cs.CVFeb 24, 2026

Region of Interest Segmentation and Morphological Analysis for Membranes in Cryo-Electron Tomography

Authors: Xingyi Cheng, Julien Maufront, Aurélie Di Cicco, Daniël M. Pelt, Manuela Dezi, Daniel Lévy

Organizations: Institut Curie, CNRS UMR168, Université PSL, Sorbonne Université, Paris, France · Inria Center at University of Rennes, SAIRPICO team, Rennes, France · Leiden Institute of Advanced Computer Science (LIACS), Universiteit Leiden, the Netherlands

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Cryo-electron tomography (cryo-ET) enables high resolution, three-dimensional reconstruction of biological structures, including membranes and membrane proteins. Identification of regions of interest (ROIs) is central to scientific imaging, as it enables isolation and quantitative analysis of specific structural features within complex datasets. In practice, however, ROIs are typically derived indirectly through full structure segmentation followed by post hoc analysis. This limitation is especially apparent for continuous and geometrically complex structures such as membranes, which are segmented as single entities. Here, we developed TomoROIS-SurfORA, a two step framework for direct, shape-agnostic ROI segmentation and morphological surface analysis. TomoROIS performs deep learning-based ROI segmentation and can be trained from scratch using small annotated datasets, enabling practical application across diverse imaging data. SurfORA processes segmented structures as point clouds and surface meshes to extract quantitative morphological features, including inter-membrane distances, curvature, and surface roughness. It supports both closed and open surfaces, with specific considerations for open surfaces, which are common in cryo-ET due to the missing wedge effect. We demonstrate both tools using in vitro reconstituted membrane systems containing deformable vesicles with complex geometries, enabling automatic quantitative analysis of membrane contact sites and remodeling events such as invagination. While demonstrated here on cryo-ET membrane data, the combined approach is applicable to ROI detection and surface analysis in broader scientific imaging contexts.

Figures & tables

Figure 1: Overview of the TomoROIS-SurfORA workflow. An asterisk marks steps with optional GUI based curation, and dotted lines indicate optional workflow steps.
Figure 2: Region of interest segmentation with TomoROIS for MCS. A: Manually annotated ROIs of two interacting membranes of vesicles and tubes bridged with protein complexes. B: Predicted ROIs generated by MSDCN are shown on the same tomogram slice. Pixel-wise prediction confidence is visualised using a colour scale, where red corresponds to the highest confidence, gradually shifting through yellow and green to blue with decreasing confidence. C: A confidence threshold was applied to retain only pixels predicted with high confidence as ROIs. Correct predictions are highlighted in blue, while false predictions are shown in red, including a ‘vesicle-in-vesicle’ region that was incorrectly classified as a MCS. D: The retained blue regions were first separated into non-connected components and then further divided using a watershed-based splitting step implemented in TomoROIS to separate touching ROIs. Each resulting ROI is displayed in a distinct colour. Scale bar: 50nm.
Figure 3: Membrane segmentation followed by surface analysis with SurfORA. A: Membrane segmentation extracted using refined ROIs shown in Fig. 2D. Segmented membranes are colour-coded by ROI identity, with the red square indicating the example region used to demonstrate SurfORA utilities. B: Surface point clouds corresponding to contacting regions of two lipid vesicles and one tube displayed as single-layer dense point sets. Non-connected components are coloured green, blue, and orange. Consistently oriented surface normals are shown as pink vectors, with vesicle normals pointing toward the tube and tube normals pointing toward the vesicles. C: Mesh generated from the smaller vesicle membrane. D Inter-membrane distance mapping between the vesicle (colour-mapped from yellow to blue by separation distance) and tube (grey) meshed surfaces. E: Distribution of averaged inter-membrane distances across >350 contact regions from 50 tomograms. The average distance is shown as a dark blue dot, with ±\pm 1 standard deviation bands in light blue. Scale bar: 50 nm.
Figure 4: Direct detection of membrane invagination with TomoROIS and surface analysis with SurfORA. A: Membrane invagination predicted by TomoROIS. Invaginating vesicles with openings aligned along either the Z or X/Y axes are shown. Each invagination event is uniquely colour-coded. B: Membrane segmentation within ROIs containing invaginating membranes. C, D: Mean curvature maps for medial surfaces of two example invaginating membranes numbered 1, 2 in B. Outer (E) and inner (F) isosurface of invagination numbered 2 with their respective curvature mapping. The colour scale for positive and negative curvature is provided for each mapping plot. Scale bar: 50 nm.
Figure S1: Interactive visualisation and curation of point cloud with normals. Manual curation of membrane point clouds and their associated normals in interactive GUI. Case 1: inversion of normals of a labeled object. A: non-corrected normals of the tube (label 1, blue points) and the vesicle (label 2, orange points) are in pink. B: normals of label 2 after flipping shown in red. Case 2: local curation of normals. C: Selection of normals displaying wrong orientation within a manually drawn polygon in dark blue. D: Selected normals after flipping shown in red.
Figure S2: Medial and isosurface extraction of an invaginated vesicle. Membrane segmentation and corresponding surface representations for a representative invaginated vesicle. A: Membrane segmentation overlaid on the tomogram. B: Extracted medial surface shown in blue, and split isosurfaces corresponding to the inner (green) and outer (red) membrane boundaries, overlaid on the membrane segmentation. C: The extracted medial surface (blue) and inner (green) and outer (red) isosurfaces overlaid directly on the tomogram. Scale bar: 5 nm.
More in this topic

Image Segmentation

Image segmentation assigns pixels or voxels to objects and regions, supporting measurement, scene understanding, and decisions across image domains.

