Combining multiple staining in histopathology is vital to relate complementary information within a tissue. We present a novel framework that enables local correspondence in the form of patch retrieval (CBIR) in multi-stain whole-slide images without the need for explicit registration. By leveraging foundation models as meaningful patch encoders, we are able to retrieve an H&E patch in a dataset of patches extracted from IHC slides. Using Fused Gromov-Wasserstein optimal transport for weak supervision, we learn a lightweight correction module that aligns embeddings and improves retrieval accuracy by up to 30% (depending on patch size and model), at negligible cost, despite staining and structural variations. Our method paves the way for efficient multi-stain integration in digital pathology workflows.

We present the Shareish web platform to foster diverse solidarity practices inspired by concepts of the gift economy. Its design is grounded in prior work (in CSCW and solidarity HCI), and in new qualitative research involving participants engaged in solidarity practices. Shareish aims at leveraging community assets through donations, free loans, requests of goods and services, free event announcements, and by enhancing the visibility of freely available resources. On a Shareish instance, users have the ability to view localized items on a map or in lists, utilize search filters, contribute new content with detailed textual and visual descriptions, engage in discussions with other users, and receive notifications when new content is added in their neighborhood. The platform can be replicated and improved by communities seeking autonomy as its source code is distributed freely under a permissive open source license (https://github.com/shareish). A research demonstration server is available (https://shareish.org).

Detecting skeletal or bone-related deformities in model and aquaculture fish is vital for numerous biomedical studies. In biomedical research, model fish with bone-related disorders are potential indicators of various chemically induced toxins in their environment or poor dietary conditions. In aquaculture, skeletal deformities are affecting fish health, and economic losses are incurred by fish farmers. This survey paper focuses on showcasing the cutting-edge image analysis tools and techniques based on artificial intelligence that are currently applied in the analysis of bone-related deformities in aquaculture and model fish. These methods and tools play a significant role in improving research by automating various aspects of the analysis. This paper also sheds light on some of the hurdles faced when dealing with high-content bioimages and explores potential solutions to overcome these challenges.

Data scarcity is a common issue when training deep learning models for digital pathology, as large exhaustively-annotated image datasets are difficult to obtain. In this paper, we propose a self-training based approach that can exploit both (few) exhaustively annotated images and (very) sparsely-annotated images to improve the training of deep learning models for image segmentation tasks. The approach is evaluated on three public and one in-house dataset, representing a diverse set of segmentation tasks in digital pathology. The experimental results show that self-training allows to bring significant model improvement by incorporating sparsely annotated images and proves to be a good strategy to relieve labeling effort in the digital pathology domain.

In this paper we perform an empirical evaluation of variants of deep learning methods to automatically localize anatomical landmarks in bioimages of fishes acquired using different imaging modalities (microscopy and radiography). We compare two methodologies namely heatmap based regression and multivariate direct regression, and evaluate them in combination with several Convolutional Neural Network (CNN) architectures. Heatmap based regression approaches employ Gaussian or Exponential heatmap generation functions combined with CNNs to output the heatmaps corresponding to landmark locations whereas direct regression approaches output directly the (x, y) coordinates corresponding to landmark locations. In our experiments, we use two microscopy datasets of Zebrafish and Medaka fish and one radiography dataset of gilthead Seabream. On our three datasets, the heatmap approach with Exponential function and U-Net architecture performs better. Datasets and open-source code for training and prediction are made available to ease future landmark detection research and bioimaging applications.

In this paper, we introduce the Shareish web platform to foster mutual aid following principles of gift economy and generalized exchange. Its design is grounded in prior work (in C&T, CSCW, and solidarity HCI) and it aims at leveraging community assets through donation, free loan, requests of goods and services, and free event announcements. Authenticated users can visualize localized items on a map or through lists, search with filters, add new content with rich textual and visual descriptions, discuss about specific content with others users, and get notifications when new content is created in their neighborhood. In addition, we evaluate AI technologies to ease content creation. The platform can be easily replicated and improved by grassroots movements or researchers seeking autonomy as its source code is made freely available and its installation relies on modern deployment strategies. A demonstration server is available (https://shareish.org/, see Section Online Resources).

