Tracking Colliding Cells
0 likes512 views
This document describes a new method for tracking cells that collide in microscope videos. It presents 3 key challenges: 1) cells' motion and appearance can change abruptly during collisions, 2) previous methods used single hypotheses that cannot handle collisions well, 3) manual tracking is tedious and expensive. The proposed method addresses these by using multiple hypotheses about collision states, Kalman filtering for reliability, and detecting/tracking cells in videos in a step-by-step process. Experiments show it achieves an 88% coverage of colliding cell positions, outperforming other methods.
![S. J. Schmugge, S. Keller, N. Nguyen, R. Souvenir, T. H. Huynh, M. Clemens, M. C. Shin. "Segmentation of Vessels Cluttered with Cells using a Physics based Model". 11th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), New York, September 6-10 2008.N. Nguyen, S. Keller, T. Huynh, M. Shin. ¡°Tracking Colliding Cells¡±. IEEE Workshop on Applications of Computer Vision (WACV), Snowbird, UT December 07-09 2009.[to be submitted to IEEE Transactions on Medical Imaging]N. Nguyen, S. Keller, Eric Norris, T. Huynh, M. Shin. ¡°Tracking Colliding Cells in In-Vivo Intravital Microscopy Images¡±. Publications](https://image.slidesharecdn.com/wbcgraduateresearchfair2010-100507133522-phpapp02/85/Tracking-Colliding-Cells-56-320.jpg){kind=tracking}
![Previous Automatic MethodsRay et al. [2002]Active ContourCui et al. [2005]Monte CarloMukherjee et al. [2004]Level Set Analysis](https://image.slidesharecdn.com/wbcgraduateresearchfair2010-100507133522-phpapp02/85/Tracking-Colliding-Cells-61-320.jpg){kind=tracking}
![Previous Automatic MethodsEden et al. [2005]SmoothnessConstraintsLi et al. [2005]Lineage ConstructionSmith et al. [2008]Probabilistic Formalization](https://image.slidesharecdn.com/wbcgraduateresearchfair2010-100507133522-phpapp02/85/Tracking-Colliding-Cells-62-320.jpg){kind=tracking}
![More ResultsEden et al. [2005]Our Method](https://image.slidesharecdn.com/wbcgraduateresearchfair2010-100507133522-phpapp02/85/Tracking-Colliding-Cells-82-320.jpg){kind=tracking}
More Related Content
Similar to Tracking Colliding Cells (20)
More from Rich Nguyen (6)
Recently uploaded (20)
![[Webinar] Scaling Made Simple: Getting Started with No-Code Web Apps](https://cdn.slidesharecdn.com/ss_thumbnails/webinarscalingmadesimplegettingstartedwithno-codewebapps-mar52025-250305183437-f03c78a3-thumbnail.jpg?width=560&fit=bounds)
Tracking Colliding Cells
- 1. Tracking Colliding Cells Nhat ¡®Rich¡¯ NguyenFuture Computing Lab
- 3. White Count is a blood test to measure the number of white blood cells.
- 4. In a drop of blood¡Number of white cellsblood cancer50,000stress, viral infection, drug intake25,000healthy5,000flu, poisoning0
- 5. It isimportant to keep track of white blood cells.
- 8. Video of white blood cells via a microscope
- 11. As many cells move at a wide range of speeds¡Collisions
- 12. abruptchange
- 13. Smoothness ConstraintsRegion Abroken tracksRegion Arobust tracks
- 16. Smoothness ConstraintRegion Abroken tracksRegion ARegion Arobust tracksOur Method
- 17. The first tracking method for colliding cells.
- 18. Training100 cell samples100background samples
- 19. Detection Classify each pixel as a Cell or Background
- 20. Trackingtime
- 21. Kalman FilterPopularExtensively used for tracking.OptimalEstimate the most probable state.SimpleTwo steps: predict and correct.
- 22. No CollisionCollisionsmoothsmooth & abruptreliabilityflexibility?Kalman filter
- 29. The reliability of the Kalman filter, the flexibility of multiple hypotheses.
- 30. Tracking Steps
- 31. 123
- 32. 123
- 33. 123
- 34. 123
- 35. 123
- 36. 123
- 38. stay colliding123split awaykeep moving
- 42. Data8300~6Kimage sequencescells tracks cell positions
- 44. Compared MethodsSC Smoothness Constraints Single Hypothesis Multiple Hypotheses SHMH
- 46. Percentage of Tracked PositionsMHSHSC
- 48. Impact of detection MH SHSC
- 49. Given adequate detection results, our method covers 88% of colliding cell positions.
- 51. ConclusionThe first tracking method for colliding cells.The reliability of the Kalman filter, the flexibility of multiple hypotheses.Excellent cell positions coverage.Non- CollidingColliding88%
- 52. Thank you.
- 53. Questions ?
