Designing visually optimal shade guides

Mostafa Analoui, PhD,a Evrika Papkosta, DDS, MS,b Michael Cochran, DDS, MS,c andBruce Matis, DDS, MSd

Indiana University, Indianapolis, Ind

Statement of problem. Dental shade guides do not provide a broad coverage of tooth colors. There is a needfor shade guides that can provide closer color matches.

Purpose. The purpose of this study was to propose and assess a visually optimal shade guide for tooth colormatching.

Material and methods. With the use of a spectrophotometric approach, the color distribution (L*a*b*) of150 extracted human teeth and 3 commercially available shade guides (Vita Lumin V, Trubyte Bioform ColorOrder Shade Guide, and Vitapan 3D-Master Shade System) was measured. With the use of a hierarchicalclustering approach, a series of shade guides was designed with a varying number of tabs. The average error (DE)between colors from each shade guide and the extracted teeth was computed.

Results. The proposed visually optimal shade guide had the smallest average error of all guides tested. The newguide achieved lower error with fewer shade tabs than the 3 commercial systems evaluated.

Conclusion. Within the limitations of this study, it was demonstrated that a hierarchical clustering techniquecan be used to design a visually optimal shade guide for a given population with a flexible degree of control overthe mean error and number of tabs. (J Prosthet Dent 2004;92:371-6.)

CLINICAL IMPLICATIONS

With the use of a hierarchical clustering technique, an optimal shade guide was designed basedon a broad spectrum of tooth color. This guide may improve color matching in restorativeprocedures.

Dental shade guides are commonly used to evaluatetooth color in restorative and bleaching procedures. Ashade guide is composed of a set of shade tabs intendedto cover the range of colors present in the human denti-tion. The successful achievement of a clinically accept-able color match between a given tooth and a shadetab is closely related to the spectral coverage of the shadeguide, clinician experience,1 and the viewing environ-ment.2-4 Although it is not possible to encompass theentire color spectrum of the teeth with a limited numberof color samples (shade tabs), the underlying assump-tion is that the difference between the true color andthe closest shade tabwould not be discernable by the hu-man eye.

Numerous reports have indicated that commonshade guides do not provide sufficient spectral coverageof the colors present in teeth.5-9 Moreover, tab colors

aSenior Director, Pfizer Global Research and Development.bPrivate practice, Athens, Greece.cProfessor, Department of Operative Dentistry, School of Dentistry,

Indiana University.dProfessor, Department of Operative Dentistry, School of Dentistry,

Indiana University.

OCTOBER 2004

may not be distributed uniformly throughout the colorspace of natural teeth, leading to close matches for someshades and gross mismatches for others. This inability torepresent the color spectrum of teeth, coupled with op-erator error and nonideal viewing conditions,10 couldlead to significant mismatch between tooth color andrestoration color. To address issues associated with theshade tabs, 2 distinct avenues have been pursued: (1)objective spectrophotometric/colorimetric assessmentand (2) the reorganization of currently available shadeguides and the development of new shade guides.

The spectrophotometric/colorimetric approach isattractive because it allows an objective assessment oftooth color, independent of viewing conditions and ex-aminer experience. Tools used with this approach in-clude spectrophotometers,11-13 colorimeters,14-16 film-based photography,17 and digital photography.18

While the spectrophotometric approach offers greateraccuracy and precision than shade guides for color mea-surement, there is a need to overcome a set of practicallimitations before such techniques gain widespread clin-ical use. The most common practical concerns are stan-dardization, cost, and ease of use in a clinical setting. Asan alternative approach, a significant amount of work has

THE JOURNAL OF PROSTHETIC DENTISTRY 371

THE JOURNAL OF PROSTHETIC DENTISTRY ANALOUI ET AL

Fig. 1. A,Hypothetical color spectrum in L*a*b* color space.B, Uniform segmentation of color spectrum. C, Central colorfor each sub-block is marked by dark circular spot.