As a part of modern technological environments, virtual microscopy enriches histological learning, with support from large institutional investments. However, existing literature does not supply empirical evidence of its role in improving pedagogy. Virtual microscopy provides fresh opportunities for investigating user behavior during the histology learning process, through digitized histological slides. This study establishes how students' perceptions and user behavior data can be processed and analyzed using machine learning algorithms. These also provide predictive data called learning analytics that enable predicting students' performance and behavior favorable for academic success. This information can be interpreted and used for validating instructional designs. Data on the perceptions, performances, and user behavior of 552 students enrolled in a histology course were collected from the virtual microscope, Cytomine®. These data were analyzed using an ensemble of machine learning algorithms, the extra-tree regression method, and predictive statistics. The predictive algorithms identified the most pertinent histological slides and descriptive tags, alongside 10 types of student behavior conducive to academic success. We used these data to validate our instructional design, and align the educational purpose, learning outcomes, and evaluation methods of digitized histological slides on Cytomine®. This model also predicts students' examination scores, with an error margin of <0.5 out of 20 points. The results empirically demonstrate the value of a digital learning environment for both students and teachers of histology.

Workflows are the keystone of bioimage analysis, and the NEUBIAS (Network of European BioImage AnalystS) community is trying to gather the actors of this field and organize the information around them. One of its most recent outputs is the opening of the F1000Research NEUBIAS gateway, whose main objective is to offer a channel of publication for bioimage analysis workflows and associated resources. In this paper we want to express some personal opinions and recommendations related to finding, handling and developing bioimage analysis workflows. The emergence of "big data” in bioimaging and resource-intensive analysis algorithms make local data storage and computing solutions a limiting factor. At the same time, the need for data sharing with collaborators and a general shift towards remote work, have created new challenges and avenues for the execution and sharing of bioimage analysis workflows. These challenges are to reproducibly run workflows in remote environments, in particular when their components come from different software packages, but also to document them and link their parameters and results by following the FAIR principles (Findable, Accessible, Interoperable, Reusable) to foster open and reproducible science. In this opinion paper, we focus on giving some directions to the reader to tackle these challenges and navigate through this complex ecosystem, in order to find and use workflows, and to compare workflows addressing the same problem. We also discuss tools to run workflows in the cloud and on High Performance Computing resources, and suggest ways to make these workflows FAIR.

In this paper, we propose variants of deep learning methods to segment head and operculum of the zebrafish larvae in microscopy images. In the first approach, we used a three-class model to jointly segment head and operculum area of zebrafish larvae from background. In the second, two-step, approach, we first trained binary segmentation model to segment head area from the background followed by another binary model to segment the operculum area within cropped head area thereby minimizing the class imbalance problem. Both of our approaches use a modified, simpler, U-Net architecture, and we also evaluate different loss functions to tackle the class imbalance problem. We systematically compare all these variants using various performance metrics. Data and open-source code are available at https://uliege.cytomine.org.

In this work, we investigate multi-task learning as a way of pre-training models for classification tasks in digital pathology. It is motivated by the fact that many small and medium-size datasets have been released by the community over the years whereas there is no large scale dataset similar to ImageNet in the domain. We first assemble and transform many digital pathology datasets into a pool of 22 classification tasks and almost 900k images. Then, we propose a simple architecture and training scheme for creating a transferable model and a robust evaluation and selection protocol in order to evaluate our method. Depending on the target task, we show that our models used as feature extractors either improve significantly over ImageNet pre-trained models or provide comparable performance. Fine-tuning improves performance over feature extraction and is able to recover the lack of specificity of ImageNet features, as both pre-training sources yield comparable performance.

Image analysis is key to extracting quantitative information from scientific microscopy images, but the methods involved are now often so refined that they can no longer be unambiguously described by written protocols. We introduce BIAFLOWS, an open-source web tool enabling to reproducibly deploy and benchmark bioimage analysis workflows coming from any software ecosystem. A curated instance of BIAFLOWS populated with 34 image analysis workflows and 15 microscopy image datasets recapitulating common bioimage analysis problems is available online. The workflows can be launched and assessed remotely by comparing their performance visually and according to standard benchmark metrics. We illustrated these features by comparing seven nuclei segmentation workflows, including deep-learning methods. BIAFLOWS enables to benchmark and share bioimage analysis workflows, hence safeguarding research results and promoting high-quality standards in image analysis. The platform is thoroughly documented and ready to gather annotated microscopy datasets and workflows contributed by the bioimaging community.

