ORIGINAL ARTICLE
Facial biometrics of peri-oral changes in Crohn’s disease
L. Zou & O. K. Adegun & A. Willis & Farida Fortune
Received: 16 October 2012 /Accepted: 8 February 2013 /Published online: 24 February 2013# Springer-Verlag London 2013
Abstract Crohn’s disease is a chronic relapsing and remit-ting inflammatory condition which affects any part of thegastrointestinal tract. In the oro-facial region, patients canpresent peri-oral swellings which results in severe facialdisfigurement. To date, assessing the degree of facialchanges and evaluation of treatment outcomes relies onclinical observation and semi-quantitative methods. In thispaper, we describe the development of a robust and reproduc-ible measurement strategy using 3-D facial biometrics toobjectively quantify the extent and progression of oro-facialCrohn’s disease. Using facial laser scanning, 32 serial imagesfrom 13 Crohn’s patients attending the Oral Medicine clinicwere acquired during relapse, remission, and post-treatmentphases. Utilising theories of coordinate metrology, the facialimages were subjected to registration, regions of interest
identification, and reproducible repositioning prior toobtaining volume measurements. To quantify the changes intissue volume, scan images from consecutive appointmentswere compared to the baseline (first scan image). Reproduc-ibility test was performed to ascertain the degree of uncertain-ty in volume measurements. 3-D facial biometric imaging is areliable method to identify and quantify peri-oral swelling inCrohn’s patients. Comparison of facial scan images at differ-ent phases of the disease revealed precisely profile and vol-ume changes. The volume measurements were highlyreproducible as adjudged from the 1 % standard deviation.3-D facial biometrics measurements in Crohn’s patients withoro-facial involvement offers a quick, robust, economical andobjective approach for guided therapeutic intervention androutine assessment of treatment efficacy on the clinic.
Keywords Crohn’s disease . Peri-oral changes . Facial laserscan . Volumemeasurement . Facial biometrics
Introduction
Crohn’s disease (CD) is a relapsing and remitting granulo-matous inflammatory condition that can affect the gastroin-testinal tract from mouth to anus. In the oro-facial region,Crohn’s disease can present as peri-oral swelling which canresult in severe facial disfigurement [1–3]. To date, variousapproaches at quantifying the degree of facial swelling havebeen proposed. Chiandussi [4] quantified dimensionalchanges in the lips by obtaining lip impressions for subse-quent casting and acquisition of linear measurements usinga calliper. Although useful and capable of providing objec-tive measurements compared to clinical photographs [5, 6],the technique is restricted to the lip area and thus is notcapable of monitoring peri-oral tissues. Furthermore, thelack of reproducibility associated with impression taking
L. ZouAdult Oral Health, Institute of Dentistry,Barts and the London School of Medicine and Dentistry,Queen Mary, University of London,London, UK
O. K. AdegunClinical and Diagnostic Oral Sciences, Institute of Dentistry,Barts and the London School of Medicine and Dentistry,Queen Mary, University of London,London, UK
A. WillisBarts and the London School of Medicine and Dentistry,London, UK
F. Fortune (*)Clinical and Diagnostic Oral Sciences, Institute of Dentistry,Barts and the London School of Medicine and Dentistry,Queen Mary, University of London,2nd floor, Turner Street,London E1 2AD, UKe-mail: [email protected]
Lasers Med Sci (2014) 29:869–874DOI 10.1007/s10103-013-1286-5
and model casting results in measurement variations. Addi-tionally, the linear measurements are not representative ofvolumetric tissue changes. A further complication is thatassociated with selecting a consistent measurement locationat each visit making the progressive evaluation of peri-oralchanges difficult to achieve.
Accordingly, an alternative measurement technique isrequired which is sensitive to the peri-oral changes associ-ated with Crohn’s disease. Such an approach involving anon-contact surface laser scanner and its performance char-acteristics is evaluated in the current study.