372

focused on the refinement of shade guides. For example,O’Brien et al19 proposed to rearrange the Bioform shadeguide to a long-range 1-dimensional color system basedon color difference. Utilizing the blending effect ofcomposite restorations, Hall and Kafalias20 constructeda shade guide composed of 9 tabs. Their procedure in-volvedmultiple mixings andmeasurements of candidatespecimens and repeated visual comparisons with manyteeth. More than 700 specimens were developed andtested over a 2-year time span. Vita Zahnfabrik (BadSackingen, Germany) recently introduced a new shadeguide, the Vita 3D-Master Shade System, that is notonly purported to improve the coverage of tooth colorsbut also arranges the tabs within a 3-dimensional colorspace according to lightness and chroma values.

The objective of the present study was to design, im-plement, and assess a visually optimal set of shade tabsfor a given population. All color measurements were ex-pressed in the CIELAB space,21 in which there are 3coordinates: L*, a*, and b*. L* indicates lightness,which ranges from 0 (black) to 100 (white). The quan-tities a* and b* are chromaticity coordinates that indi-cate color directions: positive a* corresponds to thered direction, whereas negative a* indicates the green di-rection; positive and negative b* values correspond toyellow and blue directions, respectively. The central axis,or zero of the a*b* plane, is the achromatic L* scale.CIELAB color space is commonly used in perceptualstudies and for dental color assessment because of its(approximate) visually uniform coverage of the colorspace.21,22 In this color space, the color difference be-tween 2 objects with colors C1 = (L*1,a*1,b*1) andC2 = (L*2,a*2,b*2) is defined as follows:

DE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiðL�

12L�2Þ

2 þ ða�12a�

2Þ2 þ ðb�12b�2Þ

2q

This definition can be used to assess the color differ-ence between a given tooth and tabs within a shadeguide. Alternatively, for a given set of teeth, averagecolor differences between the color of the teeth andthe colors represented by a specific shade guide can becomputed. Assuming that for a given tooth, the bestmatch can be found among the tabs of a shade guide,the average color difference (DE) represents the averageerror in matching tooth color to the given shade guide.

Minimizing the average DE would lead to a shadeguide with the smallest perceivable average color error(between teeth and tabs), for example, a visually optimalshade guide (VOS). To design such a set, a hierarchicalcluster analysis technique,23 which tends to organizea data set into a set of the representative group, wasemployed. To visualize this approach graphically, as-sume that the L*a*b* values for the tooth data set aredistributed uniformly in a cubical subspace of L*a*b*space. To cluster such data intoM clusters (tabs), dividethe cube into M subcubes, and use the central point of

VOLUME 92 NUMBER 4

THE JOURNAL OF PROSTHETIC DENTISTRYANALOUI ET AL

each subcube as the representative color of all teethwithin that subcube (Fig. 1). In this example, M=27was used. While this example describes the clusteringprocedure, 2 major issues related to the subspace ofteeth must be considered. First, the set of colors repre-sented by teeth are not distributed uniformly in theL*a*b* space, and a simple geometrical shape (such asa cube) cannot encompass this set. Second, a simple uni-form division of tooth color space would not providea visually optimal design of clusters.

The data clustering approach was devised to groupitems within a data set (tooth color) such that items inthe same group (cluster or shade tab) aremore alike thanitems in a different group. The hierarchical clusteringmethod is a mathematical procedure for creating a se-quence of partitions within a data set. In this approach,the similarity (in terms of DE) between all tooth samplesin the population is computed. Then, the pairs ofsamples with closest color (smallest DE) are grouped in-to binary clusters. The process then links the newly for-med clusters, creating bigger clusters until all objects(teeth) in the original data set are linked together in a hi-

Fig. 4. Average DE for VOS (with 5 to 40 tabs) and 3commercially available shade guides evaluated.

Fig. 2. Color distribution of 150 extracted teeth.

OCTOBER 2004

erarchical tree. The process ultimately results in a singlesupercluster (root) that contains the entire population.

Once the hierarchical tree is computed, it remains tobe determined how many clusters (shade tabs) must beused for a given population. This is commonly doneby either (1) finding the natural divisions in the originaldata set or (2) specifying an arbitrary number ofclusters.20

MATERIAL AND METHODS

In this study, several clusters were computed to com-pare the results with 3 common shade guides. To assessthe proposed technique, the spectral characteristics of150 extracted maxillary human teeth were measuredwith a spectroradiometer. The tooth sample consistedof 35 central incisors, 56 lateral incisors, 32 canines,and 27 first premolars. Specimens were selected from

Fig. 5. Histogram for error distribution of 4 shade guides overall specimens.

