods. It showcases how meta-surfaces can be harnessed not merely as passive lenses but as active information encoding elements when combined with artificial intelligence, fundamentally redefining microscope architecture. The potential to adapt these principles to other imaging modalities, including phase contrast or super-resolution techniques, suggests a fertile landscape of future innovations.

As this technology matures, questions remain regarding scalability, cost-effectiveness, and ease of adoption in diverse laboratory settings. The researchers are optimistic that ongoing advancements in nanofabrication and machine learning infrastructure will address these challenges, making multispectral extended depth-of-field fluorescence microscopy accessible to a broad scientific community. This promise aligns with growing demands for comprehensive, fast, and minimally invasive imaging tools capable of unraveling complex biological systems.

Ultimately, the fusion of meta-optical engineering and neural image reconstruction represents a landmark in microscopy development. It not only enhances technical capabilities but also catalyzes the integration of multidisciplinary expertise spanning optics, computer science, and biology. Such collaborative innovations signal a transformative future for biomedical imaging, enabling researchers to explore the microcosm with clarity and depth previously thought unattainable.

Subject of Research: Multispectral fluorescence microscopy with extended depth-of-field achieved via co-designed meta-optics and neural reconstruction algorithms.

Article Title: Multispectral extended depth-of-field fluorescence microscopy with co-designed meta-optics and neural reconstruction.

Article References:
Atalay Appak, I.A., Singh, H.J., Korpela, S. et al. Multispectral extended depth-of-field fluorescence microscopy with co-designed meta-optics and neural reconstruction. Light Sci Appl 15, 242 (2026). https://doi.org/10.1038/s41377-026-02337-y

Image Credits: AI Generated

DOI: 19 May 2026

Understanding the form and structure of organisms has always been fundamental to biology, yet the sheer complexity and minuteness of many species, particularly insects, have long posed formidable challenges for comprehensive morphological studies. Traditional techniques rely heavily on specimen mounting and preservation methods that maintain external features while internal anatomy rapidly degrades post-mortem. Moreover, conventional imaging tools such as laboratory-based CT scanners are prohibitively slow and costly for mass-scaling research, often restricting studies to individual or few specimens.

Antscan transcends these constraints by employing a synchrotron particle accelerator as a powerful X-ray source coupled with automated robotic sample handling to facilitate ultra-high-throughput scanning. By orchestrating a robotic arm to systematically exchange specimen vials, the system generates approximately 3000 two-dimensional X-ray projections per specimen. Advanced computational algorithms then reconstruct these projections into comprehensive 3D tomograms, revealing ant morphology with unmatched resolution and fidelity. Remarkably, this technology condensed what would conventionally require six years of continuous scanning down to a single week.

The breadth and depth of Antscan’s dataset are transformative for entomologists, morphologists, and the broader scientific community. Every specimen’s metadata—including taxonomy, collection locality, collector identity, and caste designation—has been meticulously standardized to ensure data integrity and facilitate comparative research. The collaboration spanned a network of museums, research institutions, and private collections worldwide, highlighting the project’s global scale and ambition.

One of the most striking facets of Antscan is its commitment to open science. All raw data are freely accessible through an online portal featuring a built-in 3D viewer, democratizing access to what has hitherto been an extraordinarily resource-intensive form of data generation. This accessibility extends opportunities beyond academia, inviting citizen scientists, educators, digital artists, and conservationists to explore and utilize the models for diverse purposes. The detailed visualizations of musculature and articulation open new avenues in biomechanical modeling, enhancing our understanding of insect locomotion and physiology.

From an ecological perspective, ants serve as keystone organisms in numerous terrestrial ecosystems, their social and structural diversity underpinning complex environmental interactions. The Antscan project not only provides a reservoir of morphological data but also paves the way for integrative studies combining genomics, biodiversity mapping, and phenotypic variation. Previous complementary efforts by OIST have already produced global distribution maps of ant biodiversity and high-quality genome assemblies covering most ant genera, setting the stage for multifaceted investigations.

The synergy of big data and advanced imaging realized by Antscan exemplifies the dawning era of quantitative phenomics, where organismal shape and form can be studied at scales and resolutions previously unimaginable. Professor Evan Economo, a co-leader of the project, highlights how this initiative advances a paradigm shift, facilitating the capture, analysis, and sharing of morphological data with unprecedented breadth. The potential for cross-disciplinary insights that merge evolutionary biology, ecology, and computational modeling is immense.

Technically, the integration of synchrotron-based micro-CT scanning into biological research represents a novel application of physics and engineering resources long dominated by materials science and medical imaging. The method involves exposing specimens to a highly collimated, intense X-ray beam, capturing slices at micron-scale increments. The resultant volumetric data enable virtual dissection and exploration without physical damage, preserving valuable specimens for future study while maximizing information yield.

Moreover, the operation’s high-throughput nature required innovations in robotics and software automation. Sorting and cataloging over 2000 ant specimens demanded coordinated efforts involving numerous experts who painstakingly classified ants by species and caste before scanning. The realized workflow sets a new benchmark for mass phenotyping projects, suggesting that parallel efforts could soon be undertaken for other taxonomically rich and morphologically complex groups.

The implications of this work extend to ecological and evolutionary research realms. In a related study published in Science Advances, Antscan data were instrumental in investigating the balance between worker ant quantity and quality within colonies. Findings indicate that favoring numerous nutritionally inexpensive workers over fewer, heavily armored individuals correlates with the evolution of larger and more resilient social systems—a revelation made possible by the rich morphological detail Antscan provides.

Looking forward, Antscan’s open-access framework promises to catalyze myriad innovative applications, spanning from biomimetic robotics design, ecological monitoring, species identification, to public engagement in biodiversity sciences. By lowering barriers to high-resolution morphological data, this initiative may inspire new models of collaborative research and education, galvanizing a broad community invested in understanding and preserving the diversity of life forms on our planet.

In sum, Antscan is a landmark achievement in phenomics, marrying cutting-edge technology and collaborative science to unlock the detailed morphology of some of Earth’s most ecologically significant insects. This resource not only enriches fundamental biological knowledge but also exemplifies how modern techniques can overcome historical bottlenecks, enabling comprehensive, scalable, and accessible study of organismal form and function.

Subject of Research: Animals

Article Title: High-throughput phenomics of global ant biodiversity

News Publication Date: 5-Mar-2026

Web References:

• Antscan project portal: https://antscan.info/
• Related global ant biodiversity map: https://www.oist.jp/news-center/news/2022/8/4/new-global-map-ant-biodiversity-reveals-areas-may-hide-undiscovered-species
• Ant genomes publication: https://www.cell.com/cell/fulltext/S0092-8674(25)00617-8
• Study on ant colony organization: https://www.oist.jp/news-center/news/2025/12/20/ant-societies-rose-trading-individual-protection-collective-power

References:
Katzke, J. et al. (2026). High-throughput phenomics of global ant biodiversity. Nature Methods, DOI: 10.1038/s41592-026-03005-0.

Image Credits: Thomas van de Kamp

Keywords: Antscan, Ant morphology, Micro-CT scanning, Synchrotron imaging, Phenomics, Biodiversity, 3D modeling, Insect anatomy, High-throughput imaging, Big data in biology, Computational morphology, Ant ecology