Article Text

Download PDFPDF

Imaging of periocular basal cell carcinoma using en face optical coherence tomography: a pilot study
  1. M Khandwala1,
  2. B R Penmetsa2,
  3. S Dey1,
  4. J B Schofield1,
  5. C A Jones1,
  6. A Podoleanu2
  1. 1Maidstone and Tunbridge Wells NHS Trust, Maidstone and Tunbridge Wells, Kent, UK
  2. 2Applied Optics Group, University of Kent, Canterbury, Kent, UK
  1. Correspondence to Mona Khandwala, 5 College Close, Lingfield, Surrey RH7 6HG, UK; mona.khandwala{at}gmail.com

Abstract

Aim To use en face optical coherence tomographic (OCT) imaging to identify features of tumour tissue and their correlation with histopathologic findings and to assess the effect of different wavelengths and resolutions of OCT in identifying tumour boundaries and features.

Methods Excision specimens of consecutive biopsy-proven periocular basal cell carcinomas (BCCs) (n=8) were assessed by OCT, performing in vitro cross-section and en face scans of the tissues. Images were collected from three different machines: systems 1 and 2 had a wavelength of 1300 nm, and system 3 had a wavelength of 840 nm. System 2 used high numerical aperture interface optics that determines higher magnification and hence allows higher transversal resolution. All the eight specimens subsequently underwent routine histopathologic examination.

Results Three common features of tumour tissue were observed in all the three systems: (1) lobular pattern of abnormal architecture, (2) dilated blood vessels and (3) high reflective margins. We compared the three systems based on their ability to pick up the three above-mentioned tumour features. In this respect, system 2 had the highest capability in picking up feature 1, followed by systems 1 and 3. In feature 2, similar results were obtained with all the three systems. System 3 was unable to pick up feature 3, whereas systems 1 and 2 performed equally.

Conclusion En face OCT imaging has the potential to identify tumour tissue from healthy tissue. It also showed correlation with corresponding histopathologic findings. Non-contact OCT imaging of the skin is a non-invasive and convenient method and can be useful for demarcating BCCs on the face and eyelids. Future larger studies on in vivo BCCs using en face ultra-high–resolution OCT should provide information on subtyping BCCs.

  • Carcinoma
  • basal cell
  • optical coherence tomography
  • BCC
  • OCT
  • eyelids
  • pathology
  • neoplasia
  • imaging
  • diagnostic tests/investigation
View Full Text

Statistics from Altmetric.com

Basal cell carcinoma (BCC) is the most frequently occurring periocular malignancy, with a rapidly increasing incidence worldwide, because of both increased sun exposure and longevity.1 Surgical excision with predetermined margins,2 or Mohs surgery where available,3–5 is recommended to ensure complete tumour removal.

Several non-invasive diagnostic methods have been reported so far but none have yet proved to be practical. High-frequency ultrasound, terahertz imaging and magnetic resonance all offer limited resolution.6–8 Reflectance confocal microscopy produces high-quality images, but it has the disadvantage of requiring contact with the tissue and having a limited field and depth penetration. Ultrasound provides better penetration but has a lower resolution when compared with confocal microscopy.9

Optical coherence tomography (OCT) is a novel imaging technology10 used to determine the structure of translucent objects. It combines a high axial resolution with good tissue penetration. Although well established for retinal imaging,11 the equipment currently used in ophthalmology clinics does not provide reliable skin images. OCT has been successfully used for high-resolution imaging of structures in other parts of the body12 13 and has been successfully used to identify tumour features and abnormal skin architecture.14 In this paper, we assess the use of en face OCT, a time domain (TD) variant of the OCT technology, to identify and delineate periocular BCCs.

Method

Clinical study

A pilot study to assess the capability of OCT to delineate tumour from normal tissue was performed. Ethical approval for the study was obtained from both the ethics committee of the Faculty of Science, Technology and Medical Studies, University of Kent, and the West Kent Ethics Committee. The research described adhered to the tenets of the Declaration of Helsinki.

In this study, excision biopsies were performed on periocular BCCs by members of the Ophthalmology Department at Maidstone and Tunbridge Wells NHS Trust. Patients attending the Ophthalmology Department at Maidstone Hospital were recruited to the study after informed consent was obtained. All the eight patients recruited had a biopsy-proven BCC within the periorbital region. After routine surgical excision using predetermined tumour margins, the tissue was marked with sutures at superior, medial and lateral margins using 5/0 silk (Ethicon, New Jersey, USA).