Fig. 3. Chromatic (a*b*) components of extracted teeth and 3shade guides.

373

THE JOURNAL OF PROSTHETIC DENTISTRY ANALOUI ET AL

a pool provided by local dentists. There was no informa-tion available regarding the age of the teeth or thereasons for their extraction. The selected specimenswere caries and restoration free on their facial surfacesand did not exhibit significant hypocalcification, enamelwear, or staining on the surfaces to be measured. Teethwere kept in 10% formalin after extraction. They weredebrided of adherent tissues and soft and hard accretionswith periodontal curettes and an ultrasonic scalar, thenpolished with a low-speed handpiece, rubber cup, anda slurry of fine flour of pumice and water paste.Specimens were rinsed and stored in water at room tem-perature until they were mounted.

Table I. Error statistics for visually optimal shade guide with26 tabs (VOS26) compared to 3 commonly used shadeguides

Shade guide Mean DE SD DE

Percentage

of sites

with DE,2

Percentage

of sites

with DE,3.7

Vita Lumin 3.11 1.74 25 72

Trubyte 3.28 2.19 28 72

Vitapan 3D-Master 2.65 1.22 33 85

VOS26 1.97 0.87 60 95

Table III. Computed L*a*b* values for shade guide with 26tabs, using visually optimal shade guide approach. Thesevalues were computed for comparison with commercialshade guides at middle third area

Tab no. L* a* b*

1 74.99 4.25 24.70

2 76.89 4.08 18.05

3 91.31 1.72 31.03

4 90.38 1.47 28.17

5 83.60 3.66 20.57

6 84.27 1.06 22.42

7 81.88 3.61 29.07

8 80.76 2.82 26.70

9 86.20 2.75 29.68

10 87.37 2.48 32.33

11 86.80 1.66 26.82

12 84.38 2.15 26.55

13 93.03 21.69 14.86

14 90.26 20.94 17.55

15 80.15 5.11 34.15

16 76.09 5.62 30.91

17 90.69 2.18 20.87

18 90.48 20.07 23.23

19 89.99 1.02 26.03

20 93.29 1.57 25.78

21 98.31 21.70 14.50

22 95.63 0.60 18.44

23 83.47 2.79 31.69

24 87.21 0.92 24.22

25 87.66 0.51 19.73

26 94.41 20.59 22.78

374

The spectroradiometer was composed of a dual-grat-ing monochromator/spectrograph (ARC SpectraPro-150; Acton Research, Acton, Mass) and cooled CCD(charge-coupled device) detector (Model 1530-A; EGand G Instruments, Princeton, NJ). All specimens weremounted on a supporting holder to control their posi-tion with respect to the spectroradiometer. The speci-men holder and background were covered by neutral

Table II. Measured L*a*b* values at middle third for Vitapan3D-Master shade guide

Tab L* a* b*

1M1 97.31 20.46 15.25

1M2 95.02 20.43 20.71

2L1 92.00 0.25 21.13

2L2 91.80 0.47 27.30

2M1 94.64 0.81 16.08

2M2 91.89 1.23 22.47

2M3 90.89 1.27 27.33

2R1 92.74 1.94 18.55

2R2 91.43 2.20 25.64

3L1 83.22 1.73 22.22

3L2 88.37 2.17 28.70

3M1 86.57 2.04 16.94

3M2 87.00 2.29 23.57

3M3 87.12 3.24 30.44

3R1 87.62 3.17 19.75

3R2 88.53 3.65 28.15

4L1 80.99 3.16 23.03

4L2 84.12 4.37 30.29

4M1 84.74 3.36 18.75

4M2 83.78 4.16 23.33

4M3 82.58 5.29 31.36

4R1 82.76 4.83 22.53

4R2 82.28 5.76 27.74

5M1 79.04 4.69 20.55

5M2 75.84 6.45 27.43

5M3 75.79 8.52 35.00

Table IV. Computed L*a*b* values for shade guide with 12tabs, using visually optimal shade guide approach. Thesevalues were computed for comparison with commercialshades guides at middle third area

