Aim To investigate in vivo optical coherence tomography (OCT) for imaging of periocular basal cell carcinoma (BCC).
Methods Consecutive patients with periocular BCC were prospectively investigated with VivoSight OCT imaging prior to surgical excision. Histology sections were compared with OCT images with regard to lesion measurements (x, y and z dimensions) and histological features.
Results A total of 15 patients with biopsy proven BCC were recruited. The OCT horizontal margins correlated positively with histology (r=0.8 and 0.66, x and y axes) and could be identified in 3/15 (x axis) and 6/15 (y axis) cases. The vertical margin correlation was r=0.43 and BCC depth could be measured in 9/15 cases. The following histological features of BCC could be identified on OCT images: (1) lobular pattern (100%); (2) dilated blood vessels (80%); (3) reflective margins of tumour lobules (100%); and (4) epidermal thinning overlying BCC lobules (100%).
Conclusions This study indicated a strong positive correlation between the margins of periocular BCCs measured using in vivo OCT and histology, and a weak positive correlation with depth of invasion. VivoSight OCT produced high resolution images of BCC morphology. The limitations in horizontal margin measurements could potentially be overcome by design modification of the scanning probe.
- Eye Lids
- Treatment Surgery
Statistics from Altmetric.com
Optical coherence tomography (OCT) is a non-invasive imaging technique described in the early 1990s1 and now routinely used in ophthalmology for in vivo imaging of the eye.2 Recently, several groups have explored OCT as a diagnostic tool in dermatology.3–5
OCT is based on the principles of Michelson interferometry and light reflection.4 The signals generated by the interference of the infrared light reflected from the tissue are compared with a reference beam and produce cross-sectional images, visualising the microstructure of the skin with the classical layered architecture.3 ,6 ,7
Basal cell carcinoma (BCC) is the commonest cutaneous malignancy in Caucasians, accounting for 80%–90% of all non-melanoma skin cancers of the eyelid.8 ,9 It is characterised by slow growth, local extensive tissue destruction with potentially disfiguring effects and very low metastatic potential of <0.1%.10–12
The treatment goals for periocular BCC include the complete eradication of the tumour, the maintenance of the eye integrity and a good cosmetic result; accurate assessment of tumour size is particularly important in the head and neck region where surgical excision carries a high cosmetic impact.
Diagnostic biopsies represent the gold standard for BCC diagnosis prior to surgical excision. However, punch biopsy carries disadvantages such as local invasiveness, trauma and inflammation, without providing an estimation of tumour size.
With regard to treatment options, Mohs micrographic surgery offers high cure rates with maximum tissue sparing; however, it requires more training, time and cost compared with traditional surgery.13 ,14Comparison studies of 20 MHz high frequency ultrasound (HFUS) versus OCT have demonstrated an overestimation of BCC thickness in both techniques, with narrower limits of agreement with OCT than HFUS.15
Whereas in vitro and in vivo OCT has been extensively investigated for the diagnosis of BCC in other parts of the body, there is limited evidence for the value of OCT imaging for periocular BCC.15 ,16 A previous paper from our institution described three different en face OCT devices for in vitro imaging of periocular BCC.17
The aim of this study was to investigate a novel Fourier domain OCT for in vivo imaging of periocular BCCs as a diagnostic tool to identify tumour margins and to describe histological correlates.
Patients and methods
This investigation was a prospective pilot study of OCT imaging for periocular BCC. This study was approved by the institutional Research Department (Maidstone and Tunbridge Wells NHS Trust) and the National Ethics Committee (12/NW/0478). The study was compliant with the Declaration of Helsinki and all participants signed an informed consent.
Inclusion criteria and outcome measures
The study included patients between 18 and 95 years of age, newly diagnosed with biopsy proven periocular nodular BCC, undergoing surgical excision and willing to participate at all reviews. A history of recurrent BCC lesions, other carcinomatous lesions of the periocular region and previous periocular surgery were considered exclusion criteria.
