Automatically Improving Cell Segmentation In Time Lapse Microscopy Images Using Temporal Context From Tracking And Lineaging

Automatically Improving Cell Segmentation in Time-Lapse Microscopy Images Using Temporal Context From Tracking and Lineaging

Over the past decade biologists and microscopists have produced truly amazing movies, showing in wonderful detail the dynamics of living cells and subcellular structures. Access to this degree of detail in living cells is a key aspect of current biological research. This wealth of data and potential discovery is constrained by a lack of software tools. The standard approach to biological image analysis begins with segmentation to identify individual cells, tracking to maintain cellular identities over time, and lineaging to identify parent-daughter relationships. This thesis presents new algorithms for improving the segmentation, tracking and lineaging of live cell time-lapse microscopy images. A new ''segmentation from lineage'' algorithm feeds lineage or other high-level behavioral information back into segmentation algorithms along with temporal context provided by the multitemporal association tracker to create a powerful iterative learning algorithm that significantly improves segmentation and tracking results. A tree inference algorithm is used to improve automated lineage generation by integrating known cellular behavior constraints as well as fluorescent signals if available. The ''learn from edits'' technique uses tracking information to propagate user corrections to automatically correct further tracking mistakes. Finally, the new pixel replication algorithm is used for accurately partitioning touching cells using elliptical shape models. These algorithms are integrated into the LEVER lineage editing and validation software, providing user interfaces for automated segmentation, tracking and lineaging, as well as the ability to easily correct the automated results. These algorithms, integrated into LEVER, have identified key behavioral differences in embryonic and adult neural stem cells. Edit-based and functional validation techniques are used to evaluate and compare the new algorithms with current state of the art segmentation and tracking approaches. All the software as well as the image data and analysis results are released under a new open source/open data model built around Gitlab and the new CloneView interactive web tool.

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015

The three-volume set LNCS 9349, 9350, and 9351 constitutes the refereed proceedings of the 18th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2015, held in Munich, Germany, in October 2015. Based on rigorous peer reviews, the program committee carefully selected 263 revised papers from 810 submissions for presentation in three volumes. The papers have been organized in the following topical sections: quantitative image analysis I: segmentation and measurement; computer-aided diagnosis: machine learning; computer-aided diagnosis: automation; quantitative image analysis II: classification, detection, features, and morphology; advanced MRI: diffusion, fMRI, DCE; quantitative image analysis III: motion, deformation, development and degeneration; quantitative image analysis IV: microscopy, fluorescence and histological imagery; registration: method and advanced applications; reconstruction, image formation, advanced acquisition - computational imaging; modelling and simulation for diagnosis and interventional planning; computer-assisted and image-guided interventions.

Genome Research