Original research study A new tool to inform intra-operative - - PDF document

SLIDE 1

Page 1 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure.

Original research study

For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74.

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

Abstract

Introduction The aim of this study was to evaluate the findings of elastic scattering spec- troscopy co-registered with histopa- thology in patients with basal cell carcinoma against normal and some common benign skin disorders. Materials and methods Clinically suspicious head and neck skin lesions were included in this

• study. Those lesions with surround-

ing innocuous skin were interrogated by elastic scattering spectroscopy, co- registered biopsies were taken and examined histopathologically; the re- sults were then compared using a va- riety of statistical techniques. Results Our analysis showed obvious and consistent spectral differences be- tween normal and pathological skin. Discriminating elastic spectral differ- ences were consistently identified be- tween basal cell carcinoma and other skin lesions of similar appearance. Conclusion This preliminary study shows that elastic scattering spectroscopy can distinguish between basal cell cancer, common benign conditions and vari- ants of normal skin. Elastic scattering spectroscopy can also help deter- mine the diagnosis between benign lesions with a high degree of accu- racy suggesting that elastic scattering spectroscopy can add significant

A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy

T Upile1,2*, W Jerjes3, O Johal2, S Lew-Gor2, J Mahil1, H Sudhoff1

• bjectivity to dermatological diagno-

sis and management of even benign conditions.

Introduction

In essence, our desire is for a technol-

• gy that delivers a pathologically ac-

curate diagnosis in real-time in situ without the need of an attendant re- moving the tissue during processing. This is especially important in cos- metically sensitive areas where any tissue volume is precious not just be- cause of its loss but also due to the difficultly of being able to provide a satisfactory surgical repair. This study discusses the elastic spectroscopic analysis of basal cell carcinoma (BCC) of the skin, the com- monest type of malignancy in the

• world. Often, there is diagnostic

doubt which leads to delayed diagno- sis or even less than satisfactory dis- ease management. Unfortunately, a tumour is often thought of in two not three dimensions with regard to treatment, although convenient and usually clinically insignificant, in ar- eas of cosmetic importance or near vital structures, a three-dimensional (3D) resection is vital to achieve the important goals of disease removal and functional preservation. Visual assessment, although im- portant, can be supplemented by more objective technology for the benefit of the patient and to guide proper treatment delivery and its ef-

• ficacy. Although there are many man-

agement pathways and a modest degree of time insensitivity in the treatment of BCC, the methodology can also be applied to other diseases requiring more stringent time sensi- tive management with a goal of rapid and complete definitive surgical treatment. The knowledge of true diagnosis and limit of the disease aids in the de- livery of an adequate surgical margin and an improved prognosis for a pa-

• tient. We examined several areas of

BCCs against normal controls and common benign skin lesions. We show that it is possible to use this op- tical technology to inform a clinician regarding the diagnosis and to guide surgical and photodynamic treat- ment in ‘real-time’1,2. BCC presents as subtle, painless, non-healing ulcers or nodules on the sun-exposed parts of the body. BCC has many clinical variants and each has its own histological pattern. BCC constitutes approximately 80% of all non-melanoma skin cancers. The tumours most often appear in indi- viduals aged 40–60 years; BCC has a male predilection, with a male-to- female ratio of 2:1. BCCs penchant for the head and neck is notable and is related to its primary aetiology— solar exposure. Approximately 75%–86% of pri- mary BCCs are found on the head or neck2–4. The most common location

• n the head is the nasal tip and alae.

Risk is related to the skin type and degree of exposure to sunlight, par- ticularly ultraviolet-B (UV-B) radia-

• tion. The tumours are more frequent

in individuals with a light complex-

• ion. The Fitzpatrick skin type scale,

which ranges from very fair (skin type I) to very dark (skin type VI), categorises cutaneous sensitivity to UV radiation. It is based on the indi- vidual’s tendency to burn and tan and is a good predictor of relative risk among Caucasians. The prevalence of

* Corresponding author Email: mrtsupile@yahoo.co.uk

1 Department of Otorhinolaryngology, University

• f Bochum, Germany

2 Department of Otolaryngology, Brighton and

Sussex University NHS Trust, Brighton, UK

3 Leeds Institute of Molecular Medicine, School

• f Medicine, University of Leeds, Leeds, UK

SLIDE 2

For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74. Page 2 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

