BackgroundThe molecular underpinnings of organ dysfunction in severe COVID-19 and its potential long-term sequelae are under intense investigation. To shed light on these in the context of liver function, we perform single-nucleus RNA-seq and spatial transcriptomic profiling of livers from 17 COVID-19 decedents.ResultsWe identify hepatocytes positive for SARS-CoV-2 RNA with an expression phenotype resembling infected lung epithelial cells, and a central role in a pro-fibrotic TGF beta signaling cell-cell communications network. Integrated analysis and comparisons with healthy controls reveal extensive changes in the cellular composition and expression states in COVID-19 liver, providing the underpinning of hepatocellular injury, ductular reaction, pathologic vascular expansion, and fibrogenesis characteristic of COVID-19 cholangiopathy. We also observe Kupffer cell proliferation and erythrocyte progenitors for the first time in a human liver single-cell atlas. Despite the absence of a clinical acute liver injury phenotype, endothelial cell composition is dramatically impacted in COVID-19, concomitantly with extensive alterations and profibrogenic activation of reactive cholangiocytes and mesenchymal cells.ConclusionsOur atlas provides novel insights into liver physiology and pathology in COVID-19 and forms a foundational resource for its investigation and understanding.

The Spatial Atlas of Human Anatomy (SAHA) represents the first multimodal, subcellular-resolution reference of healthy adult human tissues across multiple organ systems. Integrating spatial transcriptomics, proteomics, and histological features across over 15 million cells from more than 100 donors, SAHA maps conserved and organ-specific cellular niches in gastrointestinal and immune tissues. High-resolution profiling using CosMx SMI, 10x Xenium, RNAscope, GeoMx DSP, and single-nucleus RNA-seq reveals spatially organized cell states, rare adaptive immune populations, and tissue-specific cell-cell interactions and ligand-receptor pairs. Comparative analyses with colorectal cancer and inflammatory bowel disease demonstrate the power of SAHA to detect disease-associated spatial disruptions, including crypt dedifferentiation, perineural invasion, and therapy-resistant immune remodeling. All data are openly accessible through a FAIR-compliant interactive portal to support exploration, benchmarking, and machine learning model training. Through SAHA, we provide a foundational framework for spatial diagnostics and next-generation precision medicine grounded in a comprehensive human tissue atlas, enabling the development of context-aware models that simulate tissue behavior, decode complex pathologies, and accelerate therapeutic innovation at unprecedented scale.

The brain is complex and heterogeneous where cell function and cell-to-cell communication are critical for rapid and accurate performance. The ability to explore protein-driven activities at high resolution within spatial context of their immediate environment is critical to gain comprehensive pictures of brain development, activity, aging, disease or dysfunction, and inflammatory responses. Many existing approaches for high-plex single cell spatial proteomics face issues around simplicity, speed, scalability, and big data analysis. Here, we present an integrated workflow from sample through analysis that addresses key concerns around high plex proteomics. The CosMx ™ Spatial Molecular Imager and AtoMx ™ Spatial Informatics Platform comprises an end-to-end workflow that efficiently handles highly multiplex protein analysis at plex sizes exceeding 68 targets. The CosMx protein assays uses oligonucleotide conjugated antibodies, detected using universal, multi-analyte CosMx readout reagents. The CosMx Mouse Neural Cell Typing and Alzheimer’s Pathology panel is optimized to comprehensively profile neural cell lineages across the brain as well as the progression of Alzheimer’s disease (AD), including specific antibodies for humanized mouse AD models. The AtoMx spatial informatics platform provides full analysis support, including whole-slide image viewer, and methods for performing built-in or fully customizable analyses for cell typing, ligand-receptor analysis, neighborhood analysis and spatial differential expression. The CosMx protein assay reagents were validated on FFPE adult mouse brain, mouse embryo, and Alzheimer’s positive human brain. We used the CosMx Mouse Neural Cell Typing and Alzheimer’s Pathology panel with the CosMx Spatial Molecular Imager to identify multiple neuronal subtypes, different reactive states of astrocytes and microglial, cell degeneration and proliferation. Single cell exploration of mitochondria showed distinct patterning of key immune targets based on their immediate microenvironment. CosMx SMI is a high-plex spatial multi-omics platform that enables detection of > 68 proteins at subcellular resolution. In combination with the high-plex CosMx Mouse Neural Cell Typing and Alzheimer’s Pathology panel, we present a flexible and scalable informatics platform, a robust solution for comprehensive neural and disease phenotyping that captures the complexity of neuronal and glial cellular activity with full spatial context. FOR RESEARCH USE ONLY. Not for use in diagnostic procedures.