This study had two outcome measures. The first outcome measure consisted of the following measurements of nodular BCC lesions: (1) during clinical examination; (2) macroscopic BCC specimen following excision; (3) microscopic histological examination; and (4) OCT images.
The second outcome measure consisted of the assessment of VivoSight OCT in identifying features of BCCs such as lobular pattern, dilated blood vessels and highly reflective margins, as described by Khandwala et al in a previous in vitro OCT study of periocular BCC.17
Newly referred patients with eyelid lesions suggestive of primary BCC were examined and underwent a punch biopsy under local anaesthesia in the Oculoplastic clinic at Maidstone and Tunbridge Wells NHS Trust. All patients were reviewed with the biopsy results 4 weeks later and cases with biopsy proven nodular BCC were listed for a two-stage standard surgical excision and reconstruction under local anaesthetic. During the clinic visit, patients received information about the study. OCT imaging was performed on the day of the surgical excision whereas colour photographs were taken before biopsy, before surgical excision and at every clinic visit. The second stage surgical reconstruction took place 1 week later and consisted of a day case procedure under local anaesthetic. All study patients will remain under regular review for 5 years following BCC excision.
The VivoSight OCT scanner used in this study was a Swept Source Fourier Domain OCT (Michelson Diagnostics, UK) with clinical CE and FDA approval. This MultiBeam OCT system with <7.5 µm lateral resolution and <9 µm vertical resolution in tissue is capable of achieving between 1.2 and 1.8 mm imaging depth and a 6×6 mm lateral field of view. Each OCT ‘bread-slice’ scan consisted of 60 B-scans (in the x–y plane) with 4 µm pixels, separated by 100 µm in the y direction.
The system conformed to laser safety Class 1 and was therefore eye safe. During patient imaging, the scanning probe was hand held and a plastic ring spacer (6 mm in diameter) was mounted on the tip of the probe in order to maintain a constant working distance between skin surface and the lens. The probe was positioned on the area of interest with the aid of a laser aiming beam (wavelength 645–665 nm), no gel was required and all scans were performed by the same clinician (LP). The scanning probe was coupled to a digital interface where the quality of the image and the penetration of the OCT signal could be assessed before starting the image acquisition. Prior to scanning the BCC lesion, the corresponding location of the patient's fellow eye was scanned for control reference purposes.
The probe was held horizontally to the lid margin when scanning lesions on the lower or upper lid margins so that the OCT sections obtained were perpendicular to the lid margin (90° orientation, y axis). The scanning probe was held vertically when imaging lesions located in the medial or lateral canthus and the OCT sections were parallel to the lid margin (180° orientation, y axis) (figure 1).
In order to assess OCT imaging in measuring BCC lesions, three dimensions were assessed: horizontal dimensions, length and width (x and y, respectively), and the vertical dimension or depth of the BCC (z axis). The quantitative evaluation of tumour margins was blinded to histopathology measurements.
Following surgical excision of BCCs with predetermined 2 mm margins of normal skin, the four margins of each specimen were marked using 6/0 silk sutures of different length for correct orientation during histological assessment. The excised specimens were placed in 10% neutral buffered formalin, embedded in paraffin and stained with H&E. Each block was cut into interrupted vertical sections similar to a ‘bread loaf’ for histological examination. The histology cuts were aligned with the OCT scan orientation (figure 1).
The histopathology report included the macroscopic, microscopic and clear margins measurements and the BCC type. The size of BCC lesions was obtained by subtracting the clear margins from the macroscopic specimen measurement.
All data organisation was carried out in Microsoft Office Excel 2010, and statistical analysis was carried out with SPSS 16.0 for Windows (SPSS Inc., Chicago, Illinois, USA).
Mean value, range and SD were calculated for all samples with normal distribution (Kilmogorov–Smirnov test) and non-normal distribution. Pearson correlation coefficients were obtained to evaluate the relation between OCT measurements and histological measurements.
