Semantic segmentation has suffered for a while from a lack of datasets such as ImageNet for image classification. This issue was partially alleviated by the advent of the segment anything model (SAM), which provides a foundation model trained on the largest and most diverse segmentation dataset to date. However, the SAM often falls short in segmenting specific regions, mostly in regard to biomedical images; this is why unsupervised domain adaptation (UDA) remains the best option for addressing the challenge of generalization capabilities. Classical UDA methods might be ineffective in several biomedical segmentation cases because the gap between two datasets, named domain shift, is too high. To address this issue, we propose a strategy based on learning the source mask probability distribution with a segmentation diffusion model as a generative prior to propose accurate target segmentation at inference. This latter can be guided by supplementary inputs, which allows us to call for the rich information contained in SAM raw segmentation both to perform adaptation and to improve robustness. A study was conducted using a comprehensive collection of segmentation datasets: 3 domains for mitochondria, 2 for the endoplasmic reticulum, and 2 for brain tumors, allowing the creation of 10 adaptation scenarios and providing an extensive test basis. The results of the experiments reveal that our proposed method outperforms various state-of-the-art UDA methods. Furthermore, ablation studies highlight the significant role of each component of our presented strategy. The code is available at: https://github.com/alex-stenger/GUDA.

Unsupervised Domain Adaptation (UDA) methods have emerged as a promising solution to generalize a learning to close datasets (domains) without the need to produce new ground truth. Nonetheless in biomedical images, some high domain shifts between source and target images lead to poor adaptation. To address this issue, we propose a method relying on two main ideas. First, we learn the source posterior label distribution with diffusion models. Assuming the target label distribution is similar, this learning helps us to guide the diffusion process to generate relevant segmentation masks on target domain. Alongside this probabilistic constraint, we propose a reconstruction pretext task on both source and target domain to extract common images features. Our approach is compared to the state of the art on three highly shifted mitochondria segmentation datasets. Our method ranks among the best in moderately difficult adaptation cases and succeeds in difficult adaptation cases where all other tested methods fail. Code will be available.

Large annotated datasets and new models have led to significant improvements in supervised semantic segmentation. On the other side, Unsupervised Domain Adaptation (UDA) for Semantic Segmentation is still an arduous open research topic. While new ideas frequently come out based on recent findings, best methods still rely on basic techniques such as the use of pseudo-labels on target for self-training. Nonetheless, such methods fail when applied to difficult UDA cases like Biomedical Images where the domain shift is too high, leading to pseudo-labels of poor quality. In this work, we propose RESAMPL-UDA ( RE fined SAM -based P seudo L abel UDA ), an unsupervised domain adaptation method that effectively integrates zero-shot predictions from the Segment Anything (SAM) model. Given the high complexity and variability of biomedical images, SAM alone often produces detailed segmentations without necessarily capturing the intended structures. To address this, our method involves training a dedicated refinement network on source domain data to selectively enhance SAM-generated masks. These refined segmentations then serve as reliable pseudo-labels within our UDA framework, significantly facilitating the adaptation process. Experiments on 8 adaptation cases demonstrate that our method outperforms the state of the art. In addition, we extend successfully our work to Source-Free Unsupervised Domain Adaptation, demonstrating its versatility. The code will be made available.

Abstract. Hierarchical structures provide versatile and efficient solutions for representing, processing and analyzing images. In the framework of mathematical morphology, partial partition trees were proposed in order to model the grey-level images. The most popular are the component tree and the tree of shapes. Both trees are image models, i.e., they represent an image in a lossless, reversible way. Based on this property, they can be used for designing image processing operators. Indeed, by selecting some nodes of these trees and/or by modifying their associated grey-level values, one can define so-called connected operators, that act at the scale of flat zones instead of pixels, and thus avoid the generation of new contours. The definition of connected operators from the component tree has been the subject of an abundant literature. This is not the case of the tree of shapes, despite its high ability to model the topological and differential properties of grey-level images. In this article, we propose an algorithmic scheme to build consistent connected operators based on the tree of shapes. More precisely, the induced connected operators do not modify the differential properties of the image, and they also preserve the equivalence between any image and its tree of shapes. We prove that this algorithmic scheme presents an efficient time cost. We also demonstrate that it generalizes the previous approaches developed for connected operator design from morphological trees. Finally, we show how it opens the way to the development of connected versions of usual pixel-based, linear, and nonlinear operators. Codes freely available at https://github.com/jmendesf/ToSConOp .

