In biological research, microscopy is a powerful tool to investigate morphological changes in tissues and cell cultures. Modern setups are able to produce 100000 and more images per day with increasing image quality. The images need to be stored, annotated and analyzed.
This massive amount of microscopy data is too large to be analyzed manually. There is a clear need for automatic software-based analysis. This project aims to provide a collaborative platform for scientists to work with large collections of microscopy images and assist them with state-of-the-art analysis algorithms.
While research questions are evolving, the analyses need to evolve as well. Whereas early applications of microscopy image analysis were simple cell counts, today phenotypes are expected to be detected automatically with machine learning approaches.
We hope to achieve best results by implementing a collaborative platform on which biologists and computer scientists work together. It will provide a full pipeline from intuitive data management to the analysis and computation of large data sets. Managing many large data sets together will also enable us to gather meta data and perform more powerful analysis in a bigger context.
Collaboration will not only happen between biologists and computer scientists. Data can also be shared between biologists or even published, for example, to provide supplemental data to a journal publication. The developers of the analyses can furthermore collaborate on code and manage the code for the analysis modules.
The platform will be built using web-based technologies with a responsive real-time user interface using Meteor. Meta data will be stored in the non-relational database MongoDB, whereas the actual microscopy images will be stored in a separate object storage.

Single cell tracking in phase-contrast microscopy

To investigate the migration behavior of cells, we have developed a method, which tracks single cells in phase-contrast microscopy. In contrast to staining based approaches, the cellular dynamics remain unaffected. The extracted trajectories can be analyzed further with respect to experimental conditions (e.g. signal proteins, mutations).

Analysis of cell trajectories with Markov State Models

One scenario is the application of growth factors next to a cell population. Assuming that cell migration is a stochastic process, we can derive a Markov state model. This enables us to compute meta-stabilities and committor probabilities. Growth factors are known to attract cells, and we can quantify this effect with our methods.

Collective migration in scratch assays

In a second scenario, we use the trajectories to investigate collective migration. Therefore, we quantify neighborhood relations between cells and can then show how mutations or knock-outs affect the group behaviour (collectivity).

Tracking and quantification of neuronal growth cone dynamics in 3D two-photon microscopy

Two-photon microscopy enables studies about intact nervous systems and investigation of growth cone dynamics. In 3D microscopy image time series, we want to analyze the stochastic growth cone and filopodia dynamics in robust brain wiring and develop computational methods to quantify filopodia dynamics. Furthermore, we want to detect synaptic protein trafficking and turnover in synaptic maintenance in 4 (3D + time) to 5 (4D + fluorescent sensor changes) dimensions.