Graphical-model framework for automated annotation of cell identities in dense cellular images

eLife ◽

10.7554/elife.60321 ◽

2021 ◽

Vol 10 ◽

Author(s):

Shivesh Chaudhary ◽

Sol Ah Lee ◽

Yueyi Li ◽

Dhaval S Patel ◽

Hang Lu

Keyword(s):

Graphical Model ◽

Conditional Random Fields ◽

Ground Truth ◽

Computationally Efficient ◽

Model Framework ◽

Efficient Manner ◽

Whole Brain ◽

C Elegans ◽

Automated Annotation ◽

Brain Data

Although identifying cell names in dense image stacks is critical in analyzing functional whole-brain data enabling comparison across experiments, unbiased identification is very difficult, and relies heavily on researchers' experiences. Here we present a probabilistic-graphical-model framework, CRF_ID, based on Conditional Random Fields, for unbiased and automated cell identification. CRF_ID focuses on maximizing intrinsic similarity between shapes. Compared to existing methods, CRF_ID achieves higher accuracy on simulated and ground-truth experimental datasets, and better robustness against challenging noise conditions common in experimental data. CRF_ID can further boost accuracy by building atlases from annotated data in highly computationally efficient manner, and by easily adding new features (e.g. from new strains). We demonstrate cell annotation in C. elegans images across strains, animal orientations, and tasks including gene-expression localization, multi-cellular and whole-brain functional imaging experiments. Together, these successes demonstrate that unbiased cell annotation can facilitate biological discovery, and this approach may be valuable to annotation tasks for other systems.

Automated Annotation of Cell Identities in Dense Cellular Images

10.1101/2020.03.10.986356 ◽

2020 ◽

Cited By ~ 1

Author(s):

Shivesh Chaudhary ◽

Sol Ah Lee ◽

Yueyi Li ◽

Dhaval S. Patel ◽

Hang Lu

Keyword(s):

Random Fields ◽

Conditional Random Fields ◽

Gene Expression Pattern ◽

Ground Truth ◽

Ground Truth Data ◽

Automated Annotation ◽

Expression Pattern Analysis ◽

Dense Image ◽

Active Data ◽

Imaging Conditions

AbstractAssigning cell identities in dense image stacks is critical for many applications, for comparing data across animals and experiment conditions, and investigating properties of specific cells. Conventional methods are laborious, require experience, and could introduce bias. We present a generalizable framework based on Conditional Random Fields models for automatic cell identification. This approach searches for optimal arrangements of labels that maximally preserves prior knowledge such as geometrical relationships. The algorithm shows better accuracy and more robust handling of perturbations, e.g. missing cells and position variability, with both synthetic and experimental ground-truth data. The framework is generalizable across strains, imaging conditions, and easily builds and utilizes active data-driven atlases, which further improves accuracy. We demonstrate the utility in gene-expression pattern analysis, multi-cellular calcium imaging, and whole-brain imaging experiments. Thus, our framework is highly valuable to a wide variety of annotation scenarios including in zebrafish, Drosophila, hydra, and mouse brains.

An ensemble learning approach to auto-annotation for whole-brain C. elegans imaging

10.1101/180430 ◽

2017 ◽

Author(s):

S. Wu ◽

Y. Toyoshima ◽

M.S. Jang ◽

M. Kanamori ◽

T. Teramoto ◽

...

Keyword(s):

Ensemble Learning ◽

Scientific Discovery ◽

Large Data ◽

Individual Neuron ◽

Learning Approach ◽

Data Sets ◽

Whole Brain ◽

C Elegans ◽

Automated Annotation ◽

New Research

AbstractShifting from individual neuron analysis to whole-brain neural network analysis opens up new research opportunities for Caenorhabditis elegans (C. elegans). An automated data processing pipeline, including neuron detection, segmentation, tracking and annotation, will significantly improve the efficiency of analyzing whole-brain C. elegans imaging. The resulting large data sets may motivate new scientific discovery by exploiting many promising analysis tools for big data. In this study, we focus on the development of an automated annotation procedure. With only around 180 neurons in the central nervous system of a C. elegans, the annotation of each individual neuron still remains a major challenge because of the high density in space, similarity in neuron shape, unpredictable distortion of the worm’s head during motion, intrinsic variations during worm development, etc. We use an ensemble learning approach to achieve around 25% error for a test based on real experimental data. Also, we demonstrate the importance of exploring extra source of information for annotation other than the neuron positions.