BCC increases in areas of higher alti- tude and in areas of lower latitude. The incidence of BCC is rising, partly because

• f

atmospheric changes and increasing popularity of

• sunbathing. The estimated annual

death rate from this tumour is 0.44 per 100,000 persons. However, BCC rarely causes death; its morbidity is associated with uncontrolled ad- vanced disease1–3. The current gold standard of skin lesion excision or even biopsy of sus- picious lesions is under ‘naked eye’ guidance; this is followed by larger more ablative excisions based upon pathological results of the biopsy, which are usually obtained several days/weeks after the biopsy. Occa- sionally frozen section analysis or Mohs surgery is also used, but their use is limited by availability, expense, time consumption and subjectivity in terms of errors of interpretation with lower sensitivity and specificity when compared with paraffin section analysis (which often takes days/ weeks to complete and interpret). Such mediators of pathology directed resection (Mohs surgery) are re- source intensive, expensive and user dependent. Often these lesions arise on areas

• f sun exposure, where tissue is at a

cosmetic premium and where loss would result in unacceptable defor-

• mity. Due to the nature of conventional

excision, many of these lesions require a wide 3D margin of surrounding nor- mal tissue to be also removed to re- duce the chances of leaving residual disease, which may otherwise lie hid- den under any reconstruction or clo-

• sure. Real-time pathological analysis

is a desirable goal to detect and guide

• treatment. It enables tumour removal

with an adequate margin of resection to ensure completeness of disease re- moval and reduction in loco-regional

• recurrence. Ideally, an objective opti-

cal method would help in diagnosis and surgical treatment. This type of perioperative analysis would have many benefits. There are many optical diagnostic technologies available; each has dif- fering characteristic advantages and published sensitivities and specifici- ties which unfortunately are for mu- cosal rather than skin diagnosis. Review of the literature and our aim for an objective, low cost system with a potential for rapid translational use by non-technical clinicians in order to help guide treatment with easily interpretable visual results led us to consider elastic scattering spectros- copy (ESS). Assessment of other methodologies also revealed a high degree of subjectivity in interpreta- tion of results and a shallow learning curve that would not have been prac- tical for translational use in primary care settings. ESS was chosen for sev- eral reasons: ease of use, rapid appli- cability to our problem of ‘real-time’ diagnosis and cost effectiveness. The advantages of ESS also include its high predictive value. ESS is an evolving technology2–9 that generates a wavelength-dependent spectrum that correlates with struc- tural and morphological change within tissues. The spectrum reflects both scattering and absorptive prop- erties of that tissue. This scattering process has been shown to occur at gradients in the optical index of re- fraction resulting from differences in densities that occur at a cellular and sub-cellular level. The structures that induce the scattering (scattering centres) are the nucleus, chromatin concentration and sub-cellular or- ganelles5–14. At a cellular level, ESS is sensitive to nuclear size/morphol-

• gy, chromatin content, nuclear/

cytoplasmic ratio and cellular crowd- ing (and chromatin level), whilst at a tissue level, it is sensitive to changes in morphology of epithelial surface texture and thickness as well as disor- ganization of epithelial cell orienta- tions or architecture. These are all criteria that a histopathologist looks for when establishing malignancy within a tissue and can be used as a basis for comparison of lesions against tissue controls detected by the ESS

• system. ESS has the advantage of be-

ing fast, reliable and cost effective and potentially offers a diagnosis in situ, non-invasively and in real-time.

Materials and methods

The procedure was approved by an ethical committee. All recruited sub- jects were fully informed and they granted their consent to participate in this study. Exclusion criteria in- cluded patients less than 18 years of age, those who were pregnant and those who did not provide their con-

• sent. Only de novo cases who pre-

sented with new skin lesions in the head and neck region were examined using this protocol. Our methodology remained consistent throughout the series of ESS protocols and is de- tailed2; this work builds upon our previous findings. In the study group, the patient highlighted the skin lesion that was