- 55. Bonus ºÝºÝߣs
- 56. S. J. Schmugge, S. Keller, N. Nguyen, R. Souvenir, T. H. Huynh, M. Clemens, M. C. Shin. "Segmentation of Vessels Cluttered with Cells using a Physics based Model". 11th International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), New York, September 6-10 2008.N. Nguyen, S. Keller, T. Huynh, M. Shin. ¡°Tracking Colliding Cells¡±. IEEE Workshop on Applications of Computer Vision (WACV), Snowbird, UT December 07-09 2009.[to be submitted to IEEE Transactions on Medical Imaging]N. Nguyen, S. Keller, Eric Norris, T. Huynh, M. Shin. ¡°Tracking Colliding Cells in In-Vivo Intravital Microscopy Images¡±. Publications
- 57. Example of Multiple cell tracking
- 59. White Blood CellsCirculate in your bloodDefend you against bacteriaProtect you from disease
- 61. Previous Automatic MethodsRay et al. [2002]Active ContourCui et al. [2005]Monte CarloMukherjee et al. [2004]Level Set Analysis
- 62. Previous Automatic MethodsEden et al. [2005]SmoothnessConstraintsLi et al. [2005]Lineage ConstructionSmith et al. [2008]Probabilistic Formalization
- 63. Variation in cell appearances within an imagetimeVaried appearance of a cell over time
- 65. ChallengesIn a collision, cell motion and appearance1. could be different2. change abruptly
- 66. Approach To improve tracking accuracy of colliding cells by:having separate collision statesto describe cells inside and outside of collisionstesting multiple hypothesesof cell motion and appearance as transitions between abrupt motion patterns.
- 67. AdaBoostIdea: combine many ¡°rules of thumb¡± to a highly accurate prediction rule.Input: visual features from training samples.Schema: maintain a strategy to determine ¡°rules of thumb¡± using weight distribution.Output: a single strong classifier which is a linear combination of the set of weak classifiers.
- 68. Detection Procedure Scan each pixel p in the imageCompute image feature vector V from a window centered around pClassify p as a Cell pixel if the feature score in V satisfies the learned decision rule; otherwise classify p as a Background pixel.Cluster groups of Cell pixels into cell observation.
- 69. Motion and Appearance ModelCollision States:State Transition:Hypotheses:to predict the state in the next framecontrol input vectorState Vector*:Motion and Appearance Models: (for )state transitionmatrixcontrol input matrixprocess noise vector ~N(0,Qs)Observation Vector:
- 71. Performance RMSERMSE : Root mean squared errors of position (pixel)-0.03-0.17+0.36-0.21+0.33-0.20SH introduces additional error in positions.MH does not introduce any additional error.Estimating colliding cells¡¯ positions is more difficult.
- 72. Collision Duration RMSE The effect of collision duration on RMSE
- 73. Impact Detection RMSE The impact of detection on RMSE-0.17-0.13-1.05-1.09Different improvement between dataset.Different improvement between methods.
- 74. Future Work1. Add more features to improve detection.5768
- 75. Future Work2. Incorporate a probabilistic approach to transition between collision states.72737576
- 76. Future Work3. Expand to track cells with more complex motions and behaviors.49505152
- 77. Detection PerformanceRecall : TP / (TP + FN)75%Precision: TP / (TP + FP)77%
- 78. Collision Duration The effect of collision duration on tracking6112Exclude SC from being considered for collision.Classify colliding positions into bins based on the number of frames of the collision.colliding cellsbins ofcollision duration
- 79. Detection Impact The impact of detection on tracking38596Data with good detection results before and after collision (+/- 2 frames)cell positionstreatedcolliding cells
- 80. Performance TablePTP: Percentage of Tracked Positions (%)+27+9+23+3+4+24+28
- 82. More ResultsEden et al. [2005]Our Method
- 83. Tracking StepsPredict motionPredict collisionGet measurementsGet errors in position & area Match with minimal error
- 84. Collision States:Hypotheses:to predict the state in the next framecontrol input vectorState Vector*:Motion and Appearance Models: (for )state transitionmatrixcontrol input matrixprocess noise vector ~N(0,Qs)Observation Vector:Measurement Model:measurement noise vector ~N(0,R)measurement transition matrix
- 85. State Vector of cell i :Predicted State Vector:Zero MatrixZero MatrixZero VectorZero VectorPredicted Covariance:
- 86. Predicted State Vector:Hypothesized Measurement Vector:measurement transition matrixError of hypothesis :observation from the detectorweight vectorRule 1:Rule 2:error threshold of Unlikely (i, k) pairStop corresponding condition:
- 87. Remaining Observation :new cellleukocyte typicaldiameterareaRemaining Cell :Not corresponded for 3 frames:Updated State Vector :Kalman gainUpdated Covariance:depends on the cell current state s Eliminate the affect cause by abrupt change in collision
- 88. Training100 Cell Samples100 Background SamplesFeaturesMean IntensityStandard Dev. of IntensityRadial Mean? Decision Rules on feature scores
- 89. Collision DurationThe effect of collision duration on PTP
- 90. H01H00No Collision(s = 0)Collision(s = 1)H11H10
- 91. MeasurementsCell matchesCell DetectionCorrespondenceUpdateCellimageFinishedtracksTracksPredictionsMultiple HypothesesH00:No Collision ¨C No CollisionH01:No Collision ¨CCollisionH11:Collision ¨CCollisionH10:Collision ¨C No Collision
Editor's Notes
- #33: Motion model