The clinical and demographic data of the study patients are presented in table 1. A total of 15 of 16 consecutive patients with primary BCC of the eyelids were included in the study. There were three women (20%) and 12 men (80%), with an average age of 74 years (range 49–91 years). The most common BCC location was the medial canthus (six patients, 40%), followed by the lower lid (five patients, 33.3%), the lateral canthus (three patients, 20%) and the upper lid (one patient, 6.6%). One patient was found to have solar keratosis with no evidence of BCC and was excluded from the study.
All patients with BCC in the medial canthus region opted for ‘laissez faire’ technique; that is, healing by secondary intention without surgical reconstruction. The remaining patients underwent a second stage surgical reconstruction of the eyelid defect. One specimen from a lateral canthus BCC was reported as incompletely excised and the patient underwent a further margin excision before proceeding to surgical reconstruction of the eyelid defect.
BCC measurements and characteristic features
In vivo OCT imaging was non-invasive, painless and the total contact time for each patient was less than 7 min. Training for the in vivo OCT involved one laboratory based session and two practical training sessions. The image acquisition was straightforward and the quality of the scan, image resolution and scanning depth was shown on the screen prior to the image acquisition.
Four measurements of BCC were recorded: (1) macroscopic measurement during clinic assessment; (2) macroscopic specimen measurement following BCC excision; (3) microscopic measurement from histology sections; and (4) BCC lesion measurement from OCT images. A summary of histological and OCT measurements is shown in table 2.
OCT imaging provided two horizontal measurements (x and y) and one vertical measurement (z). The horizontal measurements were obtained in 3/15 and 6/15 patients, x and y axes, respectively. The vertical measurement of BCC depth was obtained in 9/15 patients. When correlating morphological measurements of BCC lesions, the following significant Pearson's correlations were detected (figure 2, table 2): (1) the measurement of the lateral margins along the x axis showed a positive correlation (r=0.8) between histology and OCT images and (2) the measurement of the lateral margins along the y axis showed a positive correlation (r=0.66) between histology and OCT images. A moderate relation was found between the histological and OCT measurements of the deep margin, z axis, r=0.43.
OCT images obtained with VivoSight OCT were assessed with regard to previously described architectural features of BCC lesions in Khandwala et al17 such as (1) lobular pattern of abnormal architecture; (2) dilated blood vessels in the upper dermis; and (3) highly reflective margins of tumour lobules (figure 3). The above features were observed in 100% (15/15), 80% (12/15) and 100% specimens, respectively. An additional feature observed in 100% samples of BCC consisted in thinning of the epidermis overlying the lobular structures.
This prospective study demonstrated that in vivo OCT imaging is a valuable non-invasive diagnostic technique with potential clinical application in the preoperative assessment of tumour margin measurements and histological correlation of BCCs of the eyelids.
Several authors have investigated the role of non-invasive diagnostic techniques such as ultrasound and OCT for the estimation of tumour margins in BCC. Although the correlation between thickness measurement obtained with HFUS and histology was found satisfactory,18 ,19 further comparison studies showed the superiority of OCT versus HFUS.15–20 In contrast to HFUS, OCT offers a higher resolution and better morphological representation of BCC lesions particularly for smaller BCC (<1 mm tumour thickness).20–22 A recent study by Hinz et al compared preoperative BCC vertical tumour thickness with OCT and HFUS versus the gold standard histopathology. Hinz et al investigated a total of 10 BCCs, six superficial and four solid BCC lesions located mostly in easily accessible sites such as chest, forehead, leg and arm; one lesion was located on the nose. The study showed that BCC vertical tumour thickness measured by OCT correlated more strongly with histology (0.83) than HFUS (0.59).16 The results reported by Hinz et al were based on vertical tumour thickness, whereas horizontal tumour margins were not described.