Methods
Subjects
Thirteen patients (six females and seven males; age range, 9–44 years; mean age, 19.2 years) with diagnosed Crohn’s dis-ease were recruited from the Oral Medicine Clinic at Barts andthe London School of Medicine and Dentistry. The prerequi-site for study inclusionwas that the patients have had prior oro-facial manifestation of the disease. A total of 32 serial facialscans were acquired at follow-up appointments which spannedbetween May 2011 and December 2012. The number of scanimages per patient varied between two and four (Table 1). Theinitial scan images (baseline) were acquired at the patient’s firstpresentation during the study period. Consequently, baselineperi-oral disease status varied between the relapse and remis-sion phases as adjudged by the clinician. Subsequent scanswere acquired and subject to image analysis and volumemeasurements with a view to identifying and quantifyingprofile and tissue changes, respectively.
Facial laser scanning
The system used in this study was an in-house facial laserscanner with performance characteristics consistent with theequipment used by Benn et al. for evaluation of facial surfacemorphology [7]. This scanning system is based on the princi-ple of optical triangulation; it utilises a 0.655-mm wavelengthclass I semi-conduct laser. A scanning interval of 5° was used.Back reflected light from the patient’s face at multiple pointswithin a rotation range of 220° was recorded by a CCDcamera. The series of facial profiles which constituted a facialscan image was acquired in approximately 7 s.
Image registration and selection of region of interest
The registration of facial scan images between appointmentsis a prerequisite for identifying areas of changes in facialprofiles. Therefore in this study, the nasal bridge, ear lobesand forehead areas, all of which satisfied previously
described criteria for registration, were selected [8]. Usingthe algorithm of Iterative Closest Point (ICP, Cloud, UCL,UK), two facial images were registered on the basis of theseselected areas. Post-registrations, individualised regions ofinterest (ROI) were defined based on the detected changes infacial profiles (Fig. 2). The ROIs were then used as tem-plates to ensure a consistent area and position was analysedon successive facial images.
Facial image positioning and volume measurement
Precise volume measurements rely on locating the registeredfacial scan images in identical positions during image
Table 1 Number of serial scans, scan images showing individualisedtemplates with maximum profile changes (The colour scale beneath eachsubtracted image spans from light purple (left side) to blue then yellowand red (right side) representing negative to positive profile changes,respectively), value of volume changes and oro-facial disease status atsecond, third and fourth scans for all patients, respectively
870 Lasers Med Sci (2014) 29:869–874
analysis. This presents a challenge as the patient’s headposition and sitting posture vary between appointments. Toobviate this challenge, a reference frame was developed toachieve uniformity in the alignment and positioning of facialscan images. The reference frame was determined by threepoints, namely the two outer canthi of the eyes (A, B) andone deepest point on the midline of nasal bridge (C), asillustrated in Fig. 1a. These points were selected as theyprovided relative stability and facilitated reproducible iden-tification regardless of the head position of the subject.
Using a 3-D analytical software (Cloud UCL, UK), thereference frame, as indicated by the red line in Fig. 1b washorizontally aligned and point C centred in the forwarddirection. This allowed registered facial images from scanappointments to be aligned consistently, such that theindividualised ROI corresponded during successive imageanalysis. The subsequent volume measurements were calcu-lated over each ROI, using identical software as for facialimage positioning.
Reproducibility test of facial volume measurement
To test the reproducibility of the measurement strategydeveloped in this study, facial registration, selection of re-gions of interest, image positioning and volume measure-ments were performed on five occasions between the twoscans of a patient. The mean and standard deviation werecalculated over the five volume measurements.
