The arterial wall consists of three concentric layers: intima, media, and adventitia. Beyond their resident cells, these layers are characterized by an extracellular matrix (ECM), which provides both biochemical and mechanical support. Elastin, the major component of arterial ECM, is present in the medial layer and organized in concentric elastic lamellae that confer resilience to the wall. We explored the arterial wall structures from C57Bl6 (control), db / db (diabetic), and ApoE ^−/− (atherogenic) mice aged 3 months using synchrotron X‐ray computed microtomography on fixed and unstained tissues with a large image field (8 mm ³ ). This approach combined a good resolution (0.83 µm/voxel), large 3D imaging field. and an excellent signal to noise ratio conferred by phase‐contrast imaging. We determined from 2D virtual slices that the thickness of intramural ECM structures was comparable between strains but automated image analysis of the 3D arterial volumes revealed a lattice‐like network within concentric elastic lamellae. We hypothesize that this network could play a role in arterial mechanics. This work demonstrates that phase‐contrast synchrotron X‐ray computed microtomography is a powerful technique which to characterize unstained soft tissues.

Diabetes is a major concern of our society as it affects one person out of 11 around the world. Elastic fiber alterations due to diabetes increase the stiffness of large arteries, but the structural effects of these alterations are poorly known. To address this issue, we used synchrotron X-ray microcomputed tomography with in-line phase contrast to image in three dimensions C57Bl6J (control) and db/db (diabetic) mice with a resolution of 650 nm/voxel and a field size of 1.3 mm³. Having previously shown in younger WT and db/db mouse cohorts that elastic lamellae contain an internal supporting lattice, here we show that in older db/db mice the elastic lamellae lose this scaffold. We coupled this label-free method with automated image analysis to demonstrate that the elastic lamellae from the arterial wall are structurally altered and become 11% smoother (286,665 measurements). This alteration suggests a link between the loss of the 3D lattice-like network and the waviness of the elastic lamellae. Therefore, waviness measurement appears to be a measurable elasticity indicator and the 3D lattice-like network appears to be at the origin of the existence of this waviness. Both could be suitable indicators of the overall elasticity of the aorta.

U-Net is a convolutional neural network model developed in 2015 and has proven to be one of the most inspiring deep-learning models for image segmentation. Numerous U-Net-based applications have since emerged, constituting a heterogeneous set of tools that illustrate the current reproducibility crisis in the deep-learning field and that remain slow to spread in application fields such as 3D bioimaging. Here we propose a solution named Biom3d, a modular deep learning framework facilitating the integration and development of novel models, metrics, or training schemes for 3D image segmentation. The development philosophy of Biom3d provides an improved code sustainability and reproducibility in line with the FAIR principles and is available as a graphical user interface and an open-source deep-learning framework to target a large community of users, from end users to deep learning developers.

Bioimage frameworks based on artificial intelligence (AI) offer powerful tools for image segmentation, but their technical overhead often creates a gap between developers and the broader bioimaging community. Biom3d addresses this challenge by providing a modular, PyTorch-based architecture designed to enable reproducible and interoperable 2D and 3D segmentation pipelines while maintaining strict adherence to FAIR principles. The framework's engineering quality is demonstrated through low intra-module complexity, high modularity, and seamless interoperability, exemplified by the successful substitution of transformer-based MONAI models. Its architecture is organized around seven core module types, allowing fine-grained control over data handling, model configuration, optimization, and evaluation. Its default configuration, nnCore, autoconfigures optimal pipelines based on new datasets and competes with state-of-the-art 3D segmentation methods, outperforming classical tools such as NucleusJ/NODeJ and matching the robustness of nnU-Net across diverse modalities. Biom3d is accessible through multiple interfaces including a graphical user interface, a Jupyter Notebook, a command-line tool, a Docker image and as a Python library. Biom3d is compatible with OMERO (Open Microscopy Environment Remote Objects) for streamlined data management and reproducible computation, including execution on HPC servers. Collectively, these results position Biom3d as a sustainable and extensible framework for building, sharing, and reusing advanced bioimage analysis solutions.

Over the past decades, the understanding of peptides and proteins biological functions has been an active research topic. Latest research works in this field have suggested that protein conformations may be a key feature for gaining insights into protein biological functions. However, analyzing small and highly flexible protein chunks, namely oligopeptides made of a handful of amino acids, remains challenging because of their dynamics and wide range of conformations. In this paper, a statistical methodology based on unsupervised statistical learning is proposed for analyzing 3D conformations small and highly flexible elastin-derived peptides. The goal of this study is twofold: first, is it aimed at identifying the most frequent conformations of each peptide and to study their stability. Second, and most important, it is aimed at comparing main conformations of different elastin-derived peptides to identify the “signature” than can be linked to a biological activity. The main strength of the present work is to propose a method for confirmation recognition that is not affected by peptide rotations or translations and, hence, avoids the use of the complex superposition methods. In addition, the proposed approach uses Kernel PCA to eliminate atypical peptide conformations. Due to the instability of those peptides, removing outliers is crucial since they may dramatically impact clustering results. To extract the most frequent conformations, we propose to use a hierarchical clustering method. Eventually, a peptide activity detector is defined based on comparison of main conformation found in different peptides. The main interests of the proposed method are twofold: first, it is fully automatic method, second, it does not require any additional information or expertise and, third, it can identify conformations accurately that make peptides enabling a given biological activity. Experimental results on a large dataset of peptides conformations highlight the relevance and efficiency of the proposed method.