While working towards my PhD Thesis between 2006-2009 at the University of Kent, and in collaboration with the British Museum, the National Gallery and Nottingham Trent University, I looked at the potential for a high resolution optical imaging technique known as Optical Coherence Tomography, or OCT, to assist with the conservation of paintings and other historical artifacts. Financial support for this project was provided by the Leverhulme Trust, and my supervisor was Prof Adrian Podoleanu. I'm very grateful to the funders, collaborators and all the Applied Optics and School of Physical Sciences team in Canterbury for a fantastic three years.
The motivation for this work was that conservationists and historians often find it useful to be able to study an object using high resolution cross-sections. A common practice is to remove a very small sample of the object, often near existing cracks or damage, and examine this under a microscope. Less invasive methods are preferable, but most current techniques, such as infrared photography and x-radiography, provide little or no depth information - they produce 2D representations. OCT is able to provide 'selection in depth', allowing cross-sectional or even 3D images to be acquired without compromising the painting or artifact.
OCT cross-section of a paint test sample, acquired using a spectral domain system built for the art conservation project. A varnish layer can clearly be seen on the right part of the image - knowing how thick this layer is could help with conservation work.
Optical Coherence Tomography (OCT)
OCT is often compared with ultrasound because the images it generates have similar characteristics. In ultrasound, a depth-resolved image is formed by making time-of-flight measurements on sound waves. Reflections from shallow depths arrive back at the transducer earlier than reflections from deeper parts of the object, and we can use this information to build up a cross-sectional image, showing how much sound was reflected from each depth.
The portable time-domain OCT system that I built, in use at the British Museum, looking at an Egyptian shabti figure.
The idea behind OCT is similar, but it uses light rather than sound. This means we can achieve much better resolution, but with the disadvantage that we can’t probe as deep into the object. Since light travels much faster than sound, we can’t make direct measurements of its flight time. Instead, we use a particular type of interferometry, called low coherence interferometry.
In its simplest form, this involves splitting light source into ‘object’ and ‘reference’ beams. The object beam is reflected off the object that we want to image, while the reference beam is reflected off a mirror. The two reflected beams are then recombined at a photo-detector. We see interference effects (the formation of an interference pattern) only when the two distances travelled are equal to within the ‘coherence length’ of the light source. The coherence length is inversely proportional to the optical bandwidth (the range of wavelengths produced). By choosing a source with a large enough bandwidth we can reduce the range over which we see interference effects down to a few micrometers.
If we measure the amplitude of the interference pattern then we are really measuring the amount of light reflected from a specific layer within the object. Only light reflecting from this layer has travelled the same distance as light reflected from the reference arm mirror, and so only it will contribute to the interference pattern. Moving the reference mirror changes the path length of the reference arm, and so alters the depth within the object from which we collect information. So by scanning the mirror over a range of depths, we can build up a complete cross-section.
This approach is known as ‘time domain OCT’. Spectral domain OCT, which has several advantages over time domain, uses an interference pattern encoded in the frequency domain representation of the OCT signal. The implementation is a little different, but the fundamental principle is the same. I worked with both types of system during my PhD.
Art Conservation Project
My work focused on designing OCT systems suitable for use in studying historical artifacts. I built a portable time domain system which was deployed to the British Museum, and I investigated techniques for speckle reduction, mirror terms rejection and dynamic focus. Full details can be found in my PhD Thesis. A complete list of all published work in this field is available at oct4art.eu.
En-face slices from sample of Egyptian faience at varous depths, obtained using the portable time domain OCT system that I built during my PhD. Scale bar is 1 mm.
Speckle is a type of noise which appears in all OCT images. It can be seen as a grainy pattern of light and dark spots where we would expect a uniform intensity. Speckle arises because of interference between light scattered from elements on a scale smaller than the resolution of the imaging system. There are several ways of removing speckle, and I spent some time implementing a method based on ‘angular compounding’ specifically for use in art conservation. It involves acquiring multiple images of the same object from slightly different angles at high speed and then combining these images using computer software. Reference 2 below describes the method in detail, and it’s also discussed in Chapter 4 of my thesis.
Speckle reduction during imaging of ex vivo canine tooth. Top image: single conventional OCT B-scan; Middle image: Incoherent summation of 20 B-scans; Bottom image: Incoherent summation of 20 B-scans with speckle decorrelated by angular compounding.
Talbot Bands in OCT
A significant topic of research in OCT at Kent has been to modify the spectrometer in a spectral domain OCT system to create an OCT system which has a longer (or better optimised) depth range. This isn't always so relevant when imaging thick tissue - where tissue scattering is the primary limitation on penetration depth - but it could be particularly advantageous for some conservation applications, where the object of interest is thick but relatively transparent.
During my PhD I performed some experimental validations of a theory that had been developed Adrian Podoleanu and Daniel Woods; reference 2 describes the results. A paper by Woods and Podoleanu describes the effect from one perspective (wavetrains), and a slightly different approach is taken in Chapter 6 of my thesis, where I derive the same result in a different way, and present a model where the spectrometer pixel size is taken into account. Work on this topic continued at Kent – see this paper, for example.
Publications from this Project
1. M. Hughes, "Optical Coherence Tomography for Art and Archaeological Conservation: Methods and Applications," PhD Thesis, University of Kent 2010. [PDF].
2. M. Hughes, M. Spring, and A. Podoleanu, "Speckle noise reduction in optical coherence tomography of paint layers," Applied Optics 49, 99 (2010) (Also in Virtual Journal of Biomedical Optics 5). [Pre-print PDF]
3. M. Hughes, D. Woods, and A. Podoleanu, Control of visibility profile in spectral low-coherence interferometry," Electronics Letters 45, 182 (2009). [Pre-print PDF]
4. M. Hughes, and A. Podoleanu, Simplified dynamic focus method for time domain OCT," Electronics Letters 45, 623 (2009). [Pre-print PDF]
5. M. Hughes, D. Jackson, and A. Podoleanu, "A swept source OCT at 1300 nm with angular compounding for art and archaeological conservation", Proc. SPIE 713917 (2009). [Presentation and Proceedings]
6. H. Liang, B. Peric, M. Hughes, A. Podoleanu, M. Spring, and S. Roehrs, "Optical Coherence Tomography in archaeological and conservation science-a new emerging field," Proc. SPIE 713519 (2008). [Proceedings]
An extract from this work was also published in:
7. P. Targowski and M. Iwanicka, Optical Coherence Tomography for structural examination of cultural heritage objects and monitoring of restoration processes – a review" Applied Physics A, Special Issue on "Optical Technologies in Art and Archaeology" DOI: 10.1007/s00339-011-6687-3 (2011)
During this project I also contributed to the following, slightly less related, publications:
8. I. Trifanov, M. Hughes, A. Podoleanu, and R. Rosen "Quasi-simultaneous optical coherence tomography and confocal imaging," Journal of Biomedical Optics 13, 044015 (2008).
9. C. Sinescu, M. Negrutiu, C. Todea, C. Balabuc, L. Filip, R. Rominu, A. Bradu, M. Hughes, and A. Podoleanu, Quality assessment of dental treatments using en-face optical coherence tomography," Journal of Biomedical Optics 13, 054065 (2008).