Results
Facial profiles changes associated with disease relapseand treatment
To demonstrate the ability of the measurement strategy toidentify changes in facial profile, each second facial scanimages for example in patients 1 and 2 (T1P1b and T1P2b)
acquired during disease relapse were registered against thecorresponding baselines at a quiet period (T1P1a andT1P2a) as shown in Table 2. This allowed for thelocalisation of areas of facial involvement and definition ofindividualised ROI for each patient shown in Table 1. Forexample, patient 1 displayed a swelling extending from theinfraorbital margin to the cheeks, upper and lower lip on theleft side of the face (closed orange line in Fig. 2a). Whereasin patient 2, the entire lower lip extending to the angles ofthe mouth and the chin areas showed considerable swelling(closed orange line Fig. 2b). Interestingly, the variation inthe areas of involvement (ROI) observed in the study pa-tients (shown in column 3 of Table 1) mirrors the reportedunpredictability of the clinical presentation of oro-facialCrohn’s disease [18]. Furthermore, the degrees of facialswellings were also observed to vary within the areas ofinvolvement in each patient. As an example, the severity ofthe facial swelling was seen to progressively increase fromthe peri-nasal to the peri-oral area with the site of maximumchange on the left side of the upper lip (red colourrepresenting profile change of +7.1 mm in Fig. 2a). Simi-larly, patients 2 and 8 (Table 1) at their second scan appoint-ment (disease relapse) displayed specific sites of maximumchanges within their regions of interest. By contrast, patients3, 5 and 9 (Table 1) at their post-treatment phase displayedsignificant reduction in their facial swellings with areas ofmaximum changes also recognisable. A specific example ofnegative change is demonstrated in Fig. 2c.
Volume changes at successive patient appointments
In this study, the individualised regions of interest in eachpatient were used as templates for volume measurements.This approach ensured the location of volume measure-ments were consistent in successive registered images. Priorto calculating volume measurements associated with facialprofiles changes at the different disease stages, a reproduc-ibility test was performed on two scans from a patient. This
Fig. 1 a, b: a Displaying thethree reproducible landmarksA and B—outer canthi of theeye, C—deepest point of themiddle of the nasal bridge;b demonstrating the referenceframe drawn across pointsA, B and C (red line)horizontally aligned with pointC centred in the forwarddirection
Lasers Med Sci (2014) 29:869–874 871
Table 2 T1P1a/b/c/d and T1P2a/b/c/d—patient 1 and 2 facial scanimages acquired at four consecutive appointments and positioned inthe reference frame, respectively. T1P1e/f/g and T1P2e/f/g—second,third and fourth facial scan images registered against the baseline scan,
showing the use of individualised templates for the acquisition ofvolume measurements. The colour coding illustrates the profile differ-ence which is indicative of oro-facial manifestation at the differentstages of Crohn’s disease
872 Lasers Med Sci (2014) 29:869–874
produced a mean and standard deviation of 10,966.9±104.8 mm3 representing a variability of approximately 1%.
To quantify changes in tissue volume at successive visits,the individualised template for each patient displayed in the3rd column of Table 1 was applied to their correspondingregistered images, a specific example of this is shown inT1P1/e/f/g and T1P2e/f/g (Table 2).
As previously stated, the variation in the facial appear-ance of each patient at the initial scan images (baseline) wasindicative of their respective disease status at the first pre-sentation during the study period. Accordingly, using clini-cal descriptors, patient 1,2 and 8 presented at remission, 3, 5and 9 at relapse while 4, 6 and 7, 10–13 at the quiet stage.Each patient had second scans registered against their cor-responding baselines which revealed differential profile andvolume changes in each patient’s template as shown inTable 1. Likewise, a direct correlation between the valuefor volume changes and the extent of swelling was also seeni.e. the larger the positive value, the greater the facial swell-ings and vice versa. Because the timings of each patientinitial scan varied across the study period, only 3 out of the13 patients had 4 successive scans at the time of preparationof this paper. Coincidently, all 3 with disease relapse at theirsecond scan displayed progressive reductions in their facialvolumes in response to treatment at the 3rd scan. Notable ispatient 3’s positive volume changes at the 4th scan, suchobservation highlights the unique capability of this tech-nique to objectively characterise the spontaneous relapsingand remitting course of this protracted disease.