A Cluster Graph Approach to Land Cover Classification Boosting

Data ◽

10.3390/data4010010 ◽

2019 ◽

Vol 4 (1) ◽

pp. 10

Author(s):

Lloyd Hughes ◽

Simon Streicher ◽

Ekaterina Chuprikova ◽

Johan Du Preez

Keyword(s):

Land Cover ◽

Graphical Model ◽

Land Cover Classification ◽

Probabilistic Graphical Model ◽

Observation Data ◽

Computationally Efficient ◽

Efficient Manner ◽

Uncertainty Estimates ◽

Spatio Temporal ◽

Cluster Graph

When it comes to land cover classification, the process of deriving the land classes is complex due to possible errors in algorithms, spatio-temporal heterogeneity of the Earth observation data, variation in availability and quality of reference data, or a combination of these. This article proposes a probabilistic graphical model approach, in the form of a cluster graph, to boost geospatial classifications and produce a more accurate and robust classification and uncertainty product. Cluster graphs can be characterized as a means of reasoning about geospatial data such as land cover classifications by considering the effects of spatial distribution, and inter-class dependencies in a computationally efficient manner. To assess the capabilities of our proposed cluster graph boosting approach, we apply it to the field of land cover classification. We make use of existing land cover products (GlobeLand30, CORINE Land Cover) along with data from Volunteered Geographic Information (VGI), namely OpenStreetMap (OSM), to generate a boosted land cover classification and the respective uncertainty estimates. Our approach combines qualitative and quantitative components through the application of our probabilistic graphical model and subjective expert judgments. Evaluating our approach on a test region in Garmisch-Partenkirchen, Germany, our approach was able to boost the overall land cover classification accuracy by 1.4% when compared to an independent reference land cover dataset. Our approach was shown to be robust and was able to produce a diverse, feasible and spatially consistent land cover classification in areas of incomplete and conflicting evidence. On an independent validation scene, we demonstrated that our cluster graph boosting approach was generalizable even when initialized with poor prior assumptions.

Thresholded Partial Least Squares: Fast Construction of Interpretable Whole-brain Decoders

10.1101/2021.02.09.430524 ◽

2021 ◽

Author(s):

Sangil Lee ◽

Eric T. Bradlow ◽

Joseph W. Kable

Keyword(s):

Least Squares ◽

Partial Least Squares ◽

Brain Activity ◽

Mental States ◽

Computationally Efficient ◽

Activity Data ◽

Data Set ◽

Whole Brain ◽

Novel Approach ◽

Brain Data

AbstractRecent neuroimaging research has shown that it is possible to decode mental states and predict future consumer behavior from brain activity data (a time-series of images). However, the unique characteristics (and high dimensionality) of neuroimaging data, coupled with a need for neuroscientifically interpretable models, has largely discouraged the use of the entire brain’s data as predictors. Instead, most neuroscientific research uses “regionalized” (partial-brain) data to reduce the computational burden and to improve interpretability (i.e., localizability of signal), at the cost of losing potential information. Here we propose a novel approach that can build whole-brain neural decoders (using the entire data set and capitalizing on the full correlational structure) that are both interpretable and computationally efficient. We exploit analytical properties of the partial least squares algorithm to build a regularized regression model with variable selection that boasts (in contrast to most statistical methods) a unique ‘fit-once-tune-later’ approach where users need to fit the model only once and can choose the best tuning parameters post-hoc. We demonstrate its efficacy in a large neuroimaging dataset against off-the-shelf prediction methods and show that our new method scales exceptionally with increasing data size, yields more interpretable results, and uses less computational memory, while retaining high predictive power.

Classification of unlabeled online media

Scientific Reports ◽

10.1038/s41598-021-85608-5 ◽

2021 ◽

Vol 11 (1) ◽

Author(s):

Sakthi Kumar Arul Prakash ◽

Conrad Tucker

Keyword(s):

Social Media ◽

Real World ◽

Graphical Model ◽

Ground Truth ◽

Classification Problem ◽

Machine Learning Algorithms ◽

Social Media Networks ◽

Online Social Media ◽

Wide Range

AbstractThis work investigates the ability to classify misinformation in online social media networks in a manner that avoids the need for ground truth labels. Rather than approach the classification problem as a task for humans or machine learning algorithms, this work leverages user–user and user–media (i.e.,media likes) interactions to infer the type of information (fake vs. authentic) being spread, without needing to know the actual details of the information itself. To study the inception and evolution of user–user and user–media interactions over time, we create an experimental platform that mimics the functionality of real-world social media networks. We develop a graphical model that considers the evolution of this network topology to model the uncertainty (entropy) propagation when fake and authentic media disseminates across the network. The creation of a real-world social media network enables a wide range of hypotheses to be tested pertaining to users, their interactions with other users, and with media content. The discovery that the entropy of user–user and user–media interactions approximate fake and authentic media likes, enables us to classify fake media in an unsupervised learning manner.