In addition, contributors who also aim at individually Pubmed indexed articles have to submit an extended abstract (less than 1500 words). Selected articles will be published: either in "Diagnostic Pathology" (editor-in-chief Klaus Kayser, co-chairs Catherine Bor, Philippe Camparo and Myriam Oger) or for computer science orientated articles in the "Computerized Medical Imaging and Graphics" Elsevier journal (editor-in-chief Daniel Racoceanu, co-chair Philippe Belhomme)

Air pollution is becoming a major health problem since it is responsible for millions of deaths related to cardiovascular and lung diseases. Particulate matter and gazes such as ozone produced have been shown to increase lung cancer morbidity in industrialized countries. However, no clear link between air quality levels and cancer cell dissemination to lung tissue has yet been established. A reliable murine model of concomitant pulmonary O3 exposure and tumor cell injection was used to evaluate metastatic burden in the lungs after pulmonary ozone exposure. The implication of neutrophils in this process was studied by using anti-Ly6G antibodies to prevent recruitment of neutrophils to the lungs. Additionally, the implication of neutrophil extracellular traps (NETs) in metastatic processes was evaluated using MRP8cre-Pad4lox/lox mice or by treating mice with DNase I. Pulmonary ozone exposure induces 1) a strong inflammatory response in lung tissues characterized by the recruitment of neutrophils and, 2) colonization of lung tissues by cancer cells and this already at early steps of the metastatic dissemination process. Additionally, an increased production of NETs by ozone-primed neutrophils was observed. Interestingly, neutrophil depletion and inhibition of NET formation greatly diminished the metastatic burden in lungs of mice exposed to ozone. The ability of O3-primed neutrophils to enhance lung colonization by tumor cells was further confirmed after their adoptive transfer in Balb/c mice unexposed to O3. Pulmonary neutrophils induced by O3 promote metastatic dissemination to lungs by producing NETs. These findings open new perspectives to improve treatment and prevention strategies in patients affected by metastatic diseases.

BACKGROUND:Air pollution, including particulates and gazes such as ozone (O3), is detrimental for patient's health and has repeatedly been correlated to increased morbidity and mortality in industrialised countries. Although studies have described a link between ambient particulate matter and increased lung cancer morbidity, no direct relation has yet been established between O3 exposure and metastatic dissemination to lungs.OBJECTIVES:To outline the mechanisms through which pulmonary O3 exposure modulates metastasis kinetics in an experimental mouse model of O3 exposure.METHODS:Metastatic responses to pulmonary O3 exposure were assessed using a reliable experimental mouse model of concomitant pulmonary O3 exposure and tumour cell injection. Roles of neutrophils in O3-induced lung metastasis were highlighted using blocking anti-Ly6G antibodies; moreover, the implication of neutrophil extracellular traps (NETs) in metastatic processes was evaluated using MRP8cre-Pad4lox/lox mice or by treating mice with DNase I.RESULTS:Pulmonary O3 exposure strongly facilitates the establishment of lung metastasis by (1) Inducing a pulmonary injury and neutrophilic inflammation, (2) Influencing very early steps of metastasis, (3) Priming neutrophils' phenotype to release NETs that favour tumour cell colonisation in lungs. The ability of O3-primed neutrophils to enhance lung colonisation by tumour cells was proven after their adoptive transfer in Balb/c mice unexposed to O3.CONCLUSIONS:Pulmonary neutrophils induced by O3 promote metastatic dissemination to lungs by producing NETs. These findings open new perspectives to improve treatment and prevention strategies in patients affected by metastatic diseases.

In this paper, we describe open practices and open resources in the field of digital pathology with a specific focus on approaches that ease collaboration in research and education settings. Our review includes open access journals and open peer review, open-source software (libraries, desktop tools, and web applications), and open access collections. We illustrate applications and discuss current limitations and perspectives.

