dynamic time warping for automated cell cycle labelling
DESCRIPTION
Dynamic Time Warping for Automated Cell Cycle Labelling. A. El-Labban, A. Zisserman University of Oxford. Y. Toyoda, A. Bird, A. Hyman Max Planck Institute of Molecular Cell Biology and Genetics. Objectives. Segment and track mitotic cells Label mitotic phases Fully automated system. - PowerPoint PPT PresentationTRANSCRIPT
Dynamic Time Warping for Automated Cell Cycle Labelling
Dynamic Time Warping for Automated Cell Cycle LabellingA. El-Labban, A. ZissermanUniversity of OxfordY. Toyoda, A. Bird, A. HymanMax Planck Institute of Molecular Cell Biology and GeneticsObjectivesSegment and track mitotic cells
Label mitotic phases
Fully automated system
InterphaseProphasePrometaphaseMetaphaseAnaphaseTelophaseData3D time lapse image stacks
Use max intensity z-projections
1-5 minute temporal resolution
0.2 micron xy-resolution
3ApproachExisting approaches (e.g. Harder et al. 2009, Held et al. 2010 [CellCognition]):Track cellsLabel cell cycle phase frame-by-frameSmooth result with HMM (CellCognition)
Our Approach:Track cellsLabel all frames by using temporal signals of featuresTemporal signals of features
Temporal signals of featuresInterphaseProphasePrometaphaseMetaphaseAnaphaseTelophase
OverviewPart ITrack cells in videos
Part IILabel mitotic phases
Part I TrackingTracking Tracking by detectionDetect first, then associate objectsHere we use detection by classification.
Segmentation: Our approachLogistic regression classifier
Graph Cuts
Input imageProbability mapBinary mapLogistic regression
classifierGraph CutLogistic Regression ClassifierFeature:10 bin intensity histogram in 5x5 window around pixel
Non-uniform bins
Get local neighbourhood information as opposed to single pixel
Histogram gives rotational invariance
Logistic RegressionGives a probability map:
Graph Cuts
Probability from Logistic Regression ClassifierGradient dependent pairwise termUses local neighbourhood information to make decisionsPairwise term penalises different labels for adjacent pixelsGraph Cuts
TrackingAssociate objects based on distance between centroids in consecutive frames.Easy given segmentation and slow movement of cells.
Tracking
Associate objects based on distance between centroids in consecutive frames.Easy given segmentation and slow movement of cells.
TrackingAssociate objects based on distance between centroids in consecutive frames.Easy given segmentation and slow movement of cells.
Tracking
Part II Phase Labelling
Simple featuresMaximum Intensity:
Interphase
Simple featuresMaximum Intensity:
InterphaseProphase21
Simple featuresMaximum Intensity:
InterphaseProphasePrometaphase
Simple featuresMaximum Intensity:
InterphaseProphasePrometaphaseMetaphase
Simple featuresMaximum Intensity:
InterphaseProphasePrometaphaseMetaphaseAnaphase
Simple featuresMaximum Intensity:
InterphaseProphasePrometaphaseMetaphaseAnaphase
Simple featuresMaximum Intensity:
InterphaseProphasePrometaphaseMetaphaseAnaphaseTelophaseReference signalAverage over training set (1 standard deviation shaded):
Dynamic time warpingStretch signal onto labelled reference:
Dynamic time warpingStretch signal onto labelled reference:
Dynamic time warping
InterphaseProphasePrometaphaseMetaphaseAnaphaseTelophaseInterphaseDynamic time warpingFind a cost matrix of pairwise distances between points on the two signalsFind minimum cost path through matrix
Test SignalReference SignalFeatures
Hidden Markov ModelHidden states, xMitotic phases
Observations, yFeatures
Transition probabilities, aFrom one phase to the next
Emission probabilities, bOf features having a given value in a given phaseImage: http://en.wikipedia.org/wiki/Hidden_Markov_model33Hidden Markov ModelDTW essentially a special case of HMMEasy to extend approachCan add other classes e.g. cell deathSplit phases into sub-phases to account for variation
34Experiments and Data54 movies
119 mitotic tracks
27 movies (61 tracks) training, 27 movies (58 tracks) testingResultsInterphaseProphasePrometaphaseMetaphaseAnaphaseTelophase
Results
Outputs
OutputsSynopsis video1 of mitotic cells
Aligned to start of anaphase1Rav-Acha et al., 2006
ConclusionsNovel approach to cell cycle phase labelling
Utilises temporal context
Extendable with HMM Questions?