Synchrotron X-ray microtomography (µCT) gives access to images with a micrometric resolution. In the context of vascular imaging, this allows the study of structural properties of arterial walls, even for small animals such as the mouse. However, the images available with µCT are non-usual, and there is no method specifically designed for their processing and analysis. This article describes a first pipeline dedicated to the segmentation of µCT images of mice aorta. This pipeline builds upon conventional image processing paradigms and more recent deep learning approaches, and tackles the issue of multiscale analysis of huge-sized, high-resolution data. It provides promising results, assessed by comparison with manual annotation of sampled data. This methodological framework is a step forwards to a finer analysis of the internal structure of the aortic walls, especially for understanding the consequences of ageing and/or disease (e.g. diabetes) on the vessels architecture.

Vessel enhancement (aka vesselness) filters, are part of angiographic image processing for more than twenty years. Their popularity comes from their ability to enhance tubular structures while filtering out other structures, especially as a preliminary step of vessel segmentation. Choosing the right vesselness filter among the many available can be difficult, and their parametrization requires an accurate understanding of their underlying concepts and a genuine expertise. In particular, using default parameters is often not enough to reach satisfactory results on specific data. Currently, only few benchmarks are available to help the users choosing the best filter and its parameters for a given application. In this article, we present a generic framework to compare vesselness filters. We use this framework to compare seven gold standard filters. Our experiments are performed on three public datasets: the hepatic Ircad dataset (CT images), the Bullit dataset (brain MRA images) and the synthetic VascuSynth dataset. We analyse the results of these seven filters both quantitatively and qualitatively. In particular, we assess their performances in key areas: the organ of interest, the whole vascular network neighbourhood and the vessel neighbourhood split into several classes, based on their diameters. We also focus on the vessels bifurcations, which are often missed by vesselness filters. We provide the code of the benchmark, which includes up-to-date C++ implementations of the seven filters, as well as the experimental setup (parameter optimization, result analysis, etc.). An online demonstrator is also provided to help the community apply and visually compare these vesselness filters.

Quasi-regularity has been proved to be a sufficient condition for simple-connectedness preservation during the digitization process of 2D continuous objects. The original definition of quasi-regularity relies on set-based morphological operations of erosion and dilation. With this definition, quasi-regularity is algorithmically difficult to assess. In this paper, we propose a tractable framework for quasi-regularity verification, especially designed for polygons. Our approach mainly relies on the computation and analysis of the medial axis of these objects, and determines their potential quasi-regularity, and thus their ability to undergo a digitization without alteration of their topological properties. The framework is applied in the context of topology-preserving rigid motions of digital objects.

The analysis of axillary and supraclavicular lymph nodes is a primary prognostic factor for the staging of breast cancer. However, due to the size of lymph nodes and the low resolution of PET data, their segmentation is challenging. We investigate the relevance of considering axillary and supraclavicular lymph node segmentation from PET/CT images by coupling Convolutional Neural Networks (CNNs) and Component-Trees (C-Trees). Building upon the U-Net architecture, we propose a framework that couples a multi-modal U-Net fed with PET and CT with a hierarchical model obtained from the PET that provides additional high-level region-based features as input channels. Our working hypotheses are twofold. First, we take advantage of both anatomical information from CT for detecting the nodes, and functional information from PET for detecting the pathological ones. Second, we consider region-based attributes extracted from C-Tree analysis of 3D PET/CT images to improve the CNN segmentation. We carried out experiments on a dataset of 240 pathological lymph nodes from 52 patients scans, and compared our outputs with human expert-defined ground-truth, leading to promising results.

The accurate knowledge of vascular network geometry is crucial for many clinical applications such as cardiovascular disease diagnosis and surgery planning. Vessel enhancement algorithms are often a key step to improve the robustness of vessel segmentation. A wide variety of enhancement filters exists in the literature, but they are often difficult to compare as the applications and datasets differ from a paper to another and the code is rarely available. In this article, we compare seven vessel enhancement filters covering the last twenty years literature in a unique common framework. We focus our study on the liver vascular network which is under-represented in the literature. The evaluation is made from three points of view: the whole liver, the vessel neighborhood and the bifurcations. The study is performed on two publicly available datasets: the Ircad dataset (CT images) and the VascuSynth dataset adapted for MRI simulation. We discuss the strengths and weaknesses of each method in the hepatic context. In addition, the benchmark framework including a C++ implementation of each compared method is provided. An online demonstration ensures the reproducibility of the results without requiring any additional software.