• f concern. If excision was clinically

indicated, the lesion would be photo- graphed and then interrogated using the ESS system. This consisted of ob- taining multiple ESS signatures from anatomically similar but uninvolved skin for control readings, i.e. internal control group. The lesion would then be subject to ESS analysis by taking multiple readings from different areas of the lesion. In our protocol, the tip of the fibre was momentarily placed perpendicularly in direct contact with the suspected lesion and the measurement activated at the keyboard or foot switch. The system automatically takes a background measurement without firing the lamp, followed immediately (within 100 ms) by an ESS measurement with a pulsed lamp and then sub- tracts the background spectrum from the ESS spectrum. Three types of op- tical measurements can be acquired from each of the suspected lesions: 1st type of measurement from the centre, 2nd type from the periphery

• f the lesion and the 3rd type from in
• between. All readings were taken at

SLIDE 3

Page 3 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure. For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

ESS measurements were matched with histopathological specimens taken from individual sites (Table 2). Statistical analysis (Table 3) showed the ability of ESS to discriminate be- tween normal, benign (Figures 1–3) between-class distance to the within- class distance and support vector ma- chine, which maximises the margin between the two data sets. Errors were further minimised by orthogonal subtraction for increasing accuracy24.

Results

The demographics of the subjects studied indicated that the average age was 54 years (±15 years) for the BCC group (range 30–75 years), most were male (62%), all were Caucasian,

• nly 6 identified themselves as smok-

ers and 73% took regular alcohol. Of the skin lesions examined (Table 1), 19 were excluded because they represented less common be- nign skin lesions leaving 21 BCCs, 16 intra-dermal naevi, 10 fibro-epithelial polyps and 7 seborrhoeic keratotic lesions for comparative analysis. No squamous cell carcinomas or mela- nomas were identified in our sample. least six times from each site. An elas- tic scattering spectrum is recorded at each reading and stored electronically. Following the optical readings, a co-registered excisional biopsy was taken, preserved in formalin, pro- cessed in haematoxylin and eosin stain and examined by a histopathol-

• gist; digital pictures were taken and

sketches made for all the suspected sites to ensure that all sites were eas- ily identified by a pathologist, and hence, reduce any chance of false

• results. The skin defect was repaired

in a cosmetically sensitive manner. Cases were discussed with a surgical

• ncologist at a Multidisciplinary

Forum for continuing follow-up. Statistical approach for model generation and model validation Some previous studies have made use of normalised spectra, i.e. the spectra from the lesion are divided by the spectrum of normal skin adjacent to the lesion2. We collected normal skin spectra, but found that classifi- cation using normalised spectra was less accurate. There is a paucity of in- formation on the relationship be- tween specific types of pigmented lesions and their reflectance spectra. In the absence of such information, we approached the problem of lesion classification with statistical tech- niques and sought classifiers that could separate lesions into distinct pathology groups. Both principle component analysis (PCA) and linear discriminate analysis (LDA) strate- gies were used to interrogate data. PCA helped to reduce the dimension- ality of the data set and helped to identify new meaningful underlying variables; LDA was useful for classifi- cation, enabling supervised training

• f neural networks and testing by the

‘leave one out’ (‘jacknife’) validation

• methodology. In summary, these two

methods of analysing data allowed us to examine which discriminated best between spectra from normal tis- sue and BCCs: linear discriminant anal- ysis, which maximises the ratio of the Table 1 The 4 most common pathologies of the patjent populatjon studied Diagnosis % BCC 28.8 Intra-dermal naevus 21.9 Fibro-epithelial polyp 13.7 Seborrhoeic keratosis 9.6 Other benign lesions 26 Table 2 The number of pathological sites interrogated and number of site specifjc spectra obtained Number

• f sites

Number

• f spectra

BCC 59 570 Intra-dermal naevus 31 633 Fibro-epithelial polyp 23 502 Benign seborrhoeic keratosis 12 269 Normal skin 125 551

400 −2 500 600 Wavelength 700 800 −1 Intensity 1 2

Figure 1: A representative elastic scattering spectral plot of the mean fibro-epithelial polyp (blue) and intra-dermal naevus (red) spectra (solid lines), with one standard devi- ation plotted either side of the mean (dashed lines).