Detecting and analyzing large numbers of proteins using whole-slide imaging is critical for a comprehensive picture of immune response to cancer. Many existing approaches for high-plex proteomics face issues around simplicity, speed, scalability, and big data analysis. Here, we present an integrated workflow from sample preparation through downstream analysis that addresses many key concerns around high plex proteomics. The CosMx Spatial Molecular Imager (SMI) and AtoMx Spatial Informatics Platform (SIP) comprise of a turnkey, end-to-end workflow that efficiently handles highly multiplex protein analysis at plex sizes exceeding 110 targets. We demonstrate an extension of our commercially available 64-plex human immuno-oncology panel to higher numbers of targets and show how the cloud computing-enabled AtoMx SIP allows flexible construction of analytic pipelines for cell typing and spatial analyses. The CosMx protein assay uses antibodies conjugated with oligonucleotides, which are detected using universal, multi-analyte CosMx readout reagents. The CosMx Human Immuno-oncology panel was optimized to comprehensively profile lymphoid and stromal lineages within the tumor microenvironment as well as markers of cancer signaling and progression. Each CosMx SMI antibody was validated on multi-organ FFPE tissue microarrays covering prevalent solid tumor types with matched controls, and 52 human FFPE cell lines, including overexpression lines for key targets such as GITR, CD278, PD-L1, and PD-1. CosMx SMI uses a deep learning algorithm to segment whole cells and a semi-supervised algorithm to classify cell types. The AtoMx SIP provides full analysis support, including a whole-slide image viewer, and methods for performing built-in or fully customizable analyses for cell typing, ligand-receptor analysis, neighborhood analysis and spatial differential expression. Within the cancer sample profiled, we performed in-depth single-cell proteomic profiling across different cell populations. We detected TLS, characterized TLS maturation, and identified immune interactions with the tumor microenvironment. The CosMx SMI assay profiled the composition and spatial organization of infiltrating immune cells within and around the tumor microenvironment. We found that markers of T cell activation and exhaustion varied across the tumor landscape. CosMx SMI is a high-plex spatial multi-omics platform that enables detection of more than 110 proteins at subcellular resolution in real-world FFPE tissues. The extensibility of the CosMx protein assay to large numbers of protein targets and our flexible, scalable bioinformatic platform provides a straightforward and robust solution for comprehensive immune phenotyping with full spatial context. FOR RESEARCH USE ONLY. Not for use in diagnostic procedures. Citation Format: Tien Phan-Everson, Zachary Lewis, Giang Ong, Yan Liang, Emily Brown, Liuliu Pan, Aster Wardhani, Mithra Korukonda, Carl Brown, Dwayne Dunaway, Edward Zhao, Dan McGuire, Sangsoon Woo, Alyssa Rosenbloom, Brian Filanoski, Rhonda Meredith, Kan Chantranuvatana, Brian Birditt, Hye Son Yi, Erin Piazza, Jason Reeves, John Lyssand, Vik Devgan, Michael Rhodes, Gary Geiss, Joseph Beechem. A complete pipeline for high-plex spatial proteomic profiling and analysis on the cosmxtm spatial molecular imager and atomtm spatial informatics platform. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 4617.

Shanshan He1, Ruchir Bhatt1, Brian Birditt1, Carl Brown1, Emily Brown1, Kan Chantranuvatana1, Patrick Danaher1, Dwayne Dunaway1, Brian Filanoski1, Ryan G. Garrison1, Gary Geiss1, Mark T. Gregory1, Margaret L. Hoang1, Emily E. Killingbeck1, Tae Kyung Kim1, Youngmi Kim1, Mithra Korukonda1, Alecksandr Kutchma1, Erica Lee1, Zachary R. Lewis1, Yan Liang1, Jeffrey S. Nelson1, Giang Ong1, Evan Perillo1, Joseph Phan1, Tien Phan-Everson1, Erin Piazza1, Tushar Rane1, Zachary Reitz1, Michael Rhodes1, Alyssa Rosenbloom1, David Ross1, Hiromi Sato1, Aster W. Wardhani1, Corey Williams-Wietzikoski1, Lidan Wu1, Joseph M. Beechem1*

