The in vitroand in vivoreproducibility of a video-based digital imaging system for tooth colour measurement

Abstract

Objectives

To assess the robustness of a new custom built video-based digital imaging system (VDIS) for measuring tooth colour and whiteness under in vitro and in vivo conditions.

Methods

The VDIS imaging system was developed for tooth colour measurement and evaluated in vitro and in vivo . The in vitro validation used extracted human teeth (HT, n = 14) stored in water and VITA Classical shade guide tabs (SG, n = 16). These were measured by the VDIS at baseline, 5 min, 2 h, 1 week and 2 weeks to evaluate the system repeatability. For in vivo validation, adult volunteers (male/female, n = 34) with two natural, unrestored central incisors had their teeth imaged using the VDIS at baseline, 5 min and 2 h (3 images each) by two different operators to evaluate time and operator effects. Between taking individual images, subjects moved from the imaging-frame to assess the effect of re-positioning on reproducibility. From the in vitro and in vivo images, the average tooth RGB values were obtained, and the CIELAB values and a tooth whiteness index WIO value were calculated. Repeatability and reproducibility of VDIS imaging system was assessed using appropriate repeated measurement analysis techