Extracellular matrix (ECM) micropatterns provide well-defined adhesive environments for investigating how geometric confinement influences cell migration. Here, we combined ECM protein micropatterning with quantitative live-cell microscopy to examine NIH 3T3 fibroblast migration on straight and sinusoidal adhesive tracks. Time-lapse bright-field imaging, actin fluorescence imaging, centroid tracking, migration-rate quantification and aspect-ratio analysis revealed distinct migration behaviours under confined adhesion. On straight tracks, cells exhibited either oscillatory migration, characterised by repeated cycles of protrusion, retraction, and migration-rate bursts, or confined migration, characterised by limited centroid displacement and slow morphological remodelling. Sinusoidal tracks further introduced curvature-associated changes in migration dynamics, with cells adjusting leading-edge orientation along the curved path or becoming locally confined within curved regions. These observations demonstrate that adhesive geometry influences cell migration beyond simple path guidance and highlight the value of engineered ECM micropatterns combined with live-cell microscopy for resolving dynamic migration behaviours under confined conditions.
Manual annotation of microscopic images of thin tissue sections remains time-consuming step limiting the throughput of diagnostic analysis. We present a novel automated approach using the total intensity image as an input to a U-Net architecture of Convolutional Neural Network with a pretrained ResNet-34 encoder for image segmentation. The network was trained on limited-size annotated dataset to distinguish four classes in the images of thin sections of murine uterine cervix: background, internal os, cervical tissue, and vaginal wall. With only 74 manually annotated images in total (51 used for model training), the model achieved 90.22% pixel accuracy on the held-out test dataset. This framework requires minimal data pre-processing and is readily extensible to other tissue types, with publicly available graphical annotation tools for practical deployment.
Multimodal optical microscopy has recently converged with artificial intelligence (AI) to enable in silico labelling-the computational prediction of fluorescence-like molecular contrast from label-free measurements. Fluorescence plays a crucial role in linking microscopy and spectroscopy at molecular level, enabling image formation by means of linear and non-linear investigation modalities. Over the past two decades, fluorescence optical microscopy has progressed into optical nanoscopy and single molecule localisation methods that operate at the nano- and even Ångstrom level under ambient conditions. Methods such as Stimulated Emission Depletion (STED), Photo-Activated Localisation (PALM), Stochastic Optical Reconstruction (STORM), super-resolution fluorescence lifetime imaging microscopy (FLIM) and image scanning microscopy (ISM), Minimal Fluorescence photon Flux (MINFLUX) microscopy allows us to investigate living cells at the molecular level. A significant and challenging development in this field is the coupling of fluorescence with label-free polarisation and phase optical methods-for example, Mueller-matrix (MM) microscopy-with the aim of extracting specific molecular information from label-free datasets. In this review, we discuss how this convergence, together with modern generative modelling for in silico labelling, is turning the optical microscope into an intelligent instrument.
Drift is one of the primary sources of uncertainty limiting the accuracy, repeatability, and reliability of scanning probe microscopy (SPM) measurements, with a particularly significant effect on the comparison of consecutive surface topography images. In this study, a two-module hybrid physical-computational framework is presented for the drift-aware interpretation and reliability assessment of consecutive AFM topography measurements. The first module integrates deep learning (DL)-based inter-frame correspondence extraction, geometric validation, physical coordinate reconstruction, scan-order-based acquisition-time assignment, inter-scan delay correction, and temperature synchronisation to generate a time-labelled drift representation. The second module uses this structured representation for comparative data-driven modelling and evaluates drift velocity ( v d ), model performance, and stability through temperature-included and temperature-excluded configurations. This structure enables the joint analysis of validated correspondences in terms of lateral displacement, relative height variation, acquisition time, v d , drift direction, and synchronously recorded ambient temperature, together with their physical coordinate labels. The results indicate that the proposed framework should be positioned not as a universal drift-correction method, but as a complementary and modular analysis approach that supports the drift-aware interpretation, comparison, and reliability assessment of consecutive AFM topography measurements.
Conventional transmission electron microscopy (TEM) staining improves image contrast but requires additional sample preparation, consumes reagents and tissue, and may introduce deformation, local mismatch, or tissue loss. When strictly paired stained and unstained sections are difficult to obtain, unpaired virtual staining provides a practical route for improving the readability of unstained TEM images while retaining correspondence with source-image structural cues. This paper presents WAFMGAN, a frequency-aware CycleGAN framework that combines Wavelet Asymmetric Frequency Mamba (WAFM) blocks for low-frequency-dominant global staining transfer and long-range structural coordination with a Frequency-Adaptive Gated Refinement (FAGR) block for later-stage boundary and texture refinement. On the mouse renal TEM test set, WAFMGAN achieved an FID of 14.8064 and a KID( × 100 ) of 0.5074 ± 0.0417, representing reductions of 23.0% and 47.3% relative to CycleGAN. At the original-micrograph-pair level, WAFMGAN achieved SSIM, MS-SSIM, CSS, and Gradient-SSIM values of 0.9796, 0.9874, 0.9825, and 0.9889, respectively, and all four structure-related metrics were significantly higher than those of every evaluated baseline after Holm-Bonferroni correction. Output-level frequency analysis further showed lower wavelet-domain discrepancy than CycleGAN on all four tissue sets and lower high-frequency Fourier discrepancy on the renal, heart, and liver sets. These results demonstrate that WAFMGAN improves stained-domain distribution and frequency matching while maintaining strong source-image structural consistency. WAFMGAN therefore provides a useful frequency-aware approach for improving the readability of unstained TEM images while maintaining source-image structural cues.
Reliable synapse quantification remains challenging, particularly in high-throughput fluorescence imaging workflows. Here we present SynapTrack, a fully automated FIJI/ImageJ-based tool for quantifying synapses in cultured neurons and brain tissue sections with minimal user intervention. SynapTrack combines channel-specific preprocessing, background subtraction, and SynQuant-based detection of colocalised pre- and postsynaptic puncta. In cultured neurons, it additionally measures cell number and total dendritic length, enabling normalisation of synapse counts per cell and per 10 micrometer of dendrite. Using hippocampal cultures, SynapTrack reliably quantified excitatory and inhibitory synapses, captured the expected predominance of excitatory contacts, and reduced inter-sample variability through structural normalisation. Compared with SynBot, SynapTrack detected more synapses while better preserving biologically meaningful differences between synapse types. For tissue sections, in which cell counting and dendritic segmentation are impractical, we developed SynapTrack_Tissue, an adapted workflow based on presynaptic-postsynaptic colocalisation alone. Despite the lack of cellular normalisation, this approach produced stable measurements across heterogeneous neuropil regions. Because all spatial parameters are defined in micrometres, analysis settings are portable across imaging systems. SynapTrack therefore provides a standardised and scalable framework for synapse quantification in both cultured neurons and tissue sections.
Compact Open-Source Multimodal Illumination Cluster ('COSMIC') is a condenser-free illumination module that enables bright field, dark field, diascopic fluorescence and differential phase contrast (DPC) imaging on both commercial and open-source microscope platforms. COSMIC is constructed using low-cost, widely available components and is controlled through µManager, facilitating straightforward integration into existing imaging systems. We demonstrate multimodal transmitted-light imaging of live cell cocultures and living Caenorhabditis elegans specimens, including quantitative phase imaging (QPI) derived from DPC measurements. We further validate and calibrate quantitative DPC (qDPC) using a quantitative phase target on a range of open-source and commercial systems and introduce a practical method for optimising the axial position of the illuminator without access to the objective's back focal plane. COSMIC provides an effortless route to multimodal and quantitative phase imaging on any optical microscopy system, using minimal optical hardware and supporting both research-grade and resource constrained microscopy environments.
Optical microscopy is known to be a powerful technique for the characterization of industrial clinkers, allowing for obtaining valuable information on many aspects of the production process. The time required for the preparation of samples and collection, and interpretation of pictures strongly limits its application on a large scale. Notably, the analysis of samples requires skilled operators and is often influenced by a significant degree of subjectivity. Consequently, the automation of this task is of great interest to the field. For this reason, in this study, we explore the possibility of adopting supervised artificial intelligence (AI) methodologies on clinker micrographs. This study aims to assess whether an AI model can be employed to support human operators in the interpretation of experimental results. Specifically, this paper focuses on the model's ability to autonomously determine the parameters of the experiment. The results demonstrate a promising capability of AI models to perform this task.
Atomic-scale characterisation of thin buried layers in oxide heterostructures is often limited by the reduced sensitivity of conventional Z-contrast high-angle annular dark-field (HAADF) imaging to light elements and low-density regions. Here, a Sc2O3/Er2O3/Si heterostructure was investigated using aberration-corrected scanning transmission electron microscope (STEM) with simultaneous HAADF and integrated differential phase contrast (iDPC) imaging, with site-specific cross-sectional specimens prepared by scanning electron microscope-plasma focused ion beam (SEM-PFIB) using a Xe-ion beam under identical acquisition conditions to enable direct contrast comparison. Although HAADF clearly resolves the heavy rare-earth oxide layers and reveals a broad interfacial region with contrast variations indicative of structural complexity, simultaneous STEM-iDPC imaging resolves a distinct ∼3 nm buried interfacial layer with substantially greater clarity. Additional high-resolution transmission electron microscopy (HRTEM) and electron energy-loss spectroscopy (EELS) analyses reveal oxygen enrichment and local structural ordering within this region, indicating that it is chemically and structurally distinct from both crystalline Si and bulk Er2O3. The combined observations are consistent with an oxygen-rich Er-Si-O transition layer. These results demonstrate the capability of STEM-iDPC to reveal buried interfacial structure and highlight the value of combining HAADF, iDPC, HRTEM and spectroscopy for comprehensive characterisation of oxide heterostructures.
In an effort to assess the reproducibility of bioimage analyses in current publications, we took part in a Global BioImage Analysts' Society (GloBIAS) initiative to try and reproduce results from published articles. We attempted to reproduce core findings from Bingham et al. (2023), who investigate the actin organisation in presynaptic structures by using diffraction-limited and super-resolution microscopy. While the original paper unveiled novel biological insight, it lacked sufficient detail in the bioimage analysis methodology, limiting the depth of reproducibility we could achieve. Through frequent contacts with the corresponding author, we were able to replicate qualitative aspects of the analysis of actin nanostructures in bead-induced presynapses. We performed image reconstruction from super-resolution microscopy data, automatic image registration and visual inspection, followed by manual annotation of structures of interest in approximately 35 images. From this exercise, we provide concrete evidence that key practises such as sharing of example datasets, depositing manual annotations and documenting manual decision criteria and consensus procedures are essential for making bioimage analysis workflows reproducible. Our experience further highlights that transparent data sharing, adherence to bioimage analysis standards, and close collaboration between experimentalist and bioimage analysis specialists are critical to ensure the reproducibility of today's complex biological imaging studies.
Dual-energy X-ray absorptiometry (DXA) remains the clinical gold standard for assessing bone mineral density (BMD), guiding diagnosis and therapeutic decisions. However, conventional DXA analysis suffers from several limitations, including insensitivity to early microarchitectural changes, operator dependence, limited availability in primary care settings, and an insufficient ability to predict fracture risk when used alone. Artificial intelligence (AI), incorporating machine learning (ML) and deep learning (DL), offers transformative potential in enhancing DXA-based bone health assessment. The purpose of this review is to describe the integration of AI algorithms into DXA image interpretation, highlighting improvements in diagnostic accuracy, and fracture risk stratification beyond traditional methods. Using AI-driven models, complex features of DXA images can be extracted, increasing sensitivity to microstructural deterioration that is typically not detected by standard BMD measurements. A combination of quantitative image features and comprehensive demographic and clinical data enhance the early detection of osteoporosis and fracture susceptibility, enabling personalised treatment strategies. In comparison to classical DXA or fracture risk assessment tool (FRAX) algorithms, convolutional neural networks (CNNs) and ensemble methods demonstrate superior predictive performance, with average area under the curve (AUC) values often about 0.90. In addition to minimising inter-operator variability and improving reproducibility, AI improves DXA technical challenges such as region-of-interest selection and image segmentation. In addition to providing indirect measurements of bone microarchitecture, AI-enabled indices, such as the trabecular bone score (TBS), contribute to the improvement of fracture risk prediction. It has been demonstrated in large-scale clinical validations that AI-assisted DXA can enhance bone health diagnostic capability.
In situ biasing electron microscopy is the necessary step forward in being able to characterise nanoscale devices and device interfaces. STEM-EBIC is a new approach that can help fill the missing gaps in creating reproducible methodologies for in situ biasing TEM. The ability to visualise electric fields in in operando lamella scale devices makes it possible to move past the previous approach of dissecting devices post failure. For STEM-EBIC to be useful for in operando device characterisation it is necessary to understand the effects of electron beam interactions with the sample and the behaviour of the device under a variable bias. This paper shows how sample charging is controlled by the scan direction and how this affects the interpretation of STEM-EBIC experiments. This work also maps the local changes in electric field in a metal/oxide/semiconductor 4H-SiC capacitor as it changes between the accumulation and depletion regimes.
Ploem's filter cube, central to modern epifluorescence microscopy, integrates three matched optical components: an excitation filter, an emission filter, and a dichroic mirror. Together, these elements enable the separation of excitation and emission light paths, producing high-contrast fluorescence images. Despite their importance, the origins of these optical filters and the concept of a matched filter set remain unattributed. This review traces the development of these principles to discoveries made half a century before the fluorescence microscope. Through a narrative historical analysis, we identify George Stokes, Wilhelm Haidinger, and David Brewster as key contributors whose studies of light, colour, and polarisation established the foundations of fluorescence filtering. We also show that the matched filter concept can be dated to 173 years ago and that early dichroic films predated Ploem's cube by more than a century. By uncovering these origins, this work highlights how early optical discoveries shaped the design of modern fluorescence microscopy.
Reproducibility challenges in bioimage analysis are common. The proliferation of tools and workflows creates a complex ecosystem that is difficult to reproduce without deliberate, explicit reporting and support. We assess the reproducibility of Gehrels et al.(2023) as a representative bioimage analysis workflow. We do not evaluate the biological claims; instead, we examine the reproducibility of the bioimage analysis methods used and the implications for current peer-review practices in the field. Although some data were deposited in the BioImage Archive and code was available on GitHub, the materials were insufficient to completely reproduce the analysis. Moreover, the repository lacked the configuration to enable straightforward replication of the workflow or results. This case study highlighted how incomplete reporting can significantly hinder the reproduction of bioimage analysis workflows. Because scientific progress relies on verifiable replications, insufficiently reproducible results can erode confidence in the scientific record.
Modern microscopy enables us to measure structural and dynamical properties of many biological processes and is therefore an indispensable research tool. However, the amount and complexity of the produced imaging data is steadily increasing. Thus, handling the data as well as reproducibly and automatically extracting accurate scientific information require dedicated 'bioimage analysis' expertise. To facilitate the dissemination of this ubiquitously required expertise we developed an open-access bioimage analysis training resource. The resource is designed to help trainers to design and run courses on bioimage analysis for life scientists. The material is modular where each module covers one concise topic and provides corresponding activities using microscopy images from biological samples. The activities can be executed using various popular software packages (e.g. ImageJ, Python). The material is hosted on a public software repository allowing the bioimaging community to readily contribute new training modules or improve existing modules. Within the last 3 years, the material has been used by several trainers in numerous courses and continuously improved.