−5 0.0 5 Score 10 0.1 0.2 0.3 Frequency 0.4 0.5

400 −2 −1 Intensity 1 2 400 600 Wavelength 700 800

Figure 2: A histogram of canonical scores for classification distinguish- ing between fibro-epithelial polyp (blue) and intra-dermal naevus (red) and showing per spectra sensitivity

• f 80.4%, specificity of 82.9% and ac-

curacy of 81.4%. Per site sensitivity was 100%, specificity was 65.2% and accuracy was 84.9%. Inset summary plot2 shows mean BCC (dark blue), normal (red), intra-dermal naevus (light blue), fibro-epithelial polyp (violet red) and seborrhoeic kerato- sis (orange) spectra.

SLIDE 4

For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74. Page 4 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

81.5%. Per site sensitivity was 94.4%, specificity 43.2%, accuracy 68.3% (Figure 10). Analysis of the receiver

• perating characteristic (ROC) curve

results may suggest that false nega- tive (FN) measurements could have been as a result of a small sample size spectra, but found that classification using normalised spectra was less ac-

• curate. We do not further describe

these results. Discrepancy occurring from combining spectra from differ- ent anatomical sites and also could be related to the co-registration of the suspected site. Comparison of the histological di- agnosis and ESS in the diagnosis of the skin pathology group (BCC) showed per spectra sensitivity of 88.4%, spec- ificity of 74.2% and accuracy of and malignant lesions (Figures 4–9). Since BCC is the most common skin cancer, further subgroup analysis showed sensitivity (high seventies) and specificities (high eighties). Exact figures varied depending upon the analysis methodology and underly- ing assumptions used (Table 4). Some previous studies have made use of normalised spectra, i.e. the spectra from the lesion are divided by the spectrum of normal skin adjacent to the lesion. We collected normal skin Table 3 Analysis results and ability of the elastjc spectroscopy technique to discriminate between normal, benign and malignant lesions. Also shown is the similar discriminatjng ability of the technique when comparing readings taken from the periphery of a BCC and the centre of the BCC Sensitjvity Specifjcity Accuracy Per spectra Per site Per spectra Per site Per spectra Per site BCC vs intra-dermal naevus 0.8789474 0.9859155 0.9787928 0.8846154 0.9306847 0.9587629 BCC vs fjbro-epithelial polyp 0.8385965 0.9577465 0.8884462 0.7826087 0.8619403 0.9148936 BCC vs benign seborrhoeic keratosis 0.9543860 1 0.9665428 0.9166667 0.9582837 0.9879518 BCC vs normal 0.8842105 0.9436620 0.7422868 0.4324324 0.8144514 0.6827586 BCC periphery vs BCC centre 0.115880 0.3750000 0.6720257 0.3939394 0.4319853 0.3846154 Intra-dermal naevus vs fjbro-epithelial polyp 0.8036466 1.0000000 0.8286853 0.6521739 0.8139918 0.8490566 Intra-dermal naevus vs benign seborrhoeic keratosis 0.6579926 0.7500000 0.9929874 0.8666667 0.9012220 0.8333333 Fibro-epithelial polyp vs benign seborrhoeic keratosis 0.8996283 0.9166667 0.9581673 0.7391304 0.9377432 0.8000000

0.0 0.2 0.4 0.0 0.6 0.8 1-specificity 1.0 0.2 0.4 0.6 Sensivity 0.8 1.0

Figure 3: An illustrative set of ROC curves for fibro-epithelial polyp and intra-dermal naevus, which charac- terise the trade-off of sensitivity against specificity as the classifier threshold is adjusted.

400 −2 500 600 Wavelength 700 800 −1 Intensity 1 2

Figure 4: A representative elastic scattering spectral plot of the mean fibro-epithelial polyp (blue) and BCC (red) spectra (solid lines), with one standard deviation plotted either side of the mean (dashed lines).

−5 0.0 5 Score 10 0.2 0.4 0.8 Frequency 0.6 1.0 1.2

400 −2 −1 Intensity 1 2 400 600 Wavelength 700 800

Figure 5: A histogram of canonical scores for classification distinguish- ing between fibro-epithelial polyp (blue) and BCC (red) and showing per spectra: sensitivity 83.9%, spe- cificity 88.8% and accuracy 86.2%. Per site sensitivity was 95.8, specifi- city was 78.3% and accuracy was 91.5%. Inset summary plot2 shows mean BCC (dark blue), normal (red), intra-dermal naevus (light blue), fibro-epithelial polyp (violet red) and seborrhoeic keratosis (orange).

