Sparse 4-D multi-spectral optical computed tomography for accurate quantification of voxel composition

This project is about developing a new type of microscope called a multi-spectral absorption optical computed tomography scanner.

Starting with the development of x-ray computed tomography - the so called CAT scanners that revolutionised medical diagnosis in the 1970's and '80s - three-dimensional imaging has become ubiquitous in modern medicine and science. Principles of computed tomography are employed in fields as diverse as remote monitoring of flames in chemical engineering, tomb-hunting in Egyptian pyramids (via muon tomography) and non-invasive mapping of radiation dose to provide quality assurance in radiotherapy treatments.

Optical computed tomography - or "optical projection tomography" (OPT) as the configuration widely used in biological microscopy has come to be known - provides high quality images that are especially useful at the so-called mesoscale, with sample dimensions of the order of cm and able to resolve details as small as a 5-10 microns. Two principal forms of OPT are available, emission and absorption tomography. This project is concerned with absorption tomography, in which a beam of light shines on a sample and we form a 3-D image from the light that passes through the sample, thus measuring the amount by which the transmitted light decreases because of optical absorption by the various tissue components in the sample.

Tissue absorption occurs to different degrees at different wavelengths. Optical absorption spectra are highly distinctive and may be used to separate out important components of tissue. This technique is routinely used in traditional 2-D microscopy in order to locate regions on the slide containing different types of tissue, which are "stained" with different types of chemical in order to make them more visible and more distinct from each other. Measurements are made at a (normally, relatively small) number of discrete wavelengths and analysed by computer in a process that is know as "spectral unmixing".

In order for this unmixing process to work well, we need to adjust the concentrations the chemical additives (stains) such that they let through an appropriate amount of light in the range measureable by our camera. However, for the sorts of 3-D animal model that we use in cancer, this adjustment is very difficult to do, because there are significant unknowns in the processes by which these "contrast agents" are delivered to tissue. Principles of responsible animal experimentation, involving a reduction in the number of animals used and a refinement of techniques argue strongly for the development of a "smart" readout system, as suggested in this project, to gain maximum value from each animal used.

A second reason for possible failures in the unmixing process is that we might not know in advance what the spectra of the compounds to be unmixed are. The spectra of an exciting new family of contrast agent, based on gold nanoparticles, can vary when they are introduced into tissue. This is unfortunate, as it hinders our ability to make accurate measurements of nanoparticle concentration via spectral unmixing. There has been an explosion of interest in gold nanoparticles in recent years, and particularly in one of the subtypes, known as gold nanorods. These can be attached to various molecules that recognise particular signatures of cancer within the body and, thus, the hope is that by measuring accurately where the nanorods congregate in tissue, we can detect and, perhaps, quantify the amount of malignant tissue.

The data acquisition and processing technique used in this project is very new in OPT. It will reduce the amount of data that need to be acquired by a factor of approximately. This is extremely important because spectral OPT involves very large datasets. Our project consists of three parts: (i) simulate and optimise the acquisition process; (ii) build the new microscope; (iii) test it in two problems of great interest in cancer.

Optical Computed Tomography (CT) is a method of 3-D imaging that has found widespread use in the physical and biological sciences. The imaging geometry most widely used in biology at the current time is known as Optical Projection Tomography (OPT). OPT may be used in either absorption or emission modes with emission tomography being the more common because of the wide range of fluorescent contrast agents available.

However, this project will study absorption mode OPT, because of its excellent quantitative abilities. The project will apply spectral unmixing to 4-D OPT data (one spectral and three spatial dimensions). It is non-trivial to extend traditional techniques, as developed for routine histopathology, to OPT of cancerous tumours for the following reasons:

1. It is difficult to control the concentration of chromophore in the tissues of interest in such a way as to optimise transmission over large fields-of-view.

2. In important use cases, the reference spectra of the species of interested can vary depending on the local microenvironment. This makes quantitative determination of the analyte concentrations susceptible to significant error.

3. The volume of data that must be manipulated during the acquisition and processing of 4-D data at high resolution from mesoscopic fields-of-view is extremely large.

Our hypothesis is that "true 4-D" (i.e., acquiring many spectral points) multispectral absorption tomography will allow significantly improved quantification of tissue composition and tracer concentration on a voxel basis compared with measurement at a small number of predetermined wavelengths.

We will test this hypothesis as follows:

(a) Simulate and optimise a sparse imaging (also known as compressed sensing) strategy for data acquisition.

(b) Use these results to design and build a multi-spectral absorption mode OPT scanner.

(c) Test the new scanner in a murine model of neuroblastoma and in phantoms containing gold nanorods.

Academic impact

* The success of the project will enhance the knowledge, capabilities and reputation of the ICR in the field of biomedical imaging.

* The biomedical application of multi-wavelength OPT is still in its infancy. This proposal will pave the way for the discovery and qualification of the sensitive and specific biomarkers of response in a preclinical cancer setting.

* Discoveries made using optical CT are potentially translatable into the clinic, in the form of methods for diagnosis, therapy or monitoring treatment response.

Economic and societal impact

* Close collaboration with drug developers in both academia and the pharmaceutical industry, to accelerate and achieve maximum impact on scientific and industrial development of new therapies, and ultimately on the delivery of health improvements through the NHS and other providers.

* Industrial collaborations will establish the value of both existing and emerging functional imaging biomarkers for the assessment of novel targeted therapeutics and guide their incorporation into planned clinical trials.

* Dissemination of imaging approaches to other hospitals will benefit both patients and the NHS.

* Societal impact is achieved through the widest dissemination of research results, both in the scientific literature and to the general public.

* Our work develops younger researchers conversant with the full range of techniques that are being developed and applied for imaging cancer, and which will also enable them to thrive in other areas of academic or industry- based biomedical research. Study at the ICR also instructs young researchers in a range of complementary skills.