Optical Coherence Tomography: its role in the non-invasive structural examination and conservation of cultural heritage objects—a review
- First Online:
- Cite this article as:
- Targowski, P. & Iwanicka, M. Appl. Phys. A (2012) 106: 265. doi:10.1007/s00339-011-6687-3
- 1.6k Downloads
A brief introduction to Optical Coherence Tomography (OCT) is presented, stressing the origin of the tomographic signal and the detection methods defining various modalities of the technique. The parameters of the tomographs, such as axial and lateral resolution, wavelength and intensity of the probing light, imaging range, time of examination, and sensitivity are then defined, and a paradigm for interpreting the OCT tomograms provided. The second part of the article comprises a review of the utilisation of OCT for structural examination of artworks, illustrated with some representative results. Applications to the structural imaging of semi-transparent subsurface layers such as varnishes and glazes, of underdrawings and of reverse painting on glass, are described first, and then applications in the examination of the structure and state of preservation of historic glass, jade, glazed porcelain and faience are discussed. Finally, the use of OCT combined with LIBS analysis and laser ablation of surface layers is presented.
1 Fundamentals of the OCT technique
Optical Coherence Tomography (OCT) is a white light interferometric technique providing high resolution cross-sectional views (tomograms) of objects which moderately absorb and scatter the probing light. Since it utilises light of low intensity, it is harmless to all known types of artworks. The examination is non-contact, fast, and does not require any preparation of the object examined. The resultant image has a convenient form, similar to the very popular photomicrographs of cross sections of samples collected from the objects. The technique is especially well suited to inspection of the internal structure of stratified objects because the in-depth (axial) resolution falls into the range of 1 to 10 μm, even though the lateral resolution is significantly lower, usually from 10 to 30 μm. The major constraint arises from the limited transparency of the strata of the object to the light used for examination. Nevertheless, the variety of objects to which the technique is being applied is still gradually increasing.
The OCT technique was invented in the mid-1990s by Huang et al. , and since then has become a well established diagnostic method in medicine , especially in ophthalmology. It is used mostly for imaging of human retinas in vivo through the iris of the eye, but also for examination of the geometry and diseases of the anterior chamber. Among other medical applications, utilisation in dermatology, oncology, gastrology, and gynaecology are especially important. At present, various commercial instruments dedicated to medical diagnostics are available at prices ranging from 50,000 to 80,000 Euros.
1.1 Principle of operation
Since the envelope S(k) of the interferometric signal is of arbitrary shape and finite range, the proper window, usually of Gaussian shape, must be applied to data before Fourier transformation. The resultant data represent a single A-scan, as before, but is obtained up to hundreds of times faster than with TdOCT. Additionally, due to the multidetector advantage, the SdOCT technique is significantly more sensitive than TdOCT .
Sometimes different intensity scales are used: simple monochrome (see Fig. 4 and Fig. 8f in Sect. 2, where the strongly scattering structures are registered in white), or sometimes more readable reverse grey scale (see Fig. 5 in Sect. 2 and, e.g., Fig. 4 in ) with non-scattering regions shown in white and strongly scattering ones in black.
It must be emphasised that various instrumental solutions have specific advantages and disadvantages. There is therefore no universal answer to the question as to the best OCT system. The choice should be made in the context of performance parameters. The information provided in the following chapters is intended to serve as a guide and encouragement to further reading on selection of the most relevant system in any particular case.
1.2 Parameters of the OCT instruments
As may be noticed from Fig. 1b, the axial resolution δz is lower than available from phase-sensitive methods, in which it is a fraction of the wavelength of the monochromatic light source, but here the position of the structure within the object examined is retrieved unequivocally, without the disadvantage of phase ambiguity. It is also clear from the formulas of (3) that, for light sources of given bandwidths (in nm), the resolution decreases as the central wavelength increases. This is unfortunate, since the transparency of many media increases with the wavelength of examination. The bandwidth of light sources used in OCT varies from 20 to 200 nm in the spectral range of 700 to 1500 nm, which leads to axial resolution of 2.0 to 20 μm (Fig. 3b) in air. Only some Full-Field OCT systems utilising visible light have axial resolution better than 1 μm.
It is obvious that neither resolution depends on the modality of the OCT system. This is not the case for other important parameters: central wavelength, imaging range, time of examination, and sensitivity.