Computationally Efficient Magnetic Resonance Imaging Based Surface Contact Modeling as a Tool to Evaluate Joint Injuries and Outcomes of Surgical Interventions Compared to Finite Element Modeling

Journal of Biomechanical Engineering ◽

10.1115/1.4026485 ◽

2014 ◽

Vol 136 (4) ◽

Cited By ~ 4

Author(s):

Joshua E. Johnson ◽

Phil Lee ◽

Terence E. McIff ◽

E. Bruce Toby ◽

Kenneth J. Fischer

Keyword(s):

Finite Element ◽

Finite Element Modeling ◽

Joint Mechanics ◽

Computationally Efficient ◽

Efficient Manner ◽

Element Modeling ◽

Joint Injuries ◽

Surface Contact ◽

Joint Contact ◽

Contact Pressures

Joint injuries and the resulting posttraumatic osteoarthritis (OA) are a significant problem. There is still a need for tools to evaluate joint injuries, their effect on joint mechanics, and the relationship between altered mechanics and OA. Better understanding of injuries and their relationship to OA may aid in the development or refinement of treatment methods. This may be partially achieved by monitoring changes in joint mechanics that are a direct consequence of injury. Techniques such as image-based finite element modeling can provide in vivo joint mechanics data but can also be laborious and computationally expensive. Alternate modeling techniques that can provide similar results in a computationally efficient manner are an attractive prospect. It is likely possible to estimate risk of OA due to injury from surface contact mechanics data alone. The objective of this study was to compare joint contact mechanics from image-based surface contact modeling (SCM) and finite element modeling (FEM) in normal, injured (scapholunate ligament tear), and surgically repaired radiocarpal joints. Since FEM is accepted as the gold standard to evaluate joint contact stresses, our assumption was that results obtained using this method would accurately represent the true value. Magnetic resonance images (MRI) of the normal, injured, and postoperative wrists of three subjects were acquired when relaxed and during functional grasp. Surface and volumetric models of the radiolunate and radioscaphoid articulations were constructed from the relaxed images for SCM and FEM analyses, respectively. Kinematic boundary conditions were acquired from image registration between the relaxed and grasp images. For the SCM technique, a linear contact relationship was used to estimate contact outcomes based on interactions of the rigid articular surfaces in contact. For FEM, a pressure-overclosure relationship was used to estimate outcomes based on deformable body contact interactions. The SCM technique was able to evaluate variations in contact outcomes arising from scapholunate ligament injury and also the effects of surgical repair, with similar accuracy to the FEM gold standard. At least 80% of contact forces, peak contact pressures, mean contact pressures and contact areas from SCM were within 10 N, 0.5 MPa, 0.2 MPa, and 15 mm2, respectively, of the results from FEM, regardless of the state of the wrist. Depending on the application, the MRI-based SCM technique has the potential to provide clinically relevant subject-specific results in a computationally efficient manner compared to FEM.

Neuron ID dataset facilitates neuronal annotation for whole-brain activity imaging of C. elegans

BMC Biology ◽

10.1186/s12915-020-0745-2 ◽

2020 ◽

Vol 18 (1) ◽

Cited By ~ 5

Author(s):

Yu Toyoshima ◽

Stephen Wu ◽

Manami Kanamori ◽

Hirofumi Sato ◽

Moon Sun Jang ◽

...

Keyword(s):

Brain Activity ◽

Whole Brain ◽

C Elegans

Functional connectomics from neural dynamics: probabilistic graphical models for neuronal network of Caenorhabditis elegans

Philosophical Transactions of the Royal Society B Biological Sciences ◽

10.1098/rstb.2017.0377 ◽

2018 ◽

Vol 373 (1758) ◽

pp. 20170377 ◽

Cited By ~ 6

Author(s):

Hexuan Liu ◽

Jimin Kim ◽

Eli Shlizerman

Keyword(s):

Time Series ◽

Nervous System ◽

Caenorhabditis Elegans ◽

Graphical Models ◽

Neuronal Network ◽

Graphical Model ◽

Time Series Data ◽

Series Data ◽

Neuronal Responses ◽

C Elegans

We propose an approach to represent neuronal network dynamics as a probabilistic graphical model (PGM). To construct the PGM, we collect time series of neuronal responses produced by the neuronal network and use singular value decomposition to obtain a low-dimensional projection of the time-series data. We then extract dominant patterns from the projections to get pairwise dependency information and create a graphical model for the full network. The outcome model is a functional connectome that captures how stimuli propagate through the network and thus represents causal dependencies between neurons and stimuli. We apply our methodology to a model of the Caenorhabditis elegans somatic nervous system to validate and show an example of our approach. The structure and dynamics of the C. elegans nervous system are well studied and a model that generates neuronal responses is available. The resulting PGM enables us to obtain and verify underlying neuronal pathways for known behavioural scenarios and detect possible pathways for novel scenarios. This article is part of a discussion meeting issue ‘Connectome to behaviour: modelling C. elegans at cellular resolution’.

