[ Home : 3B Microscopy | Localisation Reclassification | Publications | Teaching | Research ]

3B Microscopy Analysis Software

Contents

About 3B Microscpopy

Bayesian analysis of blinking and bleaching, or 3B microscopy, is a method which analyses data in which many overlapping fluorophores undergo bleaching and blinking events, giving the structure at enhanced resolution. By using a Hidden Markov Model (HMM), it allows useful information to be obtained from data that would be impossible to analyse with standard localisation analysis techniques. An example of the type of data suitable for 3B is given in Figure 1.

Video of input data.
Widefield Widefield image created by averaging video.
3B Reconstructed 3B results.
Figure 1: Example input data and reconstruction.
The method is described in the paper Bayesian localisation microscopy reveals nanoscale podosome dynamics, which appeared in Nature Methods on 4 December 2011. Links:

If you are interested in using this method, it is available to download under the GPL (general public license). In brief, this means that you are allowed to use the software, but any software you write which uses this software must make its source code available (so its use in commercial closed source software packages is not permitted).

As with all high resolution microscopy techniques, 3B has advantages and drawbacks. Below is some guidance as to the issues which will determine how well suited 3B is to solving your particular problem.

Going faster

The key improvement that 3B gives is that it can deal with images in which many fluorophores are overlapping, all the time. This allows you to speed up your data collection. Primarily this is interesting because it allows you to look at dynamic processes in live cells; however, it may also be of interest if your system suffers from drift which you are unable to get rid of.

However, note that 3B has a very significant computational cost. Being able to process the data for a single static cell would take several days on a state of the art desktop computer. If you wish to use 3B for large areas or video data, you will need access to a cluster. If you do not have access to a cluster, but you think that 3B offers a significant advantage for your system, then contact us at threeb@coxphysics.com and we may be able to help. Note that if you only need to deal with images in which a few fluorophores are overlapping in each frame, there are alternative analysis techniques such as DAOSTORM which do not have such a high computational cost, and are also freely available.

Imaging for 3B

Picking a suitable structure to image is probably the most important step in achieving good results. Images with a low out-of-focus background are necessary to allow the algorithm to pick up blinking and bleaching events. In practice this means structures should be thin in z (preferably < 500 nm) and not too dense for successful collection of widefield data. If you are able to make measurements in TIRF, the out-of-focus background will be much reduced.

If you are imaging fixed cells using organic dyes, you will need to introduce a reducing agent to induce blinking. Recipes for reducing embedding media are available (e.g. here). If you are imaging live cells using standard fluorescent proteins (such as GFP), you can use either laser or arc lamp illumination. Data needs to be recorded using an EMCCD camera to achieve a good signal-to-noise ratio. In order to speed up data collection, you should create a region of interest around the area you want to measure. However, leave a border of at least 10 pixels larger than the area you are interested in. It is best if there are no in-focus fluorophores in this border region.

Although 3B can extract information from what appears to be very poor quality data, it has limitations. If there is little underlying information in your data, you will not get an image characteristic of the underlying structure. If your reconstructed image includes a visible grid pattern, this is an indication that reconstruction has failed.

Download Software

Previous versions of the software and manuals are here.

ImageJ plugin

Version 1.1

Source code

Version 1.1

Version 1.0

Sample Data

We have provided some sample data and configurations which can be used to test the software: Instructions for usage are given in the manual.

Updated January 31st 2013, 05:29