It has been 20 years since the pioneering work of Shinya Yamanaka and Kazutoshi Takahashi at Kyoto University led to the first successful generation of induced pluripotent stem cells (iPSCs) from mouse embryonic and adult fibroblast cells. iPSCs have the capacity to differentiate into any type of cell in the human body, and as such, they have become ubiquitous in medical and biological research. Throughout the development and use of iPSCs and their derivatives, fluorescent microscopy has been particularly integral, playing a central role in the characterisation of cells and analysis of cellular function of both iPSCs and their derived cells. On the 20th anniversary of their discovery, this short review summarises the development of iPSCs, their application in research and clinical settings, and highlights advances in microscopy and imaging methodologies that have been crucial in developing and characterising iPSCs and their derived cell types.
Recurrent tonsillitis (RT) and obstructive sleep apnoea (OSA) are major indications for tonsillectomy, yet their tissue-level pathology remains unclear. Recurrent inflammation in RT may induce collagen remodelling through excess extracellular matrix deposition. Here, we used label-free second harmonic generation (SHG) microscopy to quantify the area fraction of the ratio of forward-to-backward SHG signals in palatine tonsils from RT and OSA patients. The ratio of forward-to-backward SHG signals showed a higher mean area fraction in OSA when compared to RT, agreeing with trends previously found using immunohistochemical staining methods; however, no statistical significance was found. Spatial analysis of epithelial dominant regions and interior regions of tissue show higher forward/backward SHG signal in OSA rather than RT; however, no statistical significance was found in either measurement. These results demonstrate a new method of using SHG microscopy to measure collagen distribution within palatine tonsils of OSA and RT.
Grain identification in polycrystalline nanoparticles, for example, determining which crystal phases are present at each spatial location, is fundamental to materials characterisation. This is particularly challenging when grains overlap extensively, as commonly occurs in four-dimensional scanning transmission electron microscopy (4D-STEM) datasets. We propose a simplex-based model (SBM) in which each simplex vertex represents the diffraction pattern (DP) of a pure grain, and the simplex edges and interior represent overlapping grains. Our SBM grain identification algorithm operates on the Bragg disk (BD) data matrix distilled from the 4D-STEM data to identify the grain membership at each scan position, together with a BD feature matrix whose columns represent the DPs for each constituent grain, which is important for identifying the crystal structure of each grain. We solve the model using a two-stage algorithm. In Stage 1, we adapt a linear mixing algorithm to estimate an initial BD feature matrix whose columns represent DPs of potentially overlapping grains. Our Stage 2 algorithm incorporates sparsity considerations to transform the initial BD feature matrix so that its columns represent DPs of pure grains. Using simulated datasets with various grain configurations, we demonstrate that SBM recovers both the BD feature matrix and membership maps more accurately than existing methods, even when a grain lacks any pure region and completely overlaps with other grains.
This thematic contribution highlights the critical role of established professional training regimes in microscopy and microanalysis in supporting high-quality research. As microscopy technologies continue to evolve in complexity, the risk of mishandling and data misinterpretation by inexperienced researchers increases significantly. To address this challenge, the authors underscore the importance of expert training and ongoing support provided by platform scientists-specialist microscopists embedded within centralised research infrastructure facilities. These experts play a vital role in equipping researchers with the necessary skills to use advanced instruments effectively and interpret data accurately. Collaboration between platform scientists and researchers is not only beneficial but essential for maximising the capabilities of modern microscopy. By fostering such partnerships and investing in structured training solutions, institutions can ensure that scientific discoveries are grounded in reliable, high-quality imaging data, advancing the broader goals of research excellence. At the microscopy core facility in Sydney, we have found that no single training method can fully substitute the comprehensive researcher experience required to become an expert microscopy scientist. Therefore, we suggest that a hybrid approach-combining multiple training tools with regular researcher check-ins throughout the research project lifecycle-is the most effective solution.