The central wavelength of the radiation utilised determines the choice of available detectors: for instruments working at wavelengths below 1 μm, both high resolution CCD/CMOS cameras and photodiodes are easily available. Therefore, any of the TdOCT, SdOCT and SSOCT systems may be employed. For longer wavelengths—up to 1.7 μm—line scan InGaAs cameras (Goodrich ISR Systems, Princeton, USA) have lately become available and made it possible to develop SdOCT systems for this spectral range, also. At present, however, this longer-wavelength range is explored mostly by means of swept source (SSOCT) systems. In order to ensure high sensitivity of the instrument, the light source must also simultaneously have high spatial coherence. Most convenient are broadband single-mode fibre sources, but these are available only for wavelengths above 700 nm.
The time of examination is an important factor, especially in diagnostic medicine, which—if short—reduces motion artefacts and increases patient comfort. In the case of examination of artworks, it is also very convenient to keep it short, since this helps in quickly selecting interesting areas for examination, which makes working with OCT highly effective, and renders possible the utilisation of OCT for real-time control of conservation treatments.
The sensitivity of OCT tomographs is defined by the maximum attenuation of the probing light within the examined object which still leaves it detectable. Here, spectral domain OCT offers a significant advantage over the single-detector methods. This parameter is especially important in the investigation of samples exhibiting some degree of IR absorbance, such as glazes and other paint layers.
The power of the probing beam must be considered carefully to ensure that there is no possibility of damage to the object under examination. Generally, very low intensities of light are used in OCT: due to the high sensitivity of the detection system, high intensities are also impractical because they lead to image artefacts caused by saturation of the detector (vertical smears in Fig. 6). Usually the radiation power at the surface of the object is below 2 mW. In such a case, for a typical SdOCT system with an acquisition time of 30 μs for an A-scan and a lateral resolution of 20 μm, the fluence of the infrared radiation may be estimated to be 10 mJ/cm2, far below the damage threshold. It is worthwhile underlining the fact that such intensities are also within the ANSI safety limit for examination of the human retina .
Typical parameters of various OCT systems
TdOCT (translated mirror)
SdOCT (2048 CCD spectrograph)
SSOCT (FDML laser)
Available central Wavelength
25 000 A-scans/s
300 000 A-scans/s
Major features of main OCT modalities
Unlimited depth of imaging
Fast acquisition (possible 3D imaging)
Very fast acquisition
Simple basic design
Adjustable resolution and depth of imaging High S/N
Limited depth of maging
Light sources still under development
Moving parts in delay line
Sensitive to specular reflections
Sensitive to specular reflections
Artefacts (side lobes) for non-Gaussian source difficult to remove
Wavelength of operation limited by CCD camera
1.3 OCT systems used for examination of cultural heritage objects
Many OCT systems designed for medical diagnostics may be used for examination of art objects. The only significant limitation originates in the optics of the object arm: instruments for retinal examination have to be excluded due to the specific optics designed for observation through the cornea and the lens of the eye, so that these systems can produce sharp images of the retina, but are unsuitable for other applications. Other current medical, as well as general-purpose commercial product, systems may be used without significant modifications.
When choosing the OCT system for examination of art, its flexibility in terms of coping with a variety of such objects should be taken into account: the art work may be of various sizes or thickness, and sometimes may be examined only in specific (e.g. vertical or horizontal) position. Very often the examination must be performed in situ. Therefore portable instruments are preferred.
Since a large range of practical solutions are usually described in detail in original papers, the authors of this review feel released from the obligation to repeat this information here. Taking note of our overview of the important parameters given above, the reader is directed to these publications. Among those devoted to examination of cultural heritage objects, the most comprehensive description of TdOCT instruments may be found in [13–17], of SdOCT systems in [7, 18], and of SSOCT in [17, 19–21]. Solutions for Full-Field OCT with the Mireau interferometer are discussed in [22–24], and with the Linnik configuration of a Michelson interferometer in . In addition, a polarisation-sensitive TdOCT tomography, in an application to the examination of parchment, is described in .
In addition to the general descriptions referred to above, some specific issues are addressed: in paper  by Hughes et al., an interesting method for the reduction of speckle noise is presented. An important matter—beyond the scope of this paper, however—is the continuing search for new methods of processing and presentation of data. Two papers by Sylwestrzak et al. fall into this category: in  new formats of presentation of 3D data are described, and in  the modern technique of ultra-fast massively parallel processing of OCT data is presented.
2 Applications to examination of the structure of artwork
The first internationally available reports on the use of the OCT technique for artwork analysis are from 2004 by Yang et al. , Targowski et al. , and Liang et al. . In these early papers, the major fields of interest were already recognised: examination of varnish and glaze layers of paintings and semi-transparent solids such as jades and glazed ceramics. Additionally Yang et al. pointed out the application of OCT for assessment of possible forgeries of jade items. Up to the present (March 2011), over 60 papers  have been published on the OCT examination of artworks.
