Confocal Images

In a perfect world we could extract sharp, instant, 3D images to use for structural analysis, and use a time series of such images to investigate the system dynamics. The reality is a compromise between resolution, timing and the size of the image volume.

To ensure good resolution in the direction of gravity, we use a confocal microscope laid on its side. This is a hazardous (or potentially expensive) thing to do, and for this reason we use a microscope that is relatively old and slow by modern standards. For us a typical 3D stack comprises 64 slices, where each slice measures 107.3 microns square (512 x 512 pixels). Since the typical slice thickness (the step distance travelled by the sample along the optic axis) is 0.48 microns the depth of our image region is 30.72 microns. It usually takes 250 seconds to acquire each stack, so movement within the sample must be sufficiently slow that a stack approximates a snapshot of the system. More slices for a given image volume would give better resolution along the optic axis (our z direction) but at the expense of increasing the stack acquisition time. Fewer slices of fixed slice thickness would speed up the imaging but make the imaged region unacceptably thin. Overall, the dynamics has to be fast enough that something actually happens, yet slow enough that on a microscopic scale individual features – droplets – move about slowly enough that we can actually image them. In addition there are issues concerned with the lens working length and residual sample opacity, both of which limit the thickness of the image volume. Our images are taken using an oil immersion lens with a numerical aperture of 1.4. The sample is housed in a modified glass cuvette with one face replaced by a cover glass. The first slice of an image volume is taken at 10 microns into the sample, rather than at the glass surface, to limit wall effects.

Below is an example of a (false colour) 3D image stack. The system comprises 30% oil, using our index and buoyancy matched mix, (see emulsion system) aggregated with 0.9% PEO. The direction of gravity is down.

Download movie clip of 3D image stack [AVI, 450Kb]

The network of aggregated droplets is clearly visible. The image deteriorates with depth into the sample, as expected. This image has undergone the regularized deconvolution “cleanup” step but has not been improved or doctored in any other way.

Another useful way to visualise the emulsion image is in terms of a 3D rendering. Below left, is the top slice of a 3D rendering of the same data set as above. The top slice matches the above image to help you get your bearings. Following it, below right, is the full 3D rendered stack:

topsliced rendered stack of emulsion	3D rendered image of an emulsion

Gravity points down from “3 o’clock” to “6 o’clock”. The z-axis is along the short distance of the image slab, with z downwards corresponding to increasing penetration into the sample. The axes of the image are isotropic in physical distance.

Two things are immediately apparent from this image. Firstly, though droplets are spherical – the upper face of the image shows disks corresponding to slices through spherical droplets – in the z direction these spheres are smeared out and overlap to give what looks like some kind of geological formation. The smearing in z is a consequence of the limited confocality of the microscope. This smearing must be taken account of in the feature finding stage.

The other point to note is that there are clear spaces containing no droplets, and that some of these extend through the entire image volume in the z direction.

We have tried imaging small volumes to enable us to follow fast dynamics. There are two problems with this. Firstly the field of view in x and y is very limited, but more significantly the z direction is so small that spherical droplets become cylinders and it’s hard to generate reliable z coordinates as a result.

We have also created images of emulsions containing lower PEO concentrations. Though the dynamics of these systems are too fast to allow full three dimensional feature location and tracking, we have been able to capture some remarkable movements in two dimensions (see section Network ripples). We have also captured the collapse of a transient gel (see section Gel Collapse Movies).