Making light work of digital imaging

In the biological arena, new and highly useful fluorescent markers are used to stain or 'label' specific structures of interest. They have transformed the range and applicability for optical observation. These labels are excited and correspondingly emit at specific wavelengths; thus, different facets of a specimen can be 'selected' by controlling the wavelength of the delivered and captured light. For example, labels such as 4',6-diamidino-2-phenylindole (DAPI) are used to highlight the nucleus of a cell and MitoTracker Orange is used for mitochondria. Figure 1 shows an example of a multiple stained section, viewed in fluorescence. There has been an explosion in fluorescent labels for examining biological structures, in fixed and live cell preparations.

These labelling methods have now moved from the fixed specimens to live samples. The advent of 'live cell imaging' now means that cell and membrane dynamics, and drug interaction can be fully appreciated. It has extended the field of microscopy considerably; fluorescent labelling is used to monitor live cells in four dimensions (X, Y, Z and time). The ability to add time as a dimension means that we have the building blocks are available to examine biological dynamics in many ways.

The need to record images in the highest possible detail has redefined the physical limits that are normally demanded from the standard microscope. Many high-end imaging systems use a microscope, digital camera and a computer to record the images with high resolution, speed and sensitivity. Automation of the microscope's focus, stage and fluorescence wavelength are frequently included; the flexibility of the system is then greatly enhanced. Key improvements have been made in the spatial, axial and temporal resolution, and the effect of these is now reviewed.

Figure 1

Confocal microscopy

The standard light microscope is limited in spatial and axial resolution. In this wide field mode, the image generated is coupled with out of focus light that degrades the overall quality in terms of resolution and contrast. The numerical aperture (NA) of the microscope objective defines these limitations. It leads to limits on the minimum sizes of structures that can be observed and what additional contributions to an image are present from other planes in the section (the depth of field is greater than desired). To resolve finer structures and improve the contrast, techniques to reduce the system's blur must be used. Such arrangements are said to deliver 'confocal' images (those with a small depth of field) and are available using a variety of technologies.

In Figure 2, we can see the difference between wide field and confocal images. The image that is collected by a system with standard optics (wide field) has restricted control over the depth of field and consequently the haze. A system that has additional flexibility to restrict the depth of illumination and observation (confocal) will remove the haze, and dramatically improve the image quality for resolution and contrast.

Figure 2

The requirements for a flexible fluorescent microscope imaging system would be the following:

• To have the capability to operate as a simple wide field or confocal instrument with great ease of wavelength change, acquisition speed with high signal to noise ratio.

• To be able to view the confocal section without a camera, but 'live' down the eyepieces.

• To have a degree of control to overcome the potential hazards of photo bleaching of the sample and possible phototoxicity to live samples.

What are the various system variants that allow sectional restoration or confocality?

Laser pinhole confocal

One thing that can be done to improve a wide field system, is to limit axial (Z) observation depth: thinly illuminating and observing any section will achieve the desired result. This can be realized by using pinholes in the illumination and observation paths (Figure 3). The effect is to increase the NA and reduce the haze as there is a reduction in the Z depth of the viewed image planes. This is, therefore, a 'light rejection' approach as the out of focus blur is eliminated by light rejection.

Figure 3

The laser confocal scanning microscope (LCSM [Figure 3]), uses a laser as the light source. This gives not only a monochromatic excitation, but is intense enough to satisfy the demand of our system requirements. The beam is projected through the source pinhole to control the illumination characteristic at the sample. The confocal image is controlled by the size of the pinhole and so the illumination plane. The fluorescent image is projected back through the detector pinhole and is 'scanned out', pixel-by-pixel, using a photomultiplier. The result is better X, Y, Z resolution and contrast, yielding structural details that were previously inaccessible by wide field observation.

A laser is directed through the source pinhole and then to the specimen using a dichroic mirror. The pinhole restricts the absolute focal depth that is illuminated. The resultant fluorescent light from the specimen is gathered through the detector pinhole, producing the thinly illuminated confocal plane for observation. Thus, only 'in focus elements' are collected as the image plane. Every point on the plane is scanned, producing the final image.

This type of confocal has been successful and commonplace for many years. It offers deep sample penetration (>50 μm), wavelength excitation and emission control by multiwavelength laser arrays; thus, many fluorescent labels can be used in one specimen. As we are using light rejection directly, the signal to noise ratio can be more challenging than a wide field approach. Additionally, this form of image generation can be slow because scanning point-by-point using a photomultiplier takes time. Faster biological events may not be within the temporal time frame offered by these systems.

Typically, the LCSM is supplied as a complete system, and limits the scan speed/image and the wavelengths that can be applied. The laser(s), however, can be harmful to live samples so photo bleaching and phototoxicity is a concern. LCSM systems are in general expensive and require careful alignment. So alternatives have now been developed to overcome these issues.

Deconvolution

Another method for improving the resolution and contrast is by deconvolution. This is not a 'light rejection strategy', but an 'image restoration' approach in which data from an image is re-assigned spatially and axially by calculation to eliminate blur. The principle is that if you know how the 'blur' is generated (your collected image is convolved with the blur), it is within the remit of the system to 'back calculate'(deconvolve) the original image, minus that blur. However, one needs to know the mathematical model that contributes to the blurring (the point spread function [PSF]) and then apply mathematics to reveal the clear image. The PSF can be mathematically generated, or mapped in the system using control fluorescent spheres.

