Clone wiki

NanoJ-SRRF / Getting set up with SRRF

How do I get set up?

Frequently Asked Questions

1) What imaging conditions do I need to run SRRF?

  • Generally, you need to be imaging at high magnification achieving a pixel size ~100-150nm with an objective of numerical aperture >= 1.2.

  • For live-cell, the rule of thumb we've used is to acquire data at 100 frames-per-second (FPS), generating 1 Super-Resolution frame for every 100 raw-frames. For most detectors to achieve 100 FPS you will need to reduce the size of your imaging field-of-view (e.g.: setting up an acquisition ROI for EM-CCDs or scanning confocals). Of course, 100 FPS is just a suggestion, one that has worked well for our imaging of cytoskeleton dynamics labelled with GFP, other frame-rates may work well also. However, for lower frame-rates keep in mind that the cell motility will impact resolution due to motion blur. With these settings we generally get ~60-150nm resolution.

  • For fixed-cell, the same suggestions apply, however longer frame sequences will generally provide higher resolution and higher signal-to-noise ratio in the reconstructions (e.g.: 1000-5000 raw frames to produce a SR reconstruction). Resolutions of tens of nanometers can be achieving by using protocols similar to those used in PALM or STORM-type of experiments.

2) How do I achieve 3D SRRF (Z-stacks)?

  • First, please note that SRRF only improves resolution in the XY-axis and not in the Z-axis (for now).

  • Similarly to 1), you will need a time-series to generate each SRRF super-resolution frame. As such, you need to collect a few frames within the same focal region before moving the objective and repeating the process. One example would be to acquire 100 frames in time, move in Z, acquire another 100 frames in time, move in Z, and keep repeating... Then on the NanoJ-SRRF software interface set the "Frame per time-point" to 100. This will mean that on the stack generated by SRRF, every frame represents a different Z-position (example here).

3) Where can I find a simple training dataset?

  • The SMLM Challenge 2013 has a beautiful collection of datasets, look for the "Real Experiments" section. Our favourite is this one.

4) What graphics card do I need to run SRRF?

  • SRRF uses OpenCL for computation, most cards have support for it. ATI cards work great. NVIDIA cards perform well but have random bugs, generally dark squares appearing in images, they are rare and will generally disappear if you run the analysis again. Make sure you have the latest graphics card drivers. We haven't found an ATI or NVIDIA that didn't work yet, but we haven't tested extensively.

  • If a compatible card is not found, SRRF will tell you and will run in the CPU instead (2 to 10 fold slower than on GPU).

  • Here's a small list of cards we use in the lab and that we've tested SRRF against: AMD Radeon R9 200 Series, Intel HD Graphics 6000, NVIDIA GeForce GTX 970.

5) Can I run SRRF on a maximum intensity projection (MIP)?

  • No. We strongly advise you not to do it, a MIP is formed by a non-linear projection that will not allow SRRF to achieve a high-quality reconstruction.

6) How do I acquire a non-continues time-lapse for SRRF (e.g.: SRRF image every 1 minute)?