SLIDE 5

Page 5 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure. For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

systems rely on the fact that the opti- cal spectrum derived from any tissue will contain information about the histological and biochemical make up

• f that tissue.

This is one of the first clinical stud- ies examining BCC of the skin using

• ESS. It demonstrates the utility of this

technique to determine the nature of the lesion to be removed before the availability of pathological data. This means that the resection can be guided to improve decision making, preoperative patient consent, intra-

• perative

cancer resection and also during the acquisition of patho- logical specimens to help improve the sample yield of disease material extrapolating to other tumour types since clones of differing aggressivity may exist in squamous cell carcinoma, melanoma and sarcomatous tumour

• types. Even within a morphoeic BCC,

the periphery may not give a true re- flection of the tumour (Figure 11).

Discussion

The current standard of assessing pathological changes in tissue is his- topathology of stained paraffin sec-

• tions. However, the processing of

biopsy material and the diagnosis of the results invariably leads to treat- ment delay and the added possibility

• f taking an unrepresentative sam-
• ple. Lately, optical spectroscopy sys-

tems have been explored to try to provide tissue diagnosis in real-time, non-invasively and ‘in situ’. These

• r due to errors in sampling and co-

registration of the data with histopa-

• thology. The false positive (FP)

measurements could be due to in- flammatory changes and vascular transformation which is a feature of BCC, but they may be incorrectly in- terpreted as an indicative of cancer. In our studied sample population, we found that sampling the periph- ery of the BCC gave similar results to sampling from the centre. We re- gard this data with caution when

0.0 0.2 0.4 0.0 0.6 0.8 1-specificity 1.0 0.2 0.4 0.6 Sensivity 0.8 1.0

Figure 6: An illustrative set of ROC curves for normal skin and BCC, which characterise the trade-off of sensitivity against specificity as the classifier threshold is adjusted.

400 −2 500 600 Wavelength 700 800 −1 Intensity 1 2

Figure 7: A representative elastic scattering spectral plot of the mean normal skin (red) and BCC (blue) spectra (solid lines), with one stan- dard deviation plotted either side of the mean (dashed lines). Haemo- globin absorption shows two charac- teristic dips at approximately 540 nm and 580 nm.

−5 0.0 5 Score 10 0.1 0.3 Frequency 0.2 0.4 0.5

400 −2 −1 Intensity 1 2 400 600 Wavelength 700 800

Figure 8: A histogram of canonical scores for classification distinguish- ing between normal skin (blue) and BCC (red) and showing per spectra sensitivity of 88.4%, specificity of 74.2% and accuracy of 81.5%. Per site sensitivity was 94.4%, specificity was 43.2% and accuracy was 68.3%. An inset summary plot2 shows mean BCC (dark blue), normal (red), intra- dermal naevus (light blue), fibro- epithelial polyp (violet red) and seborrhoeic keratosis (orange). Examination of the spectra acquired from histologically positive sites (e.g. BCC-dark blue) and histologically negative sites (normal tissue-red) re- veal that the main difference was the slope between 520 nm and 600 nm. Figure 9: Indicative results based on sensitivity and specificity at a single threshold as well as an illustrative set

• f ROC curves, which characterise the

trade-off of sensitivity against specif- icity as the classifier threshold is ad-

• justed. ROC curve (right) showing the

results of comparison between gold standard (histopathological analysis)

• f the skin pathology group (59 BCC)

and 125 normal against the results of ESS in this study. Each specimen was sampled several times, e.g. 59 BCCs gave 397 separate true ESS spectra. The results were normalised against internal controls and then linear dis- criminant analysis methodology was

• applied. Receiver operating charac-

teristic curve of the BCC study. The area under the curve (AUC) was 79%. This statistical application suggests that ESS was a good diagnostic test.

