Bioimage Informatics for Systems Pharmacology Authors : Fuhai Li - - PowerPoint PPT Presentation
Bioimage Informatics for Systems Pharmacology Authors : Fuhai Li Zheng Yin Guangxu Jin Hong Zhao Stephen T. C. Wong Presented by : Iffat chowdhury Motivation Image is worth for phenotypic changes identification High resolution
Presented by : Iffat chowdhury
Image is worth for phenotypic changes identification High resolution microscopy, fluorescent labeling Rich in terms of information of biological processes Bioimage informatics
Multicolor cell imaging-based studies Live-cell imaging-based studies Neuron imaging-based studies C. elegans imaging-based studies
Multicolor cell imaging-based studies Live-cell imaging-based studies Neuron imaging-based studies C. elegans imaging-based studies
Multiple fluorescent markers Feature extraction Drosophila cell Softwares : CellProfiler, Fiji, Icy, GcellIQ,
Multiple fluorescent markers Feature extraction Drosophila cell Softwares : CellProfiler, Fiji, Icy, GcellIQ,
Multicolor cell imaging-based studies Live-cell imaging-based studies Neuron imaging-based studies C. elegans imaging-based studies
Progression, proliferation, migration of cell Dynamic behaviors of cells Live Hela cell images Softwares : CellProfiler, Fiji, BioimageXD, Icy,
Progression, proliferation, migration of cell Dynamic behaviors of cells Live Hela cell images Softwares : CellProfiler, Fiji, BioimageXD, Icy,
Multicolor cell imaging-based studies Live-cell imaging-based studies Neuron imaging-based studies C. elegans imaging-based studies
To study brain functions and disorders Use super-resolution microscope Softwares : NeurphologyJ, NeuronJ,
To study brain functions and disorders Use super-resolution microscope Softwares : NeurphologyJ, NeuronJ,
Multicolor cell imaging-based studies Live-cell imaging-based studies Neuron imaging-based studies C. elegans imaging-based studies
Common animal model for drug and target
Consists of only hundred of cells Embryonic development
Source : Wikipedia
Common animal model for drug and target
Consists of only hundred of cells Embryonic development
Detect the locations of individual objects Facilitate the segmentation by giving the position
Two types of object detection :
Nuclei detection Distance transformation Seeded watershed Intensity information Gradient vector
Nuclei detection Distance transformation Seeded watershed Intensity information Gradient vector
Intensity remains constant Centerline detection Edge detectors Machine-learning
Intensity remains constant Centerline detection Edge detectors Machine-learning
Delineate boundaries of objects Threshold segmentation Fuzzy-C-Means method Watershed algorithm Active contour model Level set representation Voronoi segmentation Graph cut method Softwares : CellProfiler, Fiji, Ilastik, SLIC
Delineate boundaries of objects Threshold segmentation Fuzzy-C-Means method Watershed algorithm Active contour model Level set representation Voronoi segmentation Graph cut method Softwares : CellProfiler, Fiji, Ilastik, SLIC
Figure taken from http://www.dma.fi.upm.es/mabellanas/tfcs/fvd/voronoi.html
Delineate boundaries of objects Threshold segmentation Fuzzy-C-Means method Watershed algorithm Active contour model Level set representation Voronoi segmentation Graph cut method Softwares : CellProfiler, Fiji, Ilastik, SLIC
Study dynamic behaviors Three approaches :
Study dynamic behaviors Three approaches :
Study dynamic behaviors Three approaches :
Cell / nuclei are detected first Boundary comes next Contour model Different objects get different colors
2D image sequences as 3D Level set segmentation approaches
First detected and then segmented Tracking is dependent of segmentation and
Association Filters may be used
Fiji, Icy, BioimageXD are for higher dimensional
NeuronStudio for neuron image analysis Farsight and vaa3D for microscopy images For customize tools, Visualization Toolkit helps.
Quantitative measuring
Wavelet feature : characterize the images in both –
Cell cycle phase identification User defined phenotype, identification and
Automated cell cycle phase identification is needed
SVM, K-nearest neighbors, Bayesian classifiers Can be done during segmentation and tracking.
Automated cell cycle phase identification is needed
SVM, K-nearest neighbors, Bayesian classifiers Can be done during segmentation and tracking.
Exhibit novel phenotype and unpredicted behaviors. Gaussian Mixture Model with statistics Clustering analysis Classifiers
Exhibit novel phenotype and unpredicted behaviors. Gaussian Mixture Model with statistics Clustering analysis Classifiers
Clustering analysis SVM-based multivariate profiling analysis Factor-based multidimensional profiling analysis Subpopulation-based heterogeneity profiling
Clustering analysis SVM-based multivariate profiling analysis Factor-based multidimensional profiling analysis Subpopulation-based heterogeneity profiling
Experimental perturbations Softwares : Cluster 3.0, Java TreeView
Clustering analysis SVM-based multivariate profiling analysis Factor-based multidimensional profiling analysis Subpopulation-based heterogeneity profiling
Wells with treated cells compared to wells with
The differences are indicated by the outputs of
One is the accuracy and another is the normal
Clustering analysis SVM-based multivariate profiling analysis Factor-based multidimensional profiling
Subpopulation-based heterogeneity profiling
Correlation of the features within the group and
Redundancy can be removed by factor analysis. Six factors representing nuclei size, DNA
Clustering analysis SVM-based multivariate profiling analysis Factor-based multidimensional profiling analysis Subpopulation-based heterogeneity profiling
Heterogeneous behavior within a cell population. GMM model to divide into subpopulation.