Curvilinear structure restoration in image processing procedures is a difficult task, which can be compounded when these structures are thin, i.e., when their smallest dimension is close to the resolution of the sensor. Many recent restoration methods involve considering a local gradient-based regularization term as prior, assuming gradient sparsity. An isotropic gradient operator is typically not suitable for thin curvilinear structures, since gradients are not sparse for these. In this paper, we propose a mixed gradient operator that combines a standard gradient in the isotropic image regions, and a directional gradient in the regions where specific orientations are likely. In particular, such information can be provided by curvilinear structure detectors (e.g., RORPO or Frangi filters). Our proposed mixed gradient operator, that can be viewed as a companion tool of such detectors, is proposed in a discrete framework and its formulation/computation holds in any dimension; in other words, it is valid in [Formula: see text], n ≥ 1 . We show how this mixed gradient can be used to construct image priors that take edge orientation, as well as intensity, into account, and then involved in various image processing tasks while preserving curvilinear structures. The experiments carried out on 2D, 3D, real, and synthetic images illustrate the relevance of the proposed gradient, and its use in variational frameworks for both denoising and segmentation tasks.

Inter / intra operator errors and high-time consumption induced by manual delineation, are the main drawbacks nowadays in clinical PET tumor segmentation. Several methodologies have been proposed to automate this task. However, there is not yet a validated general protocol to use in clinical routine. Multimodality imaging has been shown to provide good performance, taking into account both functional and anatomical scopes together for segmentation decision. In this context, the involved images used are generally required to be spatially corresponding. However, this is not always the case due to acquisition constraints or for multidate follow-up. In this work, we propose a spatially independent algorithm that avoids image pre-processing (e.g. image registration) or acquisition adjustments for multimodal segmentation. In particular, non-spatially correspondent images (such as multitemporal ones) can be directly exploited taking advantage of hierarchical image structure properties. Regions, obtained from hierarchical models of images, are co-evaluated to match similar ones such as tumors on PET and CT. Results show good performance in terms of time-computing and robust-nesses dealing with PET/CT segmentation problems such as necrosis, compared with other methodologies.

Volume data can be represented by voxels. In many applications of computer graphics (e.g. animation, simulation) and image processing (e.g. shape registration), such voxel data require manipulations. Among the simplest manipulations, we are interested in rigid motions, namely motions that do not change the shape of voxel objects but do change their position and orientation. Such motions are well-known as isometric transformations in continuous spaces. However, when they are applied on voxel data, some important properties of geometry and topology are generally lost. In this article, we discuss this issue, and we provide a method for rigid motions of voxel objects that preserves the global convexity properties of objects, with digital topology guarantees. This method is based on the standard notion of H-convexity, and a new notion of quasi-regularity.

The analysis of thin curvilinear objects in 3D images is a complex and challenging task. In this article, we introduce a new, non-linear operator, called RORPO (Ranking the Orientation Responses of Path Operators). Inspired by the multidirectional paradigm currently used in linear filtering for thin structure analysis, RORPO is built upon the notion of path operator from mathematical morphology. This operator, unlike most operators commonly used for 3D curvilinear structure analysis, is discrete, non-linear and non-local. From this new operator, two main curvilinear structure characteristics can be estimated: an intensity feature, that can be assimilated to a quantitative measure of curvilinearity; and a directional feature, providing a quantitative measure of the structure's orientation. We provide a full description of the structural and algorithmic details for computing these two features from RORPO, and we discuss computational issues. We experimentally assess RORPO by comparison with three of the most popular curvilinear structure analysis filters, namely Frangi Vesselness, Optimally Oriented Flux, and Hybrid Diffusion with Continuous Switch. In particular, we show that our method provides up to 8 percent more true positive and 50 percent less false positives than the next best method, on synthetic and real 3D images.

The binary partition tree (BPT) is a hierarchical data-structure that models the content of an image in a multiscale way. In particular, a cut of the BPT of an image provides a segmentation, as a partition of the image support. Actually, building a BPT allows for dramatically reducing the search space for segmentation purposes, based on intrinsic (image signal) and extrinsic (construction metric) information. A large literature has been devoted to the construction on such metrics, and the associated choice of criteria (spectral, spatial, geometric, etc.) for building relevant BPTs, in particular in the challenging context of remote sensing. But, surprisingly, there exists few works dedicated to evaluate the quality of BPTs, i.e. their ability to further provide a satisfactory segmentation. In this paper, we propose a framework for BPT quality evaluation, in a supervised paradigm. Indeed, we assume that ground-truth segments are provided by an expert, possibly with a semantic labelling and a given uncertainty. Then, we describe local evaluation metrics, BPT nodes / ground-truth segments fitting strategies, and global quality score computation considering semantic information, leading to a complete evaluation framework. This framework is illustrated in the context of BPT segmentation of multispectral satellite images.