0.0 0.2 0.4 0.0 0.6 0.8 1-specificity 1.0 0.2 0.4 0.6 Sensivity 0.8 1.0

SLIDE 6

For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74. Page 6 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

with a high sensitivity and specificity; this is operator and experience inde-

• pendent. Thus, we suggest that this

novel optical diagnostic technology can be used to aid diagnosis of skin lesions in a primary care setting2. It is felt that if clinical diagnosis could be supported by non-invasive diagnostic methods, the need for surgical exci- sion of suspicious looking but non- significant skin lesions would be

• reduced. This in turn would save pa-

tients a surgical burden in terms of morbidity and mortality, thus ensuring significant saving1,2,17–20,23. Cost effec- tiveness would be shown in terms of the cost of the equipment and minor surgery service against the costs of re- ferral to secondary care and follow-up. We present a paradigm shift to that

• f improved evidence based surgical

resection in primary care settings that directly addresses our patient

• concerns. This model can also be ap-

plied throughout in a secondary care setting (clinical interface) with fur- ther benefits to patients and signifi- cant cost savings to the organisation. This manuscript answers some of the criticisms of recent debate in transla- tional research20–23 to directly benefit patient care. Challenges The ESS system was assessed very rigorously against histopathological 64%, as concluded from large clinical studies15,16. This is highly dependent upon the skills and experience of the

• clinician. In this study, we have shown

that we can diagnose BCC lesions to reduce the FP associated with sim- ple naked eye examination. We were able to use the ESS system to discrimi- nate BCC lesions from normal skin tis- sues, benign lesions and to develop an algorithm for diagnosis2. Further, our results show that BCC (the common- est skin cancer) diagnosis can be achieved by ESS with a high degree of accuracy. Clinical examination has a sensitiv- ity 40%–92% and specificity of up to

400 500 600 −2 700 Wavelength 800 −1 1 Intensity 2

Figure 10: A representative sum- mary plot of the mean normal (red), intra-dermal naevus (sky blue), fi- bro-epithelial polyp (violet red) and seborrhoeic keratosis (orange). ESS spectra showing the discriminating features of the absorption spectrum which may be useful in characterisa- tion and diagnosis between these be- nign lesions and normal. Table 4 BCC vs normal skin results Comparison of the histological diagnosis and ESS in the diagnosis of the skin pathology group Test: ESS diagnosis Disease: BCC histological diagnosis (Gold standard) Sites BCC Normal Total (n) BCC (59) True positjve (TP) 397 False positjve (FP) 29 All with positjve test = (TP + FP) 426 Normal (125) False negatjve (FN) 113 True negatjve (TN) 118 All with negatjve test = (FN + TN) 231 Total (n) All with disease 510 All without disease 147 All = (TP + FP + FN + TN) 657 Sensitjvity = TP/(TP + FN) 77.8% Specifjcity = TN/(TN + FP) 80.3% Pre-test probability = (TP + FN)/(TP + FP + FN + TN) 64.8% Positjve predictjve value = TP/(TP + FP) 93.2% Negatjve predictjve value = TN/(FN + TN) 51.1%

0.0 0.2 0.4 0.0 0.6 0.8 1-specificity 1.0 0.2 0.4 0.6 Sensivity 0.8 1.0

Figure 11: Discrimination between BCC data obtained from periphery and centre sites showing multiple spectral measurements from differ- ent areas of the same lesion, centre, periphery and whole, with indicative results based on sensitivity and spe- cificity at a single threshold as well as an illustrative set of ROC curves, which characterise the trade-off of sensitivity against specificity as the classifier threshold is adjusted. Per spectra sensitivity was 11.2%, spe- cificity was 67.2% and accuracy was 43.2%. Per site sensitivity of 37.5%, specificity of 39.4% and accuracy

• f 38.5%.

SLIDE 7

Page 7 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure. For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

• 5. Jerjes W, Swinson B, Pickard D,

Thomas GJ, Hopper C. Detection of cervi- cal intranodal metastasis in oral cancer using elastic scattering spectroscopy. Oral Oncol. 2004 Aug;40(7):673–8.

• 6. Jerjes W, Swinson B, Johnson KS,

Thomas GJ, Hopper C. Assessment of bony resection margins in oral cancer using elastic scattering spectroscopy: a study on archival material. Arch Oral Biol. 2005 Mar;50(3):361–6.

• 7. Lovat LB, Johnson K, Mackenzie GD,

Clark BR, Novelli M, Davies S, et al. Elastic scattering spectroscopy accurately de- tects high grade dysplasia and cancer in barrett’s oesophagus. Gut. 2006 Aug; 55(8):1078–83.