2.1 Examination of the structure and state of preservation of paintings
This application is straightforward, and from its very beginning was considered an alternative to sampling. Arecchi et al. [15, 32] have proved by direct comparison that the OCT tomogram of the varnish and glaze layer is equivalent to a cross-sectional photomicrograph of the sample taken in the same place.
Liang et al.  first pointed out what is probably the most important advantage of the OCT sectioning over sampling collection: the number of samples and the location of their collection are strictly limited by conservation ethics mostly to areas of existing damage. Since the OCT technique is non-invasive, the examination may be repeated as many times as necessary in any desirable place, so that results obtained with this technique may be considered more representative.
A major concern at the initial stages of development of this technique is centred on its applicability to investigation of the paint layer, due to the limited transparency of such strata to the radiation used. This limitation remains valid, but increasing instrumental sensitivity, together with improved understanding of the images obtained, has rendered it less significant. The only systematic studies on the applicability of OCT to the imaging of paint layers composed of various pigments were performed by Szkulmowska et al. . They analysed 47 different commercially available oil paints with two OCT systems working at extreme central wavelengths: 0.83 and 1.55 μm. As expected, better results were obtained with the longer wavelength observation: for that case, 16 of the 47 pigments qualified as having high transparency, while a further 6 were of medium transparency. For 0.83 μm observation, 9 of the 47 fell into the first category, a further 7 into the second. However, as can be seen from (3), increasing the wavelength of light used for examination leads to rapid deterioration of the axial resolution. This trade-off must therefore always be taken into account when new experiments are under consideration. An important account of this subject has also been presented by Liang et al. .
When a varnish or similar layer is under investigation, the choice is rather simple: short wavelengths are preferred to ensure high resolution. Superb axial (1.5 μm) and lateral (1.0 μm) resolutions were reported by Latour et al. with the use of red (visible) light and a Full-Field TdOCT system with Mireau objective for examination of pigmented lacquers on slabs  and of the coatings on an historic violin . However, a disadvantage of microscope system used lies in the limited lateral field of view. Examination of the stratigraphy of the varnish layers with OCT was successfully executed by Targowski et al.  for analysis of the locations of paint layers involved in inscriptions in relation to other strata: two paintings were analysed at the request of the conservation studio. In the first case, it proved possible to resolve some doubts concerning the history of the painting. In the second case, evidence of forgery of the signature was uncovered. Processes occurring during drying of the varnish layer were investigated by Liang et al. . They pointed out that the varnish Paraloid B-72, when drying, begins to follow the roughness of the substrate. This may be used as a non-contact method of determination of the stage of the drying process. The ability of the varnish layer to either reproduce or level the roughness of the surface of underlying paint (depending on molecular weight of the resin) was further investigated by Targowski et al.  and found to be in accordance with the pre-OCT conclusions derived much earlier by de la Rie .
2.2 Examination of the structure and state of preservation of historic glass
The suitability of OCT for imaging corrosion layers of stained glass was confirmed by Kunicki-Goldfinger et al.  by comparison with SEM (Scanning Electron Microscopy)-BSE (Back Scattered Electrons) images of a cross section of the same sample. The SEM-BSE technique is commonly used for detection of the build-up of the silica-rich gel layer at the glass surface during the corrosion process. However, in order to employ the SEM-BSE method, it is necessary to prepare a polished cross section of the sample. Therefore, this method, although not destructive of the sample, may be used in limited cases only. OCT offers here an interesting alternative, since it is truly non-invasive and may be used at any point of the glass sheet, in principle even without dismantling the window. Similar results, but obtained with a 1300 nm system, have been reported by Hughes .
Some of the ancient glass artefacts from the collections of the British Museum have been examined by Liang et al. [17, 34, 37]. These studies confirmed the ability of OCT to reveal details of the internal structure of the glasses up to 0.5 mm under the surface. The usefulness of OCT for fast screening of museum collections may be important also in diagnosing crizzling of glass (see Fig. 1 in ). This phenomenon is a deficiency of glass caused by imbalance in the ingredients of the batch, and can be recognised as a network of fine cracks, very often developing with time and possibly leading eventually to the total destruction of the object.
An OCT investigation  of such samples revealed that the crizzled layer is of uniform, though not the same, thickness on both sides of the same sheet.
