Computed microscopy: quantitative, deep-tissue imaging

Optical microscopy is the most widely used imaging tool in laboratories all around the world. Indeed, According to BCC market research, the global optical microscopy market will be worth US$6.3 billion in 2020. Several Nobel prizes have been awarded for contributions made to the development of optical microscopy, including most recently in 2014. There is, however, a major limitation facing optical microscopy: it is difficult, if not impossible, to image tissue hidden beneath layers of overlying tissue. This occurs for the same reason that it is difficult to see clearly through a window covered in rain drops - tissue is highly scattering, like rain drops, and critically degrades image quality. This is important as it prevents in-tact tissue from being imaged in its natural environment, requiring tissue to instead be sliced into thin sections.

A variety of approaches have been used in an attempt to overcome this problem. All such approaches are generally similar in that they insert hardware into the microscope in an attempt to compensate for the degradation due to the sample. This is similar to humans using spectacles to overcome imperfections of their eye. The main difference is that opticians are able to precisely determine the imperfections that each eye has, and thus design spectacles which perfectly compensate for them. No such method has been developed for measuring sample induced imperfections, or aberrations, present in microscope images.

This project proposes to do just that: measure the imperfections caused by the sample itself. This will be achieved by computing the optical structure of the sample (i.e., how light travels in the sample) via a two stage process. Firstly, the sample will be imaged by a microscope capable of performing rapid three-dimensional imaging called an optical coherence microscope (OCM). OCM works very much like ultrasound imaging, except light is used instead of sound waves. The second step involves developing a sophisticated computational procedure for calculating the sample's optical structure from the OCM image. This will be performed using a recently mathematical model, developed recently by the project team, which allows OCM images to be predicted from a given sample structure. Clearly, our task is to solve the opposite problem: calculate the sample's structure given a measured OCM image. Formal techniques have been established for solving the problem in the opposite fashion which will be adapted specifically for this project.

Once the sample's optical structure has been solved, in a follow-on project, existing methods will be employed for restoring optical fluorescence microscope images which have been degraded by the sample itself. This will enable fluorescence microscopy to be performed at depths within tissue which are currently inaccessible. This will be highly advantageous to many biological researchers in the UK and the world.

There is currently much interest in performing super-resolution fluorescence microscopy, due to the incredible insight offered to many biological research questions. Diffraction limited fluorescence microscopy remains a very useful tool and has been the mainstay of biological researchers for decades. However, super-resolution and diffraction limited fluorescence microscopy cannot be achieved beneath layers of scattering tissue, preventinh fluorescence microscopy from being performed on tissue in its natural, unsectioned, state.

Whilst hardware approaches like adaptive optics have been implemented, they are yet to be experience mainstream uptake because of the lack of a guide-star within tissue. In particular, unlike in astronomy, known features rarely exist within samples which allow for the adaptive optical elements to be conditioned to the particular sample under study. The approach suggested here will use computation to restore fluorescence microscopy images by first rigorously and robustly solving the inverse problem of optical coherence microscopy (OCM).

OCM is a scanning optical microscope which obtains depth information using low-coherence interferometry. Solutions to the OCM inverse problem which make the first Born approximation have been developed. Clearly, since we aim to overcome degradation due to highly scattering tissue, we cannot make the first Born approximation. Instead, we will iteratively solve for the sample's refractive index distribution, iteratively, using a recently demonstrated full wave forward model of optical coherence microscopy developed by the PI. The co-I of this proposal is an expert in the solution of inverse problems. This project thus seeks to employ established methods for the iterative solution of inverse problems to solve the inverse problem of OCM using a highly novel forward computational model of OCM. Restoration of fluorescence images using established techniques will be deferred until a follow-on project.

This proposal seeks to provide proof of principle of a new method of performing both fluorescence and quantitative phase microscopy using a combination of hardware and computation. The beneficiaries are thus quite broad as outlined below.

Microscope users:

If commercialised, computed microscopy could revolutionise fluorescence microscopy by allowing microscopy of thick samples rather than thin sections. This enables tissue to be imaged in its most natural state, including in-vivo. This would allow fluorescence microscopy of animal brains, in vivo, at unprecedented depths in tissue, which would be revolutionary in neurobiology. In another example, the study of tumour invasion mechanisms would benefit through obtaining fluorescence microscopy images of synthetic tumours, which is currently impossible due to their thickness. Computed microscopy should allow for established super-resolution microscopy techniques, excluding stimulated emission depletion microscopy, to be applied deeper in tissue than is currently possible.

Scientist training:

The skills applied in this project are specialised and central to optical imaging. PI Munro benefited from specialised training by his PhD supervisor to learn the theory of coherent optical imaging, and co-I Arridge is leader in the solution of inverse problems in optical imaging. Together they will provide a valuable opportunity for a post-doctoral research associate to be trained in these disciplines, which are crucial to the UK's future scientific and economic development.

UK economic development:

Computed microscopy presents a valuable opportunity for the UK to develop economically. Although too early to predict the project's success, if it is, it will lead to an export in a highly lucrative global market (exceeding US$6 Billion by 2020). Computed microscopy adds value to fluorescence microscopes. Furthermore, development of computed microscopy will be less expensive than designing a new imaging system, for example. Thus, if successful, commercialisation will likely be rapid and economically rewarding.

Optical coherence tomography researchers:

Researchers in optical coherence tomography (OCT) have been developing models of image formation for over two decades. They have had to wait until recently to have a model which is highly realistic. The image formation model central to computed microscopy will itself be of high interest to OCT researchers for analysing a range of imaging scenarios. The inversion technique will be even more beneficial as it will transform OCT from a noisy, qualitative, imaging techniques, into quantitative techniques with fewer artefacts.

Clinical translation of optical coherence tomography:

OCT is used widely in clinical practice only in ophthalmology. This is because retinal imaging is the ideal sample for OCT imaging since the retina is of the order of 200 micrometres thick, and can be imaged through the lens of the eye itself, making it possible to image through the entire retina. Despite this, it can still be quite difficult to identify different layers within retinal scans. Solving the inverse problem of OCT will result in a much clearer, quantitative image, which will enable easier and more accurate image interpretation. This could also enable improved automated diagnosis of retinal abnormalities.

Students and researchers using optical coherence tomography:

The forward model of OCT image formation will allow students and researchers an opportunity to probe details of OCT image formation and to learn the principles of image formation through simulation. This could be integral in the training of the next generation of OCT users and developers.