Tab no. L* a* b*

1 90.55 0.70 22.42

2 87.66 0.51 19.73

4 91.33 21.23 16.50

4 96.47 20.12 17.21

5 84.07 1.84 21.86

6 85.85 1.66 25.90

7 81.26 3.17 27.76

8 75.26 4.23 23.75

9 78.59 5.30 32.91

10 94.41 20.59 22.78

11 85.58 2.69 31.08

12 90.92 1.35 27.18

VOLUME 92 NUMBER 4

THE JOURNAL OF PROSTHETIC DENTISTRYANALOUI ET AL

gray, nonreflecting materials. For each specimen, 3measurements were performed at the midline,corresponding to the cervical, middle, and incisal thirdsof the facial surface. For each area, a reflectance spectrumbetween 400 and 700 nm at 10-nm intervals was mea-sured. All specimens were illuminated in an illuminationbox (SpectraLight II;Macbeth,NewWindsor, NY)withsimulated daylight (D65). After compensation for the il-luminant spectrum,22 CIE Yxy coordinates were calcu-lated with the CIE 1931 2-degree standard observercolor-matching function, and CIELAB colorcoordinates were computed.

The L*a*b* color coordinates of 3 commerciallyavailable and commonly used shade guides also weremeasured: the Vita Lumin V Shade Guide (VitaZahnfabrik); the Trubyte Bioform Color Order ShadeGuide and Porcelain System (Dentsply, York, Pa); andthe Vitapan 3D-Master Shade System (VitaZahnfabrik). Procedures identical to those used for ex-tracted teeth were used to position and measure theshade tables at the cervical, middle, and incisal thirdsof the facial surface. Specular reflection (gloss) was no-ticed in selected areas, mainly incisal, of some teeth.All such sites were excluded from the design and evalu-ation steps, because specular reflection does not repre-sent the spectral reflection of the tooth surface.

RESULTS

The color distribution of extracted teeth (all 3 sites)in L*a*b* color space is shown in Figure 2. Figure 3compares the chromatic (a*b*) components of the ex-tracted teeth to those of the 3 commercial shade guides.The chromatic span of the extracted teeth was largerthan all 3 guides. Using the volume described byL*a*b* values of extracted teeth as the total populationspace, the coverage (overlap) of this volume by each ofthe shade guides was computed. The coverage was37.45%, 30.96%, and 56.05% for Vita Lumin, Trubyte,and Vitapan 3D-Master, respectively. These results indi-cate that the Vitapan 3D-Master shade guide providedthe best color coverage for the extracted teeth.

Using the proposed method, a series of visually opti-mal shade guides (VOS) were designed with the numberof shade tabs ranging from 5 to 40. For each extractedtooth and a given commercial shade guide, the tab withthe smallest DE was determined. This process was re-peated for all 4 shade guides (VOS, Vita Lumin,Trubyte, and Vitapan 3D-Master), and an average DEfor each shade guide was computed. Figure 4 showsthe average DE for VOS (5-40 tabs) and the 3 commer-cial shade guides. Figure 5 shows the error distributionof the 4 shade guides over all specimens. The histogramfor VOSwith 26 tabs indicated that a significant numberof specimens had color differences in the lower range ofDE. Table I summarizes statistics for the data shown in

OCTOBER 2004

Figure 5. Tables II and III present the L*a*b* valuesfor the middle third of Vitapan 3D-Master and VOSwith 26 tabs, respectively. Table IV provides theL*a*b* values for VOS with 12 tabs.

DISCUSSION

As indicated by Figure 4, Vitapan 3D-Master demon-strated the lowest averageDE among the 3 commerciallyavailable shade guides examined in this study. VitaLumin, Trubyte, and Vitapan 3D-Master have 16, 24,and 26 tabs, respectively. The error associated withVOS decreased monotonically with the number of tabs.It is interesting to note that VOS with 12 tabs led to anaverage DE 2.61 compared to DE 2.65 for the Vitapan3D-Master shade guide with 26 tabs. VOS with 26 tabsresulted in an average DE 1.86 (Table IV).