Discussion
The cosmetic problem arising from diffuse swelling of thelip and face is the principal complaint in adolescents andyoung adults with oral Crohn’s disease presenting to an oralmedicine clinic [9]. To date, measurements from clinicalphotographs and the lip impressions technique remain themainstay for judging the extent and degree of the facial
involvement [4]. Although valuable, they are limited bythe assessment inaccuracies arising from the use of non-standardised measurement strategies. This is particularlythe case with lip impressions as they are prone to distortionsarising from impression taking and subsequent laboratoryprocesses. Furthermore, the restriction to linear measure-ments of the lip discounts the characteristic facial swellingreported in oral Crohn’s disease [10, 11]. Lastly, the spon-taneous relapsing and remitting course of the disease makesthe efficacy of therapy difficult to assess clinically [17]. Toobviate these shortcomings, non-contact and non-invasivesurface laser scanning capable of 3-D imaging was used toassess facial involvement in oral Crohn’s disease. To thebest of our knowledge, this represents the first study toreport this clinical application. In addition to assessing theextent of facial swelling, tissue changes associated with therelapse, remission and post-treatment (T1P1e/f/g andT1P2e/f/g in Table 2) of Crohn’s disease were quantified.
The comparison of patient facial scan images acquired atdifferent appointments is dependent on achieving a repro-ducible position. To minimise variations, previous studieshave utilised the same operator, head positioning protocolsand the superimposition of scans using bony landmarks [7,12]. Such strategies when employed with Crohn’s patientsare fraught with challenges. This is particularly the case inchildren who are reluctant to adopt an identical head posi-tion at every scan visit. Furthermore, the use of bony land-mark for superimposition of scans can be subject toinaccuracies attributable to growth and weight loss both ofwhich are characteristic of the protracted course of oral CD[1]. To address these shortcomings, a robust measurementstrategy for reproducible head positioning irrespective of itsangulations was developed in this study. This involvedpositioning facial scan images into a defined referenceframe (T1P1a/b/c/d and T1P2a/b/c/d in Table 2), after whichsuccessive scans were registered to the baseline (T1P1e/f/gand T1P2e/f/g in Table 2). Utilising this methodology, thecolour encoded images revealed different degrees of facialinvolvement, as an example, patient 1 displayed change
Fig. 2 a–c: Colour coded images showing changes in facial profile asa result of relapse of the disease. The green, yellow and red areas areindicative of progressively increasing degrees of facial swelling whilst
the purple area signifies a reduction. The closed orange lines illus-trates the boundaries of the region of interest (ROI) which include areasof facial profile changes
Lasers Med Sci (2014) 29:869–874 873
restricted to one side of the upper lip (T1P1e in Table 2)while patient 2 showed changes all around the mouth withthe lower lip more affected (T1P2e in Table 2).
Volume measurement is sensitive to slight changes in theposition, size and boundaries from which they are acquired.To mitigate these effects on successive image analysis, acustomised template illustrating the size and boundaries offacial profile changes in patients was adopted; an exampleof this is shown in Fig. 2a–c. This ensured a consistenttissue area and position was selected prior to performingvolume measurements on successive facial images in eachpatient. In contrast to previous studies where changes infacial tissue volume were reported [1, 15, 16], a high degreeof reproducibility in volume measurement was achievedusing the template system.
This approach provides an objective method for guidedtherapeutic intervention and routine assessment of treatmentefficacy [13, 14] without resorting to the traditional ap-proaches utilised in several case reports [17, 18]. It is welldocumented that oral CD shows a variable and unpredictableresponse to therapy [17]. As anticipated, each patientdisplayed varying amount of volume changes at subsequentscans despite receiving intralesional triamcinolone (40 mg/ml)and/or dietary modification. Although the exact reasons forthis differential response are largely unknown, future quanti-tative facial biometric studies may help provide useful insightsinto the potential reasons.
Although oral CD predating its systemic manifesta-tions is reported to have a low prevalence, the psycho-logical and social impact of the resultant facialdisfigurement in adolescents and young adults can bedaunting [17]. Facial biometrics measurement offers aunique opportunity for early detection of disease relapseand guided therapeutic intervention.
In conclusion, our technique offers a quick, robust,economical and objective approach for routine assess-ment of disease activity and treatment efficacy of theoro-facial manifestations of Crohn’s disease. Most im-portantly, the technique can facilitate guided therapeuticintervention contributing to improved treatment out-comes and an overall reduction in the number oflesional sites infiltrated with steroids.
Acknowledgments The authors acknowledge the financial supportof Barts and The London School of Medicine and Dentistry, QueenMary University of London. We also thank Professor Dan Bader forproviding useful comments during the preparation of this manuscript.