Tubular structure segmentation is an important task, with many applications in medical image analysis such as vessel segmentation both in 2D and 3D. However, this task is challenging due to the spatial sparsity of these objects, implying a high sensitivity to noise. An important cue in this context is the local orientation of the tubular structures. Using this information, it is possible to regularize the structures without destroying its integrity. In this article, we take advantage of recent advances in orientation estimation to propose a directional regularization prior for tubular structures, suitable for use in a variational framework. We illustrate on both synthetic and 2D real data.

Angiographic imaging is a crucial domain of medical imaging. In particular, Magnetic Resonance Angiography (MRA) is used for both clinical and research purposes. This article presents the first framework geared toward the design of virtual MRA images from real MRA images. It relies on a pipeline that involves image processing, vascular modeling, computational fluid dynamics and MR image simulation, with several purposes. It aims to provide to the whole scientific community (1) software tools for MRA analysis and blood flow simulation; and (2) data (computational meshes, virtual MRAs with associated ground truth), in an open-source/open-data paradigm. Beyond these purposes, it constitutes a versatile tool for progressing in the understanding of vascular networks, especially in the brain, and the associated imaging technologies. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Connected operators provide well-established solutions for digital image processing, typically in conjunction with hierarchical schemes. In graph-based frameworks, such operators basically rely on symmetric adjacency relations between pixels. In this article, we introduce a notion of directed connected operators for hierarchical image processing, by also considering non-symmetric adjacency relations. The induced image representation models are no longer partition hierarchies (i.e., trees), but directed acyclic graphs that generalize standard morphological tree structures such as component trees, binary partition trees or hierarchical watersheds. We describe how to efficiently build and handle these richer data structures, and we illustrate the versatility of the proposed framework in image filtering and image segmentation.

International audience

We introduce the new notion of multivalued component-tree, that extends the classical component-tree initially devoted to grey-level images, in the mathematical morphology framework. We prove that multivalued component-trees can model images whose values are hierarchically organized. We also show that they can be efficiently built from standard component-tree construction algorithms, and involved in antiextensive filtering procedures. The relevance and usefulness of multivalued component-trees is illustrated by an applicative example on hierarchically classified remote sensing images.

Mathematical morphology, initially devoted to binary and grey-level image processing, also offers opportunities to develop efficient tools for multivalued - and in particular, colour - images. In this context, connected operators are increasingly considered as a relevant way to obtain such tools, mainly for image filtering and segmentation purposes. In this article, we focus on connected operators based on component-trees and their extension to multivalued images, namely component-graphs. Beyond the classical colour-handling strategies, we show how component-graphs can be algorithmically used to efficiently handle the whole structural information gathered by colour spaces, in order to finally design original image filtering tools.

In recent papers, a new notion of component-graph was introduced. It extends the classical notion of component-tree initially proposed in mathematical morphology to model the structure of gray-level images. Component-graphs can indeed model the structure of any-gray-level or multivalued-images. We now extend the antiextensive filtering scheme based on component-trees, to make it tractable in the framework of component-graphs. More precisely, we provide solutions for building a component-graph, reducing it based on selection criteria, and reconstructing a filtered image from a reduced component-graph. In this paper, we first consider the cases where component-graphs still have a tree structure; they are then called multivalued component-trees. The relevance and usefulness of such multivalued component-trees are illustrated by applicative examples on hierarchically classified remote sensing images.

We provide conditions under which 2D digital images preserve their topological properties under rigid transformations. We consider the two most common digital topology models, namely dual adjacency and well-composedness. This paper leads to the proposal of optimal preprocessing strategies that ensure the topological invariance of images under arbitrary rigid transformations. These results and methods are proved to be valid for various kinds of images (binary, gray-level, label), thus providing generic and efficient tools, which can be used in particular in the context of image registration and warping.

We study the conditions under which the topological properties of a 2D well-composed binary image are preserved under arbitrary rigid transformations. This work initiates a more global study of digital image topological properties under such transformations, which is a crucial but under-considered problem in the context of image processing, e.g., for image registration and warping.

Thin structure filtering is an important preprocessing task for the analysis of 2D and 3D bio-medical images in various contexts. We propose a filtering framework that relies on three approaches that are distinct and infrequently used together: linear, non-linear and non-local. This strategy, based on recent progress both in algorithmic/computational and methodological points of view, provides results that benefit from the advantages of each approach, while reducing their respective weaknesses. Its relevance is demonstrated by validations on 2D and 3D images.