Although psychophysical vision testing under con-trolled conditions has shown that a human observercan detectDE1.0,24 clinically perceivable and acceptablethresholds are much higher. While the American DentalAssociation has stated a limit of DE 2 as a tolerance forshade guides,25 DE 3.7 has been reported as the averagecolor difference between teeth and matched shade tabsin the oral environment.26 The percentage of sites inthe present study with a color difference below eitherthreshold was computed. An acceptable color matchfor 60% (DE 2) and 95% (DE 3.7) of measured siteswas found for VOS with 26 tabs, compared to 33%(DE 2) and 85% (DE 3.7) for Vitapan 3D-Master.

Using the hierarchical clustering approach, it is possi-ble to design a shade guide for a target average DE. Asthe target average DE decreases, the number of shadetabs will increase. The disadvantage of decreasing DEwould be the greater number of tabs within the shadeguide. For example, Figure 4 shows that when 40 tabswere used, the average DE for VOS dropped to 1.49.An increased number of tabs is commonly associatedwith increased complexity and required time for clinicalselection of the optimal tab. It must be noted that all er-ror calculations in this study, for VOS and the 3 com-mercially available shade guides, were based ona computer search for the best match. In this process,for a given tooth or specific site within a tooth, a com-puter found the tab with the smallest DE among all tabswithin a shade guide.

This study involved extracted teeth, which have dif-ferent spectral characteristics than nonextracted teeth.The average L*, a*, and b* values at the middle thirdof the teeth used in this study were 87.8, 1.5, and25.2, respectively. These values differ from reports ofthe color spectrum of teeth in the oral environment.For example, Hasegawa et al15 reported the averageCIELAB values of L*, a*, and b* for 87 subjects to be71.76, 5.5, and 17.59, respectively. For 7 subjects,Douglas17 reported that the mean L*, a*, and b* values

375

THE JOURNAL OF PROSTHETIC DENTISTRY ANALOUI ET AL

varied from 53.4 to 60.8, 20.72 to 1.44, and 1.92 to14.70, respectively. The color difference between ex-tracted and nonextracted teeth is due primarily to dif-ferences in pulp content and hydration, in addition todifferences in population size, age, and dietary regimens.Other difficulties associated with use of extracted teethfor color assessment are the inability to ascertain theage of teeth and, more importantly, the reason(s) fortheir extraction. Thus, in the present study, the errorpresented for the 3 commercial shade guides may notbe reflective of their performance in a clinical environ-ment.

Using the color spectrum of nonextracted teeth,a hierarchical clustering approach could be used to de-sign a new visually optimal shade guide for clinical appli-cation. Because this approach directly uses colorrepresented by the population under consideration, itmay be possible to design shade guides specificallydesigned for various populations. These customguides might span difference color subspaces, such asdifferences in race and age. Is would be important tohave a large data set to represent the underlyingpopulation accurately. Moreover, the hierarchical clus-tering tree could be redesigned so that other measuresof color matching could be optimized. For example, us-ing this approach, a shade guidewith amaximumDE lessthan a target value could be designed.

VOS tabs can be organized according to their light-ness and chromatic values to streamline the selectionprocess. It should be noted that VOS provides the colorcoordinates of an optimal shade guide; its clinical perfor-mance is a function of how accurately those colors areduplicated with available materials.

CONCLUSIONS

In this study, spectrophotometric colormeasurementwas used to present a new approach to designing a visu-ally optimal shade guide. Within the limitations of thisinvestigation, the following conclusions were drawn:

1. A hierarchical clustering approach can be used todesign a visually optimal shade guide (VOS).

2. Using the color space defined by 150 extracted hu-man teeth, VOS had a much lower error in the color-matching task than the 3 commercial shade guides eval-uated.

3. The proposed design procedure lends itself to thedevelopment of objective and population-specific shadeguides.

REFERENCES

1. McMaugh DR. A comparative analysis of the colour matching ability of

dentists, dental students, and ceramic technicians. Aust Dent J 1977;22:

165-7.

376

2. Leon JM. Shade selection—the art and science of color matching. Quin-

tessence Int 1982;13:851-9.