• 8. Jerjes W, Swinson B, Pickard D,

Thomas GJ, Hopper C. Detection of cervi- cal intranodal metastasis in oral cancer using elastic scattering spectroscopy. Oral Oncol. 2004 Aug;40(7):673–8.

• 9. Bigio IJ, Bown SG. Spectroscopic sens-

ing of cancer and cancer therapy: current status of translational research. Cancer Biol Ther. 2004 Mar;3(3):259–67.

• 10. Tata DB, Foresti M, Cordero J,

Tomashefsky P, Alfano MA, Alfano RR. Fluorescence polarization spectroscopy and time-resolved fluorescence kinetics

• f native cancerous and normal rat kid-

ney tissues. Biophys J. 1986 Sep;50(3): 463–9.

• 11. Epstein JB, Feldman R, Dolor RJ,

Porter SR. The utility of tolonium chloride rinse in the diagnosis of recurrent or sec-

• nd primary cancers in patients with

prior upper aerodigestive tract cancer. Head Neck. 2003 Nov;25(11):911–21.

• 12. Nagao T, Ikeda N, Fukano H, Miyazaki H,

Yano M, Warnakulasuriya S. Outcome following a population screening pro- gramme for oral cancer and precancer in Japan. Oral Oncol. 2000 Jul;36(4): 340–6.

• 13. Gynther GW, Rozell B, Heimdahl A.

Direct oral microscopy and its value in diagnosing mucosal lesions: a pilot

• study. Oral Surg Oral Med Oral Pathol

Oral Radiol Endod. 2000 Aug;90(2): 164–70.

• 14. Andrea M, Dias O. Contact endoscopy
• f the upper aerodigestive tract. In Endo-

press, Tuttlingen; 2001:10–6.

• 15. Sharwani A, Jerjes W, Salih V,

MacRobert AJ, El-Maaytah M, Khalil HS, et al. Fluorescence spectroscopy com- bined with 5-aminolevulinic acid- induced protoporphyrin IX fluorescence

tuned for maximum sensitivity. If treatment is to be based just on the

• ptical diagnosis, then the analysis

can be tuned for maximum specific-

• ity. ESS can also help determine the

diagnosis between benign lesions with a high degree of accuracy, sug- gesting that ESS can add significant

• bjectivity to dermatological diagno-

sis and management even of benign conditions.

Abbreviations list

BCC, basal cell carcinoma; ESS, elastic scattering spectroscopy; FN, false negative; FP, false positive; LDA, linear discriminate analysis; PCA, principle component analysis; ROC, receiver

• perating

characteristic curve; TN, true negative; TP, true pos- itive; UV, ultraviolet.

Acknowledgement

We acknowledge Miss Jiao for her help with statistical analysis.

Declaration

Some of the preliminary unpublished work for this pilot study was submit- ted as a manuscript to the RCGP and received a Roché Research Prize from the Royal College of General Practitioners (UK).

References

• 1. Upile T, Jerjes W, Betz CS, El Maaytah M,

Wright A, Hopper C. Optical diagnostic techniques in the head and neck. Dent

• Update. 2007 Sep;34(7):410–2, 415–6,

419–20 passim.

• 2. Upile T, Jerjes W, Radhi H, Mahil J, Rao

A, Hopper C. Elastic scattering spectros- copy in assessing skin lesions: an “in vivo”

• study. Photodiagnosis Photodyn Ther.

2012 Jun;9(2):132–41.

• 3. Bigio IJ, Bown SG, Briggs G, Kelley C,

Lakhani S, Pickard D, et al. Diagnosis of breast cancer using elastic-scattering spectroscopy: preliminary clinical results. J Biomed Opt. 2000 Apr;5(2):221–8.

• 4. Mourant JR, Hielscher AH, Eick AA,

Johnson TM, Freyer JP. Evidence of intrin- sic differences in the light scattering properties of tumorigenic and nontumor- igenic cells. Cancer. 1998 Dec;84(6): 366–74.

results rather than clinical examina- tion alone2. We realise that informa- tion from ESS in these cases only represents superficial cells and may not reflect deep underlying invasion. Also, since many skin lesions may be heterogeneous, co-registration of ESS signature with the pathological re- port is a potential problem. We have tried to overcome this issue by mul- tiple statistical sampling. From our sample, we can also determine that for the BCCs studied, there was little difference in the spectral signatures between the periphery and centre of the lesions. Occasionally, very high or low readings would necessitate the recalibration of the spectroscope.