PURPOSE:Digital histology is being increasingly used in research and clinical applications. In parallel, new tissue imaging methods (e.g., imaging mass spectrometry) are currently regarded as very promising approaches for better molecular diagnosis in pathology. However, these new data sources are still often underexploited because of the lack of collaborative software to share and correlate information for multimodal analysis.EXPERIMENTAL DESIGN:The open science paradigm is followed to develop new features in the web-based Cytomine software to support next-generation digital pathology bridged to molecular investigations.RESULTS:New open-source developments allow to explore whole-slide classical histology with Matrix Assisted Laser Desorption Ionisation (MALDI) imaging and to support preprocessing for biomarker discovery using laser microdissection-based microproteomics.CONCLUSIONS AND CLINICAL RELEVANCE:The updated version of Cytomine is the first open and web-based tool to enable sharing data from classical histology, molecular imaging, and cell counting for proteomics preprocessing. It holds good promise to fulfill imminent needs in molecular histopathology.

The detection of anatomical landmarks in bioimages is a necessary but tedious step for geometric morphometrics studies in many research domains. We propose variants of a multi-resolution tree-based approach to speed-up the detection of landmarks in bioimages. We extensively evaluate our method variants on three different datasets (cephalometric, zebrafish, and drosophila images). We identify the key method parameters (notably the multi-resolution) and report results with respect to human ground truths and existing methods. Our method achieves recognition performances competitive with current existing approaches while being generic and fast. The algorithms are integrated in the open-source Cytomine software and we provide parameter configuration guidelines so that they can be easily exploited by end-users. Finally, datasets are readily available through a Cytomine server to foster future research.

Mesenchymal stromal cells (MSCs) are multipotent stem cells with immunosuppressive and trophic support functions. While MSCs from different sources frequently display a similar appearance in culture, they often show differences in their surface marker and gene expression profiles. Although bone marrow is considered the "gold standard" tissue to isolate classical MSCs (BM-MSC), MSC-like cells are currently also derived from more easily accessible extra-embryonic tissues such as the umbilical cord. In this study, we defined the best way to isolate MSCs from the Wharton's jelly of the human umbilical cord (WJ-MSC) and assessed the mesenchymal and immunological phenotype of BM-MSC and WJ-MSC. Moreover, the gene expression profile of established WJ-MSC cultures was compared to two different bone marrow-derived stem cell populations (BM-MSC and multipotent adult progenitor cells or MAPC®). We observed that explant culturing of Wharton's jelly matrix is superior to collagenase tissue digestion for obtaining mesenchymal-like cells, with explant isolated cells displaying increased expansion potential. While being phenotypically similar to adult MSCs, WJ-MSC show a different gene expression profile. Gene ontology analysis revealed that genes associated with cell adhesion, proliferation, and immune system functioning are enriched in WJ-MSC. In vivo transplantation confirms their immune modulatory effect on T cells, similar to BM-MSC and MAPC. Furthermore, WJ-MSC intrinsically overexpress genes involved in neurotrophic support and their secretome induces neuronal maturation of SH-SY5Y neuroblastoma cells to a greater extent than BM-MSC. This signature makes WJ-MSC an attractive candidate for cell-based therapy in neurodegenerative and immune-mediated central nervous system disorders such as multiple sclerosis, Parkinson's disease, or amyotrophic lateral sclerosis.

A goal ! The MOOC entitled "Introduction to Histology, A Human Tissue Exploration" correspond to our vision of the practice of General Histology, which is based on the ability to diagnose 5 families of biological tissues. Ultimately, participants must be able to recognize the different types of cells and all the surrounding elements in order to understand how they organize themselves to form tissues with specific functions. A tool ! This know-how is based on reasoning from observations of microscopic structures. Learners are therefore invited to manipulate a virtual microscope to explore biological samples on histological slides digitized. Annotations, comments, drawings or photos are associated with landmarks that enrich the study of these histological sections. A target audience ! Two educational paths allow deepening the subject in a different way and thus matching the goals or motivations of each one. After a first year of experience, usage statistics and surveys of our learners show that the MOOC Histo has allowed each of them to find an interest and federate a community of motivated learners.

In this paper, we study deep transfer learning as a way of overcoming object recognition challenges encountered in the field of digital pathology. Through several experiments, we investigate various uses of pre-trained neural network architectures and different combination schemes with random forests for feature selection. Our experiments on eight classification datasets show that densely connected and residual networks consistently yield best performances across strategies. It also appears that network fine-tuning and using inner layers features are the best performing strategies, with the former yielding slightly superior results.