See all 390 papers

Explore similar work

CardsList
Aug 31, 2026

TopoFuse: Topology-Aware Tri-Planar Fusion for 3D Cryo-Electron Tomography Segmentation

Automated segmentation of cryo-electron tomograms routinely produces masks that are voxel-accurate but topologically broken: membranes fragment, organelles merge into one another, and enclosed cavities collapse. Existing topology-aware losses reduce these violations but cannot eliminate them, because topology is encouraged through gradient pressure rather than structurally enforced. We introduce TopoFuse, which reframes topology as a differentiable projection operator rather than a loss penalty. At each forward pass, the projection operator ProjT\mathrm{Proj}_T (a PH-guided sparse edit) identifies the critical voxels responsible for topological violations via bottleneck matching and applies sparse edits to satisfy a specified topology target (diagram feature counts and lifetime budgets) for dimensions d∈{0,2}d \in \{0,2\}. If the projection converges, the output satisfies those constraints on the downsampled grid (s=2s=2); when it does not, a repair certificate exposes this explicitly, enabling downstream filtering. A topology prior head predicts the correction target directly from input features, removing any dependence on ground-truth topology at inference. Across three cryo-ET benchmarks, TopoFuse reduces Betti number error by 54% over the strongest soft-loss baseline (p<0.001p < 0.001), improves Dice by 4.6 points, and edits only 3.1% of voxels to achieve this.

Rohit Kumar Salla, Neelesh Gupta, Xingjian Li +1

Virginia Tech, Blacksburg, VA, USA · National Institute of Technology, Kurukshetra, India · Carnegie Mellon University, Pittsburgh, PA, USA

Topology-Preserving Image SegmentationImage Segmentation3D Medical Image SegmentationCryo-Electron MicroscopyElectron MicroscopyTopological Data Analysis
AskPDF
Jun 8, 2026

POPSICLE: Benchmark Datasets for Segmentation and Localization in CryoET

Cryo-electron tomography (cryoET) has emerged as a powerful tool in structural and cellular biology by enabling direct visualization of macromolecular structures within intact cells, thereby linking molecular architecture to cellular organization in a native context. Realizing the full potential of cryoET, however, increasingly depends on advances in computational analysis, particularly machine learning (ML), to interpret its complex and information-rich data. Despite rapid progress, ML development for cryoET remains bottlenecked by the lack of standardized, well-annotated benchmarks. Existing evaluations are typically small, task-specific, and are assembled in isolation, limiting robust comparisons across methods. Here, we present POPSICLE, a benchmark suite for cryoET segmentation and macromolecular localization built from the CryoET Data Portal - an open, ML-ready repository of tomographic data, metadata, and annotations. POPSICLE spans eukaryotic and prokaryotic systems, both purified and fully in situ samples, and dense voxel-wise segmentation as well as sparse localization tasks. Built on a living data resource, it can expand as new datasets and annotations become available. Baseline experiments reveal substantial variation in model rankings across tasks, underscoring the need for benchmarks tailored to the unique characteristics of cryoET rather than evaluation practices adapted from adjacent biomedical imaging domains. POPSICLE thus provides an open and extensible foundation for reproducible ML evaluation in cryoET.

Jonathan Schwartz, Utz Heinrich Ermel, C. Braxton Owens +6

1Biohub, Redwood City, CA 94063, USA · 2Brigham Young University, Provo, UT, 84602, USA

Image SegmentationCryo-Electron MicroscopyMicroscopy Image AnalysisElectron Microscopy
AskPDF
Oct 15, 2026

Learning the Shared Latent Structure of Cryo-EM Inverse Problems

Understanding biomolecular function requires characterizing three-dimensional structure and conformational variability, for which cryogenic electron microscopy (cryo-EM) is a powerful tool. Cryo-EM produces noisy two-dimensional particle images in which molecular structure, particle pose, and imaging conditions are jointly encoded. Recovering biological information from these observations spans inverse problems including pose estimation, heterogeneity analysis, and reconstruction. These tasks are usually addressed independently despite shared protein structure and image formation, motivating a common representation across inverse problems. Since biologically meaningful differences may be spatially localized, we study self-distillation with complementary global and masked patch objectives. Because experimental dataset discrimination can reflect acquisition-specific cues, we develop controlled synthetic and experimental evaluations and find that structural discrimination benefits from dense spatial features. We introduce the EMber family of vision transformers, pre-trained on >50 million experimental particle images from 214 EMPIAR datasets together with simulated projections from 268,000 molecular assemblies. EMber improves upon existing methods at conformational and compositional discrimination by 14.85 and 14.26 percentage points, estimates conformational populations to within 1.59pp error with a linear probe, and produces particle assignments yielding reference-quality reconstructions in fewer epochs. We demonstrate how to learn representations that capture biological information for use across multiple cryo-EM inverse problems.

Nithin Chalapathi, Theo Sternlieb, Ethan Han +4

University of California, Berkeley · Flatiron Institute · University of Maryland, College Park

Cryo-Electron MicroscopyImage Inverse ProblemsMasked Image Modeling
AskPDF