By contrast, we found a similar correlation between OCT and histology with regard to lateral margin measurements (0.8 and 0.6); the relation was weaker for vertical margin measurements (0.43). Additionally, we could not obtain OCT measurements for all margins in all study subjects. These results need to be considered in light of the technical challenges of scanning a difficult anatomical area such as the orbital region. The most common BCC location in our study was represented by the medial canthus region followed by the lower lid and the upper lid margin. The medial canthus is formed by rigid walls of the maxillary process of the frontal bone, the frontal process of the maxilla and the lacrimal crest of the frontal process of maxilla. The medial palpebral ligament may be prominent in people with orbital fat atrophy and may form a further ridge between the upper and lower part of the medial canthus.
Overall, the medial canthus can be approximated to a concave surface with increasing depth proportional to the height of the nasal bridge and the degree of orbital fat atrophy. By contrast, lesions located on the lid margin were difficult to scan due to the convexity of the skin surface at this site, the mobility of the eyelid and the nearby eyelashes interfering with the OCT signal. In addition to the difficult anatomy of the orbital region, the use of the ring spacer fixed on the scanning probe with an external diameter of 6 mm presented a limitation to the maximum area that could be examined with each scan.
Lesions with lateral dimensions over 6 mm could not be covered fully by a single scan. Moreover, the maximum imaging depth of an OCT scan with the current technology is limited to 1.5–2.0 mm.
The value of OCT imaging for morphological characteristics of periocular BCC lesions was previously described in an in vitro study using en face OCT images.17 en face OCT shows the cross-section of the lesion identifying the lateral margins at different depths with great similarity to Mohs sections. However, en face OCT offers poor comparability with standard histological sections obtained with the ‘bread-loaf’ technique. By contrast, for the purpose of histological correlation, B-scan OCT images can be acquired with the same orientation of postexcision histological sections, allowing the clinician to carry out a correlation study between OCT images, that is, optical biopsy, and microscopy of the excised specimen, that is, histological biopsy.
To the best of our knowledge, this is the first paper investigating in vivo OCT imaging for BCC of the periocular region. The results of this investigation showed a positive correlation between in vivo OCT imaging and histology in BCC lateral margin measurements (r=0.6, r=0.8) and a weakly positive correlation (r=0.43) for vertical margin measurements.
With regard to the morphological features of BCC identified with in vivo OCT imaging, the VivoSight device allowed identification of BCC features with a higher sensitivity compared with previous OCT systems.17
The strengths of this study include its prospective design, the in vivo OCT imaging of BCCs, the comparison with excised specimen rather than biopsy specimen and the alignment of OCT images with histology sections.
The limitations of the study include the small number of patients, the incomplete number of dimensional observations and the homogeneity of the BCC subtype with 100% nodular BCC. Whereas the study was designed as a pilot investigation, the incomplete number of dimensional observations resulted from the difficult anatomical location and the limitations imposed by the ring spacer on the hand-held probe. This issue could be resolved with an adaptation of the probe design in order to allow an easier access and manoeuvrability in anatomically challenging regions such as the periorbital area.
As the present study only investigated nodular BCCs, it is impossible to draw conclusions on the use of OCT imaging for assessing margins on different BCC subtypes such as superficial, multifocal and infiltrating BCC. Overall, OCT appears to represent a promising non-invasive tool for in vivo preoperative diagnosis of periocular BCC. The question of whether in vivo OCT optical biopsy may potentially replace invasive diagnostic biopsies is still the subject of debate. Larger prospective studies assessing the sensitivity and specificity of this system for the assessment of BCCs are necessary.
We would like to thank Daniel Woods of Michelson Diagnostics Ltd for technical support and advice during the study.
Contributors All authors fulfil the criteria of authorship based on the following contributions: LP: study conception, data collection and analysis, drafting of article, final approval. HBS: data interpretation, critical revision of the article, final approval. JBS and AD: data analysis, critical revision, final approval. AM: study design, article revision, final approval. MK: study design, critical revision, final approval, guarantor of the study.
Competing interests Adam Meeckings is employed by Michelson Diagnostics as senior clinical scientist; he offered technical support during data collection and assistance with image analysis during the study.
Patient consent All patients signed an informed consent.
Ethics approval National Ethics Committee approval and Institutional Research & Development approval.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement We do not have any additional unpublished data from the study.
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.