Image formation in structured illumination wide-field fluorescence microscopy

Abstract

We present a theoretical analysis of the image formation in structured illumination wide-field fluorescence microscopy (SIWFFM). We show that the optically sectioned images obtained with this approach possess the optical sectioning strengths comparable to those obtained with the confocal microscope. We further show that the transfer function behaviour is directly comparable to that of the true confocal instrument. The theoretical considerations are compared with and confirmed by experimental results.

Introduction

The confocal microscope possesses a number of advantages over the conventional optical microscope, and the optical sectioning is arguably the most important one (Wilson and Sheppard, 1984, Wilson, 1990, Pawley, 1995). The traditional construction of a confocal microscope employs a point light source together with a point pinhole detector. It is the presence of the pinhole aperture which is responsible for the optical sectioning since it physically blocks light from out-of-focus planes from reaching the photodetector and hence precludes this light (information) from contributing to the final image. A complete image of the object is built up by scanning. These systems, which are usually built as an ‘add-on’ to a conventional microscope, are very powerful and have found great application is a wide number of fields. They normally require laser light sources since standard microscope illumination systems are usually insufficiently bright. In order to overcome these problems a number of approaches have been taken. Among those there are two developed at the University of Oxford. The first one has concentrated on improving the design of traditional pinhole-based systems (Juškaitis et al., 1996, Wilson et al., 1996), whereas the second has attempted to modify the conventional optical microscope in such a way as to introduce optical sectioning (Neil et al., 1997). If we define optical sectioning as the requirement that all spatial frequencies within the transfer function attenuate with defocus then it is clear that the conventional microscope does not possess this property. However, if we consider for example the conventional fluorescence case then we find that it is only the zero spatial frequency which does not attenuate with defocus; all the other frequencies are imaged less strongly with increasing defocus. This suggests that if we modify the illumination system in a conventional microscope, so as to project a single spatial frequency fringe pattern onto the object, the microscope will image those parts of the fringe pattern efficiently whose projection onto the object lies within the focal region. We thus obtain an image of the object in which the focal optical section is labelled by a sharply focused grid of lines. Simple processing of three images, taken at three relative spatial positions of the fringe pattern, permits both an optically sectioned and a conventional image to be extracted in real-time. The approach has been demonstrated in both brightfield reflected light imaging (Neil et al., 1997, Wilson et al., 1996) as well as fluorescence (Neil et al., 1998, Neil et al., 2000, Lanni and Wilson, 2000). In this article we theoretically describe the fluorescence case using two methods of obtaining structured illumination: fringe projection and grid projection. In the fringe projection case the structured pattern is formed by the interference of two coherent illumination beams, so as to create the appropriate pattern on the specimen. In the grid projection approach the structured pattern is formed by imaging a physical grid object on the specimen.

There have been numerous attempts to exploit structured illumination in microscopy to achieve different goals. Notably, Gustafsson has shown that using basically the same principle as described in this paper it is possible theoretically to improve the lateral resolution of the optical microscope by a factor of two (Gustafsson, 2000). Further, exploiting the nonlinearity properties of fluorophore saturation theoretically an unlimited improvement is possible (Gustafsson, 2005). However, in the reality the improvement is limited by the signal-to-noise ratio and the photobleaching of imaged fluorophores. Bailey et al. (1993) introduce standing wave excitation in the fluorescence microscopy, by which they achieve improved axial resolution. Failla et al. (2003) name this type of microscopy spatially modulated illumination microscopy and claim it can be used for the measurement of the structures with accuracy close to 1 nm. This approach also uses structured illumination, but unlike of the approach described in the paper, its sinusoidal pattern is laid along the illumination axis, which complicates the system setup, and limits its potential applications.

The analysis in this article emphasises the optical sectioning side of the approach, and presents experimental results which confirm that the optically sectioned images of high quality can be achieved by this method.