Both apoptotic and mitotic cells exhibit nuclear condensation which can be measured in a quantitative manner using a simple DNA dye and the ImageStream multispectral imaging flow cytometry. To distinguish apoptotic from mitotic cells additional morphological measurements such as textured brightfield imagery and increased texture and intensity of dark field scatter imagery may also be required. Distinguishing the sometimes minimal difference in level of nuclear condensation between apoptotic and mitotic cells in an automated manner can be problematic when applying only individual intensity and texture image analysis parameters. Here we present an automatic classifier which generates complex weighted combinations of intensity and texture features applied to multiple images of each cell to optimize distinction and quantitation of apoptotic and mitotic events in the same sample. Imagery from Ramos, Jurkat, and Colo205 cell lines were acquired on the ImageStream multispectral imaging flow cytometer after incubation with mechanistically distinct apoptosis inducing drugs (Camptothecin and Staurosporine) and staining with simple nuclear dyes. The accuracy of an automatic classifier which can measure the apoptotic index and mitotic index of samples based on brightfield, nuclear, and darkfield scatter imagery without the need for additional fluorescent labels using the IDEAS analytical software package is demonstrated.

Human emotional behavior, personality, and body language are the essential elements in the recognition of a believable synthetic story character. This paper presents an approach using story scripts and action descriptions in a form similar to the content description of storyboards to predict specific personality and emotional states. By adopting the Abridged Big Five Circumplex (AB5C) Model of personality from the study of psychology as a basis for a computational model, we construct a hierarchical fuzzy rule-based system to facilitate the personality and emotion control of the body language of a dynamic story character. The story character can consistently perform specific postures and gestures based on his/her personality type. Story designers can devise a story context in the form of our story interface which predictably motivates personality and emotion values to drive the appropriate movements of the story characters. Our system takes advantage of relevant knowledge described by psychologists and researchers of storytelling, nonverbal communication, and human movement. Our ultimate goal is to facilitate the high-level control of a synthetic character

We present an approach using story scripts and action descriptions in a form similar to the content description of storyboards to predict specific personality and emotional states. By constructing a hierarchical fuzzy rule-based system we facilitate the personality and emotion control of the body language of dynamic story characters. Our ultimate goal is to facilitate the high-level control of synthetic characters.

Robots have diverse capabilities and complex interactions with their environment. Software development for robotic platforms is time consuming due to the complex nature of the tasks to be performed. Such an environment demands sound software engineering practices to produce high quality software. However software engineering in the robotics domain fails to facilitate any significant level of software reuse or portability. This paper identifies the major issues limiting software reuse in the robotics domain. Lack of standardisation, diversity of robotic platforms, and the subtle effects of environmental interaction all contribute to this problem. It is then shown that software components, fuzzy logic, and related techniques can be used together to provide suitable abstractions to address this problem. While complete software reuse is not possible, it is demonstrated that significant levels of software reuse can be obtained. Without an acceptable level of reuse or portability, software engineering in the robotics domain will not be able to meet the demands of a rapidly developing field. The work presented in this paper demonstrates a method for supporting software reuse across robotic platforms and hence facilitating improved software engineering practices.

Uniform sensor management and abstraction across different robot platforms is a difficult task due to the sheer diversity of sensing devices. However, because these sensors can be grouped into categories that in essence provide the same information, we can capture their similarities and create abstractions. An example would be distance data measured by an assortment of range sensors, or alternatively extracted from a camera using image processing. This paper describes how using software components it is possible to uniformly construct high-level abstractions of sensor information across various robots in a way to support the portability of common code that uses these abstractions (e.g. obstacle avoidance, wall following). We demonstrate our abstractions on a number of robots using different configurations of range sensors and cameras.

Currently, most content based image retrieval (CBIR) systems operate on all images, without sorting images into different types or categories. Different images have different characteristics, and thus often require different analysis techniques and query types. Additionally, placing an image into a category can help the user to navigate retrieval results more effectively. To categorise an image, firstly the dominant region needs to be extracted using multi level colour segmentation. Based on the region's features of colour, texture, shape and relation between regions, the image is then categorised. Users are presented with retrieval results sorted into different categories, where dominant region extraction allows object based retrieval to be performed.

User input interfaces are quickly become more natural and intuative, relying less and less on the traditional mouse and keyboard interface, and moving towards a pen based input system. Current technologies which take advancement of these developments in hardware have been based primarily on handwriting recognition to allow the user to write there instructions instead of typing. Some work has been performed regarding diagram recognition from on-line input, however these system have been developed using hardcoded parameters with fuzzy sets and common feature sets. Hidden Markov Models are a powerful mathematical recognition tool, which is current being used in feilds such as speech recognition, handwriting recognition and DNA cell searching and classification applications. This is the first attempt at using a hidden markov model to recognize input from a two dimensonal spacial environment, the result of the initial implementation have shown promising results for more complex recognition models.