3. Miller LL. Shade matching. J Esthet Dent 1993;5:143-53.

4. Pensler AV. Shade selection: problems and solutions. Compend Contin

Educ Dent 1998;19:387-4. 396.

5. Ferreira D, Monard LA. Measurement of spectral reflectance and colori-

metric properties of Vita shade guides. J Dent Assoc S Afr 1991;46:63-5.

6. Horn DJ, Bulan-Brady J, Hicks ML. Sphere spectrophotometer versus hu-

man evaluation of tooth shade. J Endod 1998;24:786-90.

7. O’Brien WJ, Boenke KM, Groh CL. Coverage errors of two shade guides.

Int J Prosthodont 1991;4:45-50.

8. O’Keefe KL, Strickler ER, Kerrin HK. Color and shade matching: the weak

link in esthetic dentistry. Compendium 1990;11:116, 118-116, 120.

9. Yap AU, Bhole S, Tan KB. Shade match of tooth-colored restorative mate-

rials based on a commercial shade guide. Quintessence Int 1995;26:

697-702.

10. Bergen SF, McCasland J. Dental operatory lighting and tooth color dis-

crimination. J Am Dent Assoc 1977;94:130-4.

11. Finagin WB. Shade taking made easy. J Md State Dent Assoc 1973;16:

195.

12. Goodkind RJ, Keenan KM, Schwabacher WB. A comparison of Chromas-

can and spectrophotometric color measurements of 100 natural teeth. J

Prosthet Dent 1985;53:105-9.

13. Macentee M, Lakowski R. Instrumental colour measurement of vital and

extracted human teeth. J Oral Rehabil 1981;8:203-8.

14. Douglas RD. Precision of in vivo colorimetric assessments of teeth. J Pros-

thet Dent 1997;77:464-70.

15. Hasegawa A, Ikeda I, Kawaguchi S. Color and translucency of in vivo nat-

ural central incisors. J Prosthet Dent 2000;83:418-23.

16. Seghi RR, Johnston WM, O’Brien WJ. Performance assessment of colori-

metric devices on dental porcelains. J Dent Res 1989;68:1755-9.

17. Roberts CJ. Shade variation in dentistry. A photographic investigation.

Aust Dent J 1984;29:384-8.

18. Du-Yong Ng DY, Allebach JP, Pizlo Z, Analoui M. Non-contact imaging

colorimeter for human tooth color assessment using a digital camera. J Im-

aging Sci Tech 2003;47:531-42.

19. O’Brien WJ, Groh CL, Boenke KM. One-dimensional color order system

for dental shade guides. Dent Mater 1989;5:371-4.

20. Hall NR, Kafalias MC. Composite colour matching: the development and

evaluation of a restorative colour matching system. Aust Prosthodont J

1991;5:47-52.

21. Hunt RW. Measuring Colour. Est Sussex, England: Ellis Horwood Limited;

1987.

22. Wyszecki G, Stiles W. Color science. New York: Wiley Interscience;

1982.

23. Dubes RD. Cluster analysis and relarted issues. In: Chen CH, Pau LF,

Wang PS, editors. Handbook of pattern recognition and computer vision.

Singapore: World Scientific; 1993. p. 3-32.

24. Kuehni RG, Marcus RT. An experiment in visual scaling of small color dif-

ference. Color Res Appl 4, 83-91. 1979.

25. Wozniak WT. Proposed guideline for the acceptance program for dental

shade guides. Chicago: American Dental Association; 1987. p. 1-2.

26. Johnston WM, Kao EC. Assessment of appearance match by visual obser-

vation and clinical colorimetry. J Dent Res 1989;68:819-22.

Reprint requests to:

DR MOSTAFA ANALOUI

PFIZER GLOBAL RESEARCH AND DEVELOPMENT

EASTERN POINT RD

GROTON, CT 06340

FAX: 860-686-1680

E-MAIL: analouim@groton.pfizer.com

0022-3913/$30.00

Copyright � 2004 by The Editorial Council of The Journal of Prosthetic

Dentistry

doi:10.1016/j.prosdent.2004.06.028

VOLUME 92 NUMBER 4