Conclusion

This preliminary study shows that ESS can distinguish between BCC, common benign conditions and vari- ants of normal skin. ESS can also help determine the diagnosis between be- nign lesions with a high degree of

• accuracy. The results are better than

those usually obtained from frozen section analysis (the closest conven- tional pathology analysis system for rapid analysis of results to guide treatment). Its advantage over frozen section is its reliability, speed and steep learning curve, cost effective- ness and non-operator dependence. One would presume that melanoma would affect even greater ESS spec- tral differences to help guide the ini- tial surgical resection and thus improve management and prognosis. ESS, as an optical diagnostic tech- nique, is an objective, quick and accu- rate way of examining tissues. The accuracy found in many of our previ-

• us studies suggests that it is cost ef-

fective and can be used for real-time and in situ diagnosis. The analysis can be varied to achieve higher speci- ficity or higher sensitivity, depending

• n the clinical situation (whether for

diagnosis or screening). If the aim is to identify high-risk sites for taking a conventional biopsy to make a defini- tive diagnosis, then the analysis is

SLIDE 8

For citation purposes: Upile T, Jerjes W, Johal O, Lew-Gor S, Mahil J, Sudhoff H. A new tool to inform intra-operative decision making in skin cancer treatment: the non-invasive assessment of basal cell carcinoma of the skin using elastic scattering spectroscopy. Head Neck Oncol. 2012 Oct 31;4(3):74. Page 8 of 8 Competjng interests: none declared. Confmict of interests: none declared. All authors contributed to the conceptjon, design, and preparatjon of the manuscript, as well as read and approved the fjnal manuscript. All authors abide by the Associatjon for Medical Ethics (AME) ethical rules of disclosure.

Original research study

Licensee OA Publishing London 2012. Creative Commons Attribution License (CC-BY)

• 22. Andrew C. Cooksey report recom-

mends central coordinating body for re-

• search. BMJ. 2006;333:1239.
• 23. Nagao T, Ikeda N, Fukano H, Miyazaki H,

Yano M, Warnakulasuriya S. Outcome fol- lowing a population screening pro- gramme for oral cancer and precancer in Japan. Oral Oncol. 2000 Jul;36(4): 340–6.

• 24. Zhu Y, Fearn T, Mackenzie G,

Clark B, Dunn JM, Bigio IJ, et al. Elastic scattering spectroscopy for detection of cancer risk in Barrett’s esophagus: experimental and clinical validation of error removal by orthogonal subtraction for increasing accuracy. J Biomed Opt. 2009 Jul-Aug;14(4): 044022. coherence tomography imaging in the diagnosis of non-melanoma skin cancer and benign lesions versus normal skin:

• bserver-blinded evaluation by derma-

tologists and pathologists. Dermatol Surg. 2009 Jun;35(6):965–72.

• 19. Korde VR, Bonnema GT, Xu W,

Krishnamurthy C, Ranger-Moore J, Saboda K, et al. Using optical coherence tomography to evaluate skin sun damage and precancer. Lasers Surg Med. 2007 Oct;39(9):687–95.

• 20. Cooksey D. A review of UK Health

Research Funding. London: Stationery Office, 2006.

• 21. Black N. The Cooksey review of UK

health research funding. BMJ. 2006 Dec; 333(7581):1231–2. in detecting

• ral

premalignancy. J Photochem Photobiol B. 2006 Apr; 83(1):27–33.

• 16. Zhao J, Lui H, McLean DI, Zeng H.

Real-time Raman spectroscopy for non- invasive skin cancer detection—prelimi- nary results. Conf Proc IEEE Eng Med Biol

• Soc. 2008;2008:3107–9.
• 17. Gniadecka M, Philipsen PA, Sigurdsson S,

Wessel S, Nielsen OF, Christensen DH, et al. Melanoma diagnosis by Raman spectroscopy and neural networks: struc- ture alterations in proteins and lipids in intact cancer tissue. J Invest Dermatol. 2004 Feb;122(2):443–9. 18. Mogensen M, Joergensen TM, Nürnberg BM, Morsy HA, Thomsen JB, Thrane L, et al. Assessment of optical