The aim of this research is to propose a new content based image retrieval (CBIR) system using categories. Different images have different characteristics and thus often require different image processing techniques. Most current CBIR systems operate on all images, without pre-sorting images into different categories. This results in limitations on retrieval performance and accuracy. Two semantic and four syntactic image categories are proposed. The category for an image is generated automatically by analysing the image for the presence of a dominant object or for correspondence to an image ‘template’. Dominant objects are obtained by performing region grouping of segmented thumbnails. The result of this research is a new Internet image retrieval and indexing system.

This paper studies the problem of creating artificial fish for real-time interactive virtual worlds aimed at desktop environments with hardware 3D support. The artificial fish developed have the ability to move, sense, and think. Each fish is modelled with a Keyframed skeletal animated body, semi-physics based movement model, sensory abilities, internal motivations, a set of behaviour routines, and a behavioural selection mechanism. These features allow the fish to act autonomously using behavioural rules in response to sensory input from the environment and other fish. This autonomous ability enables definite behaviours to be described and observed in the fish that are not simply random, cyclic, or scripted.Excellent work has been previously done on modelling sophisticated artificial fish. The contribution of this paper is to focus on the practical modelling of fish for game production.

This paper describes the work-in-progress of creating an artificial 3D environment and robot, suitable for educational simulation. A 3D vehicle robot, equipped with a monocular camera navigates in a physics based 3D environment, with some artificial intelligence capabilities. Students can interact with the robot, add new objects and set the robot for some tasks. This multimedia tool is designed for student with very little experience with robotics.

Mobile robots today, while varying greatly in design, of- ten have a large number of similarities in terms of their tasks and goals. Navigation, obstacle avoidance, and vision are all examples. In turn, robots of similar design, but with varying configurations, should be able to share the bulk of their controlling software. Any changes required should be minimal and ideally only to specify new hardware config- urations. However, it is difficult to achieve such flexibil- ity, mainly due to the enormous variety of robot hardware available and the huge number of possible configurations. Monolithic controllers that can handle such variety are im- possible to build. This paper will investigate these portability problems, as well as techniques to manage common abstractions for user-designed components. The challenge is in creating new methods for robot software to support a diverse va- riety of robots, while also being easily upgraded and ex- tended. These methods can then provide new ways to sup- port the operational and functional reuse of the same high- level components across a variety of robots.

This paper describes the work-in-progress of creating an artificial 3D environment and robot, suitable for educational simulation. A visual 3D vehicle robot, equipped with a monocular camera navigates in a physics based 3D environment, with some artificial intelligence capabilities. Students can interact with the robot, add new objects and set the robot various tasks. This multimedia tool is designed for students with very little experience with robotics, and aims at giving students unlimited access to a relatively sophisticated robotic system, incorporating artificial intelligence, with an extremely low cost compared to using real robot systems. Our current version of the simulation software has been designed to perform three main tasks, play soccer, avoid object, and wander. The simulation is being designed to closely resemble its real-world counterparts, and we hope, will ultimately become a powerful research and development tool.

Mobile robots today, while varying greatly in design, often have a large number of similarities in terms of their tasks and goals. Navigation, obstacle avoidance, and vision are all examples. In turn, robots of similar design, but with varying configurations, should be able to share the bulk of their controlling software. Any changes required should be minimal and ideally only to specify new hardware configurations. However, it is difficult to achieve such flexibility, mainly due to the enormous variety of robot hardware available and the huge number of possible configurations. Monolithic controllers that can handle such variety are impossible to build. This paper will investigate these portability problems, as well as techniques to manage common abstractions for user-designed components. The challenge is in creating new methods for robot software to support a diverse variety of robots, while also being easily upgraded and extended. These methods can then provide new ways to support the operational and functional reuse of the same high-level components across a variety of robots.

Searching for an object in a general image collection us- ing current image retrieval systems, is still a problem. The retrieval results contain many unrelated images. In pro- viding an effective and robust image database, objects in an image need to be extracted. Since the number of stored images can be very large, automation is an important as- pect. Image indexing is a technique that extracts objects in an image automatically. The aim of this research is to propose a new object based indexing system based on ex- tracting salient region representative from the image and categorising an image into different types. Different image has different characteristics and often require different image processing techniques. Currently, most content based image retrieval (CBIR) systems oper- ate on all images, without pre-sorting these images into dif- ferent types. This resulted in limitations on retrieval per- formance and accuracy. Categories described here are of statistical and syntactical descriptions rather than semanti- cal. By analysing which features are dominant in an image, two outcomes will be obtained: category for that image and salient object. Identifying salient object further reduce the retrieval results into relevant images.