Towards Gas Discrimination and Mapping in Emergency Response Scenarios Using a Mobile Robot with an Electronic Nose

Sensors ◽

10.3390/s19030685 ◽

2019 ◽

Vol 19 (3) ◽

pp. 685 ◽

Cited By ~ 10

Author(s):

Han Fan ◽

Victor Hernandez Bennetts ◽

Erik Schaffernicht ◽

Achim Lilienthal

Keyword(s):

Emergency Response ◽

Electronic Nose ◽

Gas Sensing ◽

Practical Significance ◽

Computationally Efficient ◽

Urban Search And Rescue ◽

Efficient Manner ◽

Distribution Maps ◽

Gas Identification ◽

Sampling Procedures

Emergency personnel, such as firefighters, bomb technicians, and urban search and rescue specialists, can be exposed to a variety of extreme hazards during the response to natural and human-made disasters. In many of these scenarios, a risk factor is the presence of hazardous airborne chemicals. The recent and rapid advances in robotics and sensor technologies allow emergency responders to deal with such hazards from relatively safe distances. Mobile robots with gas-sensing capabilities allow to convey useful information such as the possible source positions of different chemicals in the emergency area. However, common gas sampling procedures for laboratory use are not applicable due to the complexity of the environment and the need for fast deployment and analysis. In addition, conventional gas identification approaches, based on supervised learning, cannot handle situations when the number and identities of the present chemicals are unknown. For the purpose of emergency response, all the information concluded from the gas detection events during the robot exploration should be delivered in real time. To address these challenges, we developed an online gas-sensing system using an electronic nose. Our system can automatically perform unsupervised learning and update the discrimination model as the robot is exploring a given environment. The online gas discrimination results are further integrated with geometrical information to derive a multi-compound gas spatial distribution map. The proposed system is deployed on a robot built to operate in harsh environments for supporting fire brigades, and is validated in several different real-world experiments of discriminating and mapping multiple chemical compounds in an indoor open environment. Our results show that the proposed system achieves high accuracy in gas discrimination in an online, unsupervised, and computationally efficient manner. The subsequently created gas distribution maps accurately indicate the presence of different chemicals in the environment, which is of practical significance for emergency response.

MONACO: accurate biological network alignment through optimal neighborhood matching between focal nodes

Bioinformatics ◽

10.1093/bioinformatics/btaa962 ◽

2020 ◽

Author(s):

Hyun-Myung Woo ◽

Byung-Jun Yoon

Keyword(s):

Biological Network ◽

Protein Complexes ◽

Ground Truth ◽

Network Alignment ◽

Supplementary Information ◽

Alignment Algorithm ◽

Computationally Efficient ◽

Topological Similarity ◽

Alignment Algorithms ◽

Multiple Network

Abstract Motivation Alignment of protein–protein interaction networks can be used for the unsupervised prediction of functional modules, such as protein complexes and signaling pathways, that are conserved across different species. To date, various algorithms have been proposed for biological network alignment, many of which attempt to incorporate topological similarity between the networks into the alignment process with the goal of constructing accurate and biologically meaningful alignments. Especially, random walk models have been shown to be effective for quantifying the global topological relatedness between nodes that belong to different networks by diffusing node-level similarity along the interaction edges. However, these schemes are not ideal for capturing the local topological similarity between nodes. Results In this article, we propose MONACO, a novel and versatile network alignment algorithm that finds highly accurate pairwise and multiple network alignments through the iterative optimal matching of ‘local’ neighborhoods around focal nodes. Extensive performance assessment based on real networks as well as synthetic networks, for which the ground truth is known, demonstrates that MONACO clearly and consistently outperforms all other state-of-the-art network alignment algorithms that we have tested, in terms of accuracy, coherence and topological quality of the aligned network regions. Furthermore, despite the sharply enhanced alignment accuracy, MONACO remains computationally efficient and it scales well with increasing size and number of networks. Availability and implementation Matlab implementation is freely available at https://github.com/bjyoontamu/MONACO. Supplementary information Supplementary data are available at Bioinformatics online.