Image restoration by deconvolution is shown in Figure 4. As explained previously, all wide field microscope images are a result of the unperturbed image, 'convolved' with the image blur produced as a result of using a finite NA. This can be modelled by saying the image is convolved with the PSF (Figure 4a). Conversely, a convolved image (the one that you record) can be 'deconvolved' with the known or measured PSF to produce the unperturbed image or crisp image (Figure 4b).

Figure 4

This method effectively works using the wide field image (blur included!) and then removes the blur or haze by modelling. It has the advantage of good signal to noise and acquisition speed compared with an LCSM: images are acquired as fast as possible using a camera, not scanned point-by-point by a photomultiplier. Real-time live cell imaging can be attained in this type of system because of the advent of highly sensitive electron multiplying charge coupled device (EMCCD) cameras that allow very low exposure times.

There can be strict light control in these systems, in terms of wavelength and intensity. This reduces the threat of photo damage and, in live cell samples, phototoxicity. Different light sources (Hg, Xenon, metal halide) can be selected to generate rapid and accurate wavelength control for multidimensional live cell imaging at high precision. As data can be rapidly scanned in, fast events can be tracked in X, Y, Z, wavelength and time.

Post-processing of data is required to execute the deconvolution (Figure 4), but this can be done after the experiment has been run rather than 'on the fly'. The high signal to noise ratio (and potentially low photon flux) is also highly advantageous. Data can be collected in real time and the choice to deconvolve or not is user selectable.

Time lapse imaging of cells or organisms can be performed for long periods, without significant bleaching or phototoxicity, as the light may be kept at a low incident level. Depending on the speed of the events being measured, with the right camera and microscope hardware control, time lapse at multiple plains in 3D can be imaged. Deconvolution systems can be retrofitted to existing wide field arrangements, giving the power, flexibility and costeffectiveness that are often demanded in multi-user environments. Dedicated and configurable systems such as Applied Precision's DeltaVision (Issaquah, Washington, USA) have been performing these tasks for more than a decade.

Spinning disk

Taking the pinhole principle again, we can increase the speed of acquisition by having a dynamically rotating ensemble of pinholes, covering the sample in real-time scanning rates. The Nipkow disk (Figure 5) has multiple sets of spirally arranged pinholes placed in the image plane of the objective lens. This has the advantage that it provides high-speed, multipoint acquisition with a similar degree of confocality to that achieved by a laser-based single pinhole pair (LCSM). Importantly, it manages to achieve this without the same level of cost and complexity.

Figure 5

In Figure 5, we can see the principle of the spinning disk confocal. A rapidly rotating array of pinholes sweeps the sample much faster than the camera acquisition time. The pinholes restrict the plane of illumination and subsequent Z collection-producing confocality close to that of the LCSM.

This approach gives fast scans (up to 50–100 frames/s) as it uses a camera and the field of view, confocally, is constantly updated. These types of solution can be deployed illumination using a laser or a white light source. They provide considerable utility in a multi-user environment.

The nonlaser (white light source) can be a long-life arc source (mercury halide) coupled using an alignment free light guide allowing full spectrum (360–700 nm) imaging of virtually any fluorescent probe. Automation of internal multiposition excitation, dichroic and emission filter wheels permits fast multidimensional imaging of up to five or more fluorescent probes in the same sample, without compromising the confocality.

A highly sensitive camera is usually coupled to the (EMCCD) system. This is most important when you are studying live cell dynamics where the intensity allowable on the specimen may be limited. This type of system will lend itself to being used for very fast dynamics or as a wide field system for ultrafast acquisition at higher signal to noise — or both. These systems are compatible with a wide selection of high-end cooled and noncooled cameras with high dynamic range (16 bit), fast readout, high quantum efficiencies and small pixel sizes that produce images at a high resolution and high signal to noise ratio.

The significant advantage of the spinning disk approach is the ability to capture rapidly occurring events within living cells without compromising axial and spatial resolution. The low intensity illumination through the disk can substantially reduce photo bleaching and phototoxicity. The specimen can be viewed 'live' though the microscope binoculars, as well as using a camera.

One example of this technology is the BD CARV II (BD Biosciences, Rockville, MA, USA) confocal imager. Combined with intensified cameras, it is used to image fast fluorescence changes at rates from 50–100 frames/s, with the above criteria.

Conclusion

Microscopy is greatly enhanced by digital imaging, coupled to resolution enhancing techniques such as spinning disk, deconvolution and laser confocal. Greater flexibility has resulted from improvements in microscope automation, image detection devices and computing power that have driven these developments.

The deployment of deconvolution and spinning disk can give 'time-resolved' 3D images of living cells at high spatial and axial resolution. Thanks to the strict control of light level and the use of highly sensitive cameras, viability and structural studies are now possible in life sciences.

References

1. J.C. Russ, The Image Processing Handbook, 2nd Edition, CRC (FL, USA).

2. C. Hammond and J.P. Heath, Microsc. Anal., 20(6) (2006).

3. C. Hammond, J. Microsc., 192(1), 63–68 (1998).

4. D.B. Murphy, Fundamentals of light microscopy and electronic imaging (John Wiley and Sons Inc., New York ,NY, USA, 2001).

Dr Andrew P. Billington is an applications scientist at Image Solutions Ltd (UK).