2.3 Examination of the structure and state of preservation of other semi-transparent objects
Among other objects, artefacts made of jade have been investigated, first by Yang et al. , later by Liang et al. , and recently by Chang et al. . Yang showed, by comparing a sample of ancient jade before and after artificial ageing (by burning with a gas torch), that it is in principle possible to distinguish between them using OCT. The artificial ageing induces some small-scale alterations to the structure of the sample leaving large-scale features unchanged, making the overall structure more heterogeneous. The original ancient pieces have a fairly uniform structure as a result of the simultaneous formation of small-scale scattering centres (these are responsible for the whitening of jade) and consequent diminishing of the large-scale features. Such observation may help in recognition of possible frauds. This idea of using OCT for assessing the quality and authenticity of jade was developed further by Chang et al. . However, in contrast with the work previously referred to, these authors provided a quantitative texture analysis method based on six well-defined parameters which may be derived from the OCT signal and further combined into a 6-D texture vector. The length of this vector, together with the reflecting properties of the jade surface, may be used for differentiation among various types of jades.
Another group of objects examined with OCT from the very beginning are those made of porcelain and faience [17, 37, 41]. The most comprehensive study was made by Yang et al. , examining various items of 10th–14th century Chinese glazed ceramic shards. Major attention was paid to the ability of OCT to recognise different phases and phase boundaries within the glaze layer. The information acquired by the OCT examination may be used for identification of the manufacturer of the porcelain, and detection of some glaze flaws such as impurities and internal cracks.
2.4 OCT as a supporting tool for LIBS stratigraphy
Laser Induced Breakdown Spectroscopy (LIBS, or sometimes LIPS) is a micro-invasive analytic technique used in materials science mostly for examining the elemental composition of metal alloys. This is determined by analysis of the spectrum of atomic fluorescence emitted from the small plasma plume generated by a short, high-power laser pulse focussed on or into the material examined. Since the diameter of the crater formed during the LIBS analysis does not usually exceed 100 μm, the invasiveness of this technique is limited, and it is often used for examination of objects of cultural heritage. A comprehensive review of this application has been presented by Fotakis et al. . A very important variety of this application is LIBS stratigraphy, where the ablation crater is hollowed out gradually by a sequence of laser pulses. Since the fluorescence spectrum is registered after each laser pulse, the concentrations of elements detected may be determined as a function of the laser pulse number. This gives some indication of the depth dispersion of elements contained in the object examined. A significant drawback of this approach lies in the lack of information on the exact depths of the crater after each pulse. This is especially important, since the ablation rate may be significantly different for each layer of the object. The solution lies in monitoring the depth of the ablation crater after every laser pulse. To the best of our knowledge, the first time the idea of integrating white light interferometry with LIBS for this purpose appeared in its inclusion in the patent granted to Detalle et al.  and described in the paper by Dufour et al. . In 2004, Papazoglou et al.  presented a LIBS apparatus comprising a white light interferometer, similar to a Full-Field TdOCT system, for use in examination of cultural heritage objects. This instrumental solution was illustrated by some results of employing it in elemental depth profiling of a daguerreotype. Due to the high axial resolution of their interferometer, the authors were able to determine the thickness of the gold-enriched covering layer to be 1.7 μm. Another study on the elemental composition of a metal object was performed by Amaral et al. . Using a commercial TdOCT system (6-μm axial and lateral resolutions) and a Calibration-Free LIBS method, they were able to trace the percentage concentrations of various elements (mostly metals) in historic Brazilian coins for 20-μm depth steps. It is worth noting that, since the structure of the coins was rather homogeneous, it was not necessary to profile the crater after each laser pulse—that was measured after completion of the sequence of pulses, and depth data for each pulse was interpolated. Preliminary results on using a LIBS-OCT system to determine the stratigraphy of pigmented layers of paintings were presented by Kwiatkowska et al. both for a model  and for a historic painting . In both cases, it was confirmed that it is possible to re-calibrate the depth scale of LIBS concentration profiles with the aid of OCT. Since the ablation rate varied from layer to layer, the dependence of crater depth on pulse number was not linear. Furthermore, changes in the crater ablation rate, determined from the depth-pulse relationship, are very helpful in recognising the boundaries between consecutive strata in the paint layer. Additionally, an OCT tomogram registered before the LIBS experiment, exactly in the same place, permits conscious choice of the best location to examine, helping with avoidance of ambiguous areas such as those of previous conservation treatments.
