Making your tile scan look less... tiled

Tile scan imaging can produce stunning images containing both detailed information and wider tissue context. Done poorly though, and the image can suffer from quilting effects and misalignments. Here are some tips and tricks to get them right.

What is a tile scan?

Tile scanning is an approach to capture a large region of interest at high resolution by capturing a grid of slightly overlapping images. In fluorescence microscopy, the achievable resolution comes down largely to the numerical aperture of the objective lens. However, in general, the numerical aperture of an objective lens tends to increase with its magnification. This poses an important trade-off: resolution for field of view. If you wish to capture the finer details across a large region of interest, a tile scan can allow you to do this.

What equipment do you need?

Tilescans are not restricted to slide scanners, they are possible on any microscope with a motorised XY stage, motorised Z-drive and software that can support it, such as our Zeiss LSM800 and Olympus iX81.

How do you capture a tile scan?

The specific steps to capture a tile scan will be dependent on the system and software, so if you wish to do these on our systems, get in touch with Rebecca Saleeb for some one-to-one application training. In general: it is important to first calibrate the stage (the software uses XYZ coordinates to identify its position; calibration will ensure the stage coordinates are correctly recorded and reproducibly returned to). You will then need to specify the region that you wish to capture, the overlap between tiles, the order/direction to capture the tiles and the focus strategy you wish to use. 

Minimising the tile effect

When viewing a single image, your eye may not notice shading variation across the field, but this becomes quickly apparent as a 'quilting' effect in a tiled image (exemplified right). 

A major source of image vignetting is uneven illumination across the field, resulting from inherent limitations in the optical systems. Reduce this by ensuring the microscope is set up correctly (in particular, kohler illumination should be verified for brightfield techniques).

 

Further 'shading correction' can then be achieved through image processing post-acquisition. Ideally, capture a 'flat field' image for each of your channels (an image of known homogeneous signal, e.g. pure dye). You can then reverse the shading effect by simply dividing the raw image with your flat field image. Flat field images can usually be stored by the acquisition software for automatic image correction following capture.

If you cannot capture an experimental flat field image, check out the BaSiC plugin for Fiji, which allows you to distil the flat field information from your tiled image and then corrects each tile (producing the improvement in the image above) - this can only be done before fusing the image though!

The quilting effect can also be caused by repeated bleaching of overlap regions. In this case you will notice a border around each tile with a clear step change in signal that is further reduced in the corners. This is most apparent in highly autofluorescent tissue. Flat field corrections won't help here. In this case you should reduce the illumination intensity and avoid excessive exposure i.e. remove or minimise autofocusing and stack acquisitions.

 

If tiles can be identified by misalignment or boundary lines between images, then the images have not been correctly stitched (i.e. aligned) and fused. Most acquisition software provides stitching capability, but if you do not have access to this, make note of the number of tiles captured in X and Y, as well as the order/direction. You will then be able to stitch with the free 'Grid/Collection Stitching' plugin for Fiji

 

For successful stitching, ensure you include at least 10% overlap between tiles in the acquisition set up. As long as all channels were acquired per position, you can stitch a single reference channel and apply the same alignment to all. Select a channel with bright structured signal that is present across all tiles. If stitching stack data, try to select a single reference Z-position that is in focus across the field (or stitch the EDF image). When you are happy with the alignment, fuse the tiles into a single image; this will blend the overlap regions, reducing the appearance of lines between tiles. 

CONTACT US

Dr Rebecca Saleeb (Facility Manager)

John Vane Science Centre

Queen Mary University of London

London, United Kingdom

With thanks to our funders, the William Harvey Research Institute and Queen Mary University of London