Conflict of interests The authors declare no conflict of interest
References
1. Parente F, Maconi G, Bollani S, Anderloni A, Sampietro G,Cristaldi M et al (2002) Bowel ultrasound in assessment ofCrohn’s disease and detection of related small bowel strictures: aprospective comparative study versus X ray and intraoperativefindings. Gut 50(4):490–495
2. Tilakaratne WM, Freyskottir J, Fortune F (2008) Orofacialgranulomatosis: review on aetiology and pathogenesis. J OralPathol Med 37:191–195
3. Jenss H, Starlinger M, Skaleij M (1992) Magnetic resonanceimaging in perianal Crohn’s disease. Lancet 340(8830):1286
4. Chiandussi S, Tappuni AR, Watson TF, White A, Escudier MP,Sanderson JD et al (2007) Lip impressions: a new method formonitoring morphological changes in orofacial granulomatosis.Oral Dis 13(1):93–98
5. Mignogna MD, Fedele S, Lo Russo L, Lo Muzio L (2003) Themultiform and variable patterns of onset of orofacial granulomatosis.J Oral Pathol Med 32(4):200–205
6. Sciubba JJ, Said-Al-Naief N (2003) Orofacial granulomatosis:presentation, pathology and management of 13 cases. J Oral PatholMed 32(10):576–585
7. Benn P, Ruff C, Cartledge J, Sauret V, CopasA, LinneyA et al (2003)Overcoming subjectivity in assessing facial lipoatrophy: is there arole for three-dimensional laser scans? HIV Med 4(4):325–331
8. Zou LF, Cherukara G, Hao PW, Seymour K, SamarawickramaDYD (2009) Geometrics of tooth wear. Wear 266(5–6):605–608
9. Williams AJ, Wray D, Ferguson A (1991) The clinical entity oforofacial Crohn’s disease. Q J Med 79(289):451–458
10. Rowland M, Fleming P, Bourke B (2010) Looking in the mouth forCrohn’s disease. Inflamm Bowel Dis 16(2):332–337
11. Stricker T, Braegger CP (2000) Images in clinical medicine. Oralmanifestations of Crohn’s disease. N Engl J Med 342(22):1644
12. Benn P, Sauret-Jackson V, Cartledge J, Ruff C, Sabin CA, Moyle Get al (2009) Improvements in cheek volume in lipoatrophic in-dividuals switching away from thymidine nucleoside reverse tran-scriptase inhibitors. HIV Med 10(6):351–355
13. Fernandez-Banares F, Cabre E, Esteve-Comas M, Gassull MA(1995) How effective is enteral nutrition in inducing clinical re-mission in active Crohn’s disease: a meta-analysis of therandomised clinical trials. J Parenter Enter Nutr 19(5):356–364
14. Grave B, McCullough M, Wiesenfeld D (2009) Orofacialgranulomatosis—a 20-year review. Oral Diseases 15:46–51
15. Nkenke E, Langer A, Laboureux X, Benz M, Majer T, Kramer M,Hausler G, Kessler P, Wiltfang J, Neukam F (2003) Validation of invivo assessment of facial soft-tissue volume changes and clinicalapplication in midfacial distraction. Plast Reconstr Surg 112(2):367–380
16. Nkenke E, Vairaktaris E, Kramer M, Schlegel A, Hoist A,Hirschfelder U, Wiltfang J, Neukan FW, Sramminger M (2008)Three-dimensional analysis of changes of the malar-midfacialregion after LeFort I osteotomy and maxillary advancement. OralMaxillofac Surg 12(1):5–12
17. Dupuy A, Cosnes J, Revuz J, Delchier J et al (1999) Oral Crohndisease: clinical characteristics and long-term follow-up of 9 cases.Arch Dermatol 135:439–442
18. Kauzman A, Quesnel-Mercier A, Lalonde B (2006) Orofacialgranulomatosis: 2 case reports and literature review. J Can DentAssoc 72(4):325–329
874 Lasers Med Sci (2014) 29:869–874