2.5 Prospective application of OCT as a supporting tool for laser ablation cleaning
This specific, and still only prospective, application has been developed exclusively by our group since 2007 in close cooperation with the Institute of Optoelectronics of The Military University in Warsaw, Poland. Attempts to remove varnish and similar coatings by means of laser ablation have their own long history, beyond the scope of this contribution. Briefly: varnish layers should be removed if they suffer from discolouration or loss of transparency due to prolonged exposure to light and/or external pollutants. Although, traditionally, varnishes are either removed mechanically with a scalpel, or chemically with solvent mixtures, or by using a combination of both methods, in cases of very firm or hardly soluble layers this approach is extremely difficult, and endangers the paint layer underneath. In such cases, laser ablation may be considered an interesting possible alternative. However, possible damage to the underlying strata is a major concern, since they are at least similarly vulnerable to laser ablation. In contrast with the well-established stone-cleaning treatment, where the process usually terminates automatically when, after removal of the strongly absorbing contamination layer, the white well-reflecting stone surface is reached, the laser-ablation process of varnish removal is not self-terminated. It is therefore necessary to introduce a specific controlling mechanism into this procedure. About ten years ago, an automated workstation was developed  in which the LIBS technique was employed to control the termination of varnish removal by detecting spectral lines from paint pigments in the ablated plasma. Obviously however, this technique does not allow control of mere thinning of the varnish layer rather than its complete removal.
On the other hand, it seems that OCT may be employed here as an alternative monitoring system, since the varnish is usually transparent enough to be imaged through the whole of its depth. Up to now, experiments have been conducted with three groups of varnishes (dammar, ketone, and acrylic), employing different lasers operating both in the far infrared (λ=2.936 μm) [42, 52] and in the UV (λ=0.266 μm) [53, 54]. Depending on the laser/varnish combination, OCT enabled the distinction of three processes caused by laser radiation: true ablation of the varnish layer, its exfoliation, and its melting. However, the majority of these results, apart from those announced by Góra et al.  (electronic version ), were not obtained in real time. Instead, raw data were collected to the hard disc of a suitable computer, then post-processed in order to obtain the OCT images. In the case of the experiment described in the publications by Góra et al., real-time results were shown, but the image quality was rather poor. To summarise, for successful monitoring of laser ablation by means of OCT, two conditions have to be fulfilled: the system must operate with as high in-depth resolution as possible, and very fast data processing must be employed. As the next stage of development, one may consider monitoring the ablation process by automatic recognition of the thickness of remaining layer of varnish together with appropriate control of the laser pulse frequency.
As may be gleaned from this short review, applications of OCT in conservation–restoration practice fall into two categories: inventory techniques aiding in revealing structural information, and tools assisting in the optimal performance of other analyses and treatments.
As was also shown above, a major limitation arises from the limited transparency of many of the constituents of art objects. However in many such objects, the intended visual effect is built up by the combined perception of many subsurface layers, as in paintings with glazes and varnishes. These layers must thus be at least semi-transparent, and are therefore accessible to examination by OCT.
When OCT is used for structural analysis, it is important to note that, although it is characterised by superb in-depth resolution, it is sometimes difficult to interpret the composition of the layers imaged (Fig. 2b). The most promising possible resolution of this problem seems to be that of combining OCT with other non-invasive techniques such as multispectral imaging, XRF, and Raman analysis, or using it in cases when the general structure of the object is known, but the local thickness and condition of layers need to be revealed in as many locations as possible.
Another direction for the future development of the technique, leading directly to practical applications, is, in our opinion, further advance in post-processing of OCT images for the automatic recognition of strata in the object under investigation, possibly simultaneously revealing some information about its properties. The rapid growth of medical applications of OCT, especially in terms of image quality and the understanding of its content, leads us to believe that the applications of OCT in the area of cultural heritage will also make further significant progress in the coming years.
Financial support was provided by the Polish Government through research grants for 2009–2011. The research leading to these results was partly funded by the EU Community’s FP7 Research Infrastructures Programme under the CHARISMA Project (Grant Agreement 228330). The publication reflects only the author’s views and the Community is not liable for any use that may be made of the information contained therein. Neither the CHARISMA consortium as a whole, nor a certain participant of the CHARISMA consortium, warrant that the information contained in this document is capable of use, nor that use of the information is free from risk, and accepts no liability for loss or damage suffered by any person using this information. MI gratefully acknowledges support from the VENTURES programme of the Foundation for Polish Science co-financed by the Innovative Economy Operational Programme within the European Regional Development Fund. The authors would like to thank Dr Michalina Góra for valuable discussions, especially regarding parameters of the OCT systems, and Dr Robert Dale for critical reading of the manuscript.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.