Within 3 hours of excision, the samples were transported in formaldehyde to the Applied Optics Group laboratory within the University of Kent for OCT imaging. The technology required for OCT imaging of the skin is at a much earlier stage of development than that for retinal imaging. All the samples were scanned by three different OCT machines. These are experimental prototypes, thereby allowing in vitro examination only. The origin of the samples and the patient details were single masked from the group performing the OCT scans.

For scanning, the samples were fixed onto pieces of polystyrene board to align the longest ends of the samples along the horizontal direction. Images were obtained in vitro using each of the three OCT systems. Multiple en face scans were obtained at steps of 25 μm in depth, and horizontal cross-section scans were obtained by moving the sample in the en face plane (mostly along the vertical direction). On each of the three OCT machines, the scanning process took between 45 to 60 min, including the setup time. All the tissue samples were kept hydrated during the scanning process using drops of formaldehyde. This part of the protocol was agreed with the histopathologist before starting the study, ensuring preservation of the tissue for accurate histopathologic analysis. On completion of the OCT scans, the tissue samples were stored in formaldehyde and processed for routine histological analysis, undertaken by a single histopathologist. After fixing the specimen, it was transversely sectioned and “bread sliced” into several pieces, which were processed and embedded in paraffin wax. Two to 3-μm slices were obtained and stained with haematoxylin and eosin (H&E). The histopathologic sections were, therefore, orientated in the axis of the OCT cross-section scan. The pathologist was masked from the findings of the OCT scans, until routine histology reporting was done. Once histopathologic analysis was completed on all the eight specimens, the OCT team and histopathologist examined all the histopathologic sections to identify a section that correlated with the OCT image for each sample. Where it was not possible to exactly match the histopathologic section to the OCT image, a section was chosen that displayed corresponding morphological features.

OCT instrumentation

OCT is based on a Michelson interferometer,10 where the light emitted by a source is split into an object beam and a reference beam using a beam splitter. The reference beam is reflected back from a reference mirror. Both beams are combined, and the interference pattern is read by an interferometer. Depth scanning is achieved by moving the reference mirror; lateral scanning is performed by moving the scan beam laterally, using a dual scanning head (equipped with two galvanometer scanners) and a lens. En face OCT as used here generates lateral reflectivity profiles, called T-scans, obtained by scanning the beam horizontally over the target.15 Multiple T-scans acquired at 500 Hz are recorded for sequential values of the vertical coordinate to form an en face OCT image (a two-dimensional (2D) map, termed C-scan, or constant depth image). The vertical scanning is obtained by driving the frame galvo scanner in the dual scanning head to move the beam vertically at 2 Hz. By changing the position of the reference mirror, C-scan images at different depths can be obtained, and for sufficient dense sampling of the tissue, 3D volumes can be reconstructed. En face OCT has the advantage that by simple rerouting of the signals driving the two scanners, the system can be switched from C-scan into B-scan regime. A B-scan is obtained by stopping the frame scanner and by moving the reference mirror to explore the object in depth. In this way, B-scans are built as 2D assemblies of T-scans repeated at sequential depths. B-scans represent cross-section slices in the tissue; they are orthogonal on C-scans and represent the traditional OCT view. However, the en face OCT method provides the added advantage of producing images with the orientation of microscopic images (C-scans). Features with en face orientation allow easier identification of tumour margins when compared with B-scan OCT images. All the three systems used here can be operated in either C- or B-scan regime, and their details are specified in table 1.

Table 1

Specifications of the three OCT systems used in the study

Results

Three common features of tumour tissue were observed with all the three systems. These features were (1) a lobular pattern of abnormal architecture, (2) dilated blood vessels in the upper dermis and (3) high reflective margins of tumour lobules. Table 2 shows the detailed results for each system. System 2 showed the best performance in picking up the abnormal architecture, followed by systems 1 and 3. There was no difference between the systems in exhibiting dilated blood vessels. However, system 3 could not identify high reflective margins in any samples, whereas systems 1 and 2 could.

Table 2

Pick-up rates of tumour features for each of the three OCT systems

Figure 1 illustrates the appearance of normal skin in an OCT image in comparison with a histologic feature with H&E stain. All the scans showed a well-defined epidermis. However, the thickness of the epidermis varied across the samples.