Image classification methods based on histogram analysis generally require to use relevant distances for histogram comparison. In this article, we propose a new distance devoted to compare histograms associated to semantic concepts linked by (dis)similarity correlations. This distance, whose computation relies on a hierarchical strategy, captures the multilevel semantic relations between these concepts. It also inherits from the low complexity properties of standard bin-to-bin distances, thus leading to fast and accurate results in the context of multiresolution image classification. Experiments performed on satellite images emphasize the relevance and usefulness of the proposed distance.

The extraction of complex urban patterns from Very High Spatial Resolution (VHSR) images presents several challenges related to the complexity of the data. Based on the availability of images of a same scene at various resolutions (Medium to Very High Spatial resolutions), a hierarchical approach has been recently proposed to segment/classify objects of interest in a top-down fashion in order to determine patterns from VHSR images. To perform, this method requires the interactive definition of segmentation examples for each considered resolution image. In the context of large dataset processing, such interactive task becomes time consuming. To deal with this issue, we propose in this article, an extension of the domain adaptation paradigm enabling the transfer of the segmentation examples defined on a source dataset to automatically process a target one. Experiments performed on urban images provide satisfactory results which may be further used for operational needs.

The estimation of one-to-one mappings is one of the most intensively studied topics in the research field of nonrigid registration. Although the computation of such mappings can be now accurately and efficiently performed, the solutions for using them in the context of binary image deformation is much less satisfactory. In particular, warping a binary image with such transformations may alter its discrete topological properties if common resampling strategies are considered. In order to deal with this issue, this paper proposes a method for warping such images according to continuous and bijective mappings while preserving their discrete topological properties (i.e., their homotopy type). Results obtained in the context of the atlas-based segmentation of complex anatomical structures highlight the advantages of the proposed approach.

Component-trees provide efficient ways to define filtering-based procedures on grey-level images. We propose an extension of the notion of component-trees to the case of ¿non grey-level¿ images (i.e. images taking their values in partially-ordered sets) including in particular - but not exclusively - colour images. Experiments performed on such images emphasise the interest of the approach.

Segmentation and analysis of blood vessels is an important issue in medical imaging. In 3D cerebral angiographic data, the vascular signal is however hard to accurately detect and can, in particular, be disconnected. In this article, we present a procedure utilising both linear, Hessian-based and morphological methods for blood vessel edge enhancement and reconnection. More specifically, multi-scale second-order derivative analysis is performed to detect candidate vessels as well as their orientation. This information is then fed to a spatially-variant morphological filter for reconnection and reconstruction. The result is a fast and effective vessel-reconnecting method.

Magnetic resonance angiography (MRA) is currently used for cerebral flowing blood visualization. Many segmentation methods have been proposed for brain vessel segmentation, in order to help analyzing the huge data (generally more than 10/sup 7/ voxels) provided by MRA acquisitions. Recently, a new family of segmentation algorithms, involving high level anatomical knowledge, has been studied. These new algorithms require a way to model and store this knowledge. An efficient and general approach to reach that goal consists in using atlases. In this paper a method is proposed to create vascular atlases of the brain, containing information useful for vessel segmentation purpose. This atlas creation process, designed for phase-contrast MRA (PC-MRA), is composed of four steps: segmentation, quantification, registration and data fusion. It uses a region-growing algorithm for vessel segmentation, a skeleton and vessel size determination algorithm, based on discrete geometry, for determination of quantitative properties, and a topology preserving non-rigid registration method to fuse the information. This method, which has been applied to a 18 PC-MRA database, enables to create vascular atlases containing information on brain vessels position, density, size and orientation. The generated atlases are essentially devoted to segmentation purpose but can also be used for anatomical description or pathology detection.

Reliable segmentation of 3D magnetic resonance angiography (MRA) is fundamental for planning and performing neurosurgical procedures, but also for detecting vascular pathologies. We propose here a method for brain vessel segmentation based on mathematical morphology tools. This method, devoted to phase-contrast MRA (PC-MRA) performs vessel segmentation by applying an adaptive set of grey-level hit-or-miss operators on each point of the MR data. High level anatomical knowledge modeled by a vascular atlas is used in order to adapt the parameters of these operators (number, size, and orientation) to the current position. The method has been performed on 30 PC-MRA cases composed of both phase and magnitude images. The results have been validated and compared to segmented data obtained by applying a region-growing algorithm on the same database. They tend to prove that the method is reliable for brain vessel detection and additionally provides information on vessel size and orientation without requiring any post-processing step.