Figure 1

(A) Horizontal B-scan OCT image obtained with system 1; an image depth of 1.5 mm was measured in air. The boundary between the epidermis and the dermis is seen as a low-reflectivity band (arrow); the image from the normal eyelid skin. (B) Histologic image showing normal epidermis and dermis. The arrow indicates the dermoepidermal junction (H&E stain, original magnification ×40).

The typical lobular architecture of the BCCs was noted both on the histologic examination and on the OCT images. Details are as shown in figure 2.

Figure 2

(A) C-scan OCT image of the tissue that underwent biopsy, obtained with system 2. Large low-reflectivity lobular structures are seen (white arrow; image size X=Y= 3.5 mm) at a depth of 0.45 mm (measured in air). (B) Corresponding histologic picture showing BCC lobules (H&E stain, original magnification ×20).

Figure 3 shows a lobular pattern of abnormal architecture and high-reflectivity bands. Lobular low-reflectivity areas were seen scattered within the tissue at depths beyond 300 μm (figures 2, 3). Multiple, lobular, low-signal areas clumped together were also seen at depths beyond 250 μm. Both the lobular low-signal areas and the altered architecture of the skin were seen in the upper dermis area. Dilated blood vessels were seen as lower- signal, well-defined, oval structures with regular outlines (figure 4).

Figure 3

(A) C-scan OCT image obtained with systems 1, showing high reflective margins dividing tissue into multiple segments. Image size X=8 mm and Y=9.5 mm at a depth selected approximately 0.45 mm (measured in air). (B) Corresponding histologic finding with an arrow indicating the stromal tissue separating the lobules of the cancerous tissue (H&E stain, original magnification ×20).

Figure 4

(A) C-scan OCT image of the tissue that underwent biopsy, obtained with system 3. Dilated blood vessels are seen as multiple low-reflectivity oval structures (white arrow). Image size X=1.37 mm and Y=1.1 mm at a depth selected approximately 0.325 mm (measured in air). (B) Corresponding histologic image showing vessels, with red blood cells in the lumen, in the superficial dermis overlying a lobule of BCC (H&E stain, original magnification ×20).

Discussion

Previous attempts to produce skin images using OCT have been reported,14 16 but this technology is still under review. There are a lack of systematic studies that prove the efficacy of OCT as a tool for the diagnosis and delineation of BCC, and all the studies undertaken to date have examined B-scan images only. Live en face images provide a transverse view at a desired depth. This allows identification of all four margins at any particular depth of tissue. Combining the interpretation of B- and C-scan OCT images, the tumour margins can be better delineated in all the three dimensions, as achieved by Mohs surgery. In the future, this could be further enhanced using computer imaging to produce a 3D reconstruction of the tumour tissue. To the best of our knowledge, this is the first paper to look at en face OCT imaging in examining BCCs.

In this study, we observed the tumour margins as signal intense lines separating the tumour from the surrounding tissue and also dividing the tissue into different lobules. Tumour tissue from the deeper layers of the skin appeared as low-signal oval and round areas with no clear arrangement. The epidermis was clearly distinguishable from the dermis in samples where skin surface was even and the border was seen as a low-reflectivity band. The OCT images showed areas of high reflectivity corresponding to the tumour lobules on histologic examination. In addition, scattered high-reflectivity areas surrounded by low-reflectivity areas were observed within the abnormal tissue. Similar observations have been made in a previous study. Bechara et al16 reported similar morphologic structures and formations both in histologic examination and OCT in their series of six patients with BCC and naevi. In their study, the OCT images were taken in vivo before lesion excision. Although in vivo imaging has obvious clinical benefits, our study was limited to in vitro imaging because of the prototype setup of the machines available. Strasswimmer et al17 devised a pilot study using polarisation-sensitive OCT to identify features of BCC. Before surgical excision, they scanned different portions of the tumour. A loss of birefringence was noted in the tumour itself, which was more marked in the infiltrative type of BCC. They were unable to correlate exact features of OCT and histologic findings to determine precise lateral tumour extension.

It is recognised that water content and temperature of the body tissue can affect image quality.18 Normal skin temperature is 37°C, and it contains 60% to 70% water. These constants are affected in excised tissue and can modify the optical properties of the skin. Because the samples used in our study were examined in vitro, the temperature difference from normal body conditions might have played a significant part in the modification of tissue optical parameters. Schmitt et al19 have shown that differences exist in the optical properties when measurements are taken at different temperatures. In addition, Bonner et al20 reported that the absorption spectrum of water in the near-infrared region depends on temperature.

Different wavelengths for imaging the BCC samples were selected (1300 and 840 nm), taking into consideration the fact that the absorption coefficient for different layers of the skin is different. The second OCT system with high–numerical aperture interface optics, which enables a high transverse resolution better than 5 μm and an axial resolution of 13.6 μm, was able to image tumours in greater detail. However, higher magnification and better transverse resolution result in a smaller field of view. In our study, the highest tumour detection rate was obtained with system 2 (high numerical aperture) followed by system 1, both operating at 1300 nm. System 3, which had a wavelength of 840 nm, did not show significant findings. This may be because of the reduced penetration depth at 840 nm and a variation of contrast with wavelength.

Previous studies with OCT imaging of BCC showed features similar to those found in our study.15 17 Our study examined BCCs exclusively from the face and eyelids, used two different wavelengths and also performed en face scans. Boundaries demarcating the tumour from the normal tissue and dividing the tumours into different segments were clearly visible in the en face images at a 1300-nm wavelength. The disappearance of this feature at 840 nm could indicate a reflectivity and absorption profile for the tumour tissue. Altered architecture of the skin was seen very prominently in images from system 2, which had a transverse resolution comparable with what can be achieved in confocal microscopy.

En face (C) scans and cross sections (B) were performed on all the eight samples, with all three OCT systems; this provided the advantage of being able to correlate information at a particular depth in the C-scans with that along the depth in B-scans, guiding us in delineating the margins of the tumour in all directions in 3D volume. B-scan OCT images may lead to missed features across the lateral aspect of the tumour.

There are no studies currently available that have reported en face scans on BCC. In traditional B-scan OCT images, margins appeared as high-reflectivity bands separating the tissue into segments. No such findings were demonstrated in previous studies on BCC.17 In this study, altered skin architecture was the most common finding in the majority of the samples, and it had varied presentations.

Given the low contrast of OCT images, the way forward is to combine imaging modalities. Some encouraging results have been reported using fluorescence or polarisation. Thus, simultaneous or sequential imaging in different regimes where the hardware allows accurate superposition of images could be a way forward.

For ideal parameters, the wavelength of 1300 nm seems the most appropriate. Radiation of longer wavelength may have better penetration; however, the water absorption peaks up, and the transversal resolution deteriorates. Radiation of shorter wavelength encounters more scattering, and perhaps, this explains our results here with the third machine at 840 nm being less effective. As regards the optimum regime of operation, TD versus spectral domain (SD), it is largely accepted that SD-OCT offers better sensitivity than TD-OCT. However, SD-OCT acquires images under a fixed focus, and as a consequence, it is incompatible with higher magnification, which requires focus change in synchronism with depth exploration. Higher magnification is also required for improved transversal resolution. Therefore, the en face OCT (a TD-OCT method), which allows focus change, seems better suited to investigations that require higher magnification despite its lower sensitivity in comparison with SD-OCT. Lastly, considering the issue of image size versus transversal resolution, perhaps different probes with different magnifications may overcome the optics limitation that prevents good transversal resolution when under low magnification.

There were two main limitations to our study. The first was the small sample size. It was designed as a pilot study to demonstrate proof of concept, with a view to a larger study in the future depending on the results. The technical and logistical difficulties of transporting the samples to a different site for OCT scanning and then back to the laboratory for histopathologic analysis necessitated a small sample size. The other limitation was the in vitro scanning. Ideally, the study would have looked at in vivo scanning, as this is where the clinical application of OCT lies, but we were limited by the equipment available to us. A further study is planned to compare the results of in vitro versus in vivo, scanning once the technology is available.

In conclusion, en face imaging has the potential to demarcate tumour tissue from healthy tissue. It has also showed correlation with corresponding histopathologic findings, as demonstrated in the figures above. Non-contact OCT imaging of the skin is a very convenient method and may be very useful for imaging BCCs on the face and eyelids.

However, larger studies of BCC in vivo are required for better identification of tumour margins and subtyping tumours.

References

View Abstract

Footnotes

  • Competing interests AP acted as consultant for and has patents with Ophthalmic Technology Inc, Canada, and University of Kent.

  • Ethics approval This study was conducted with the approval of the ethics committee of the Faculty of Science, Technology and Medical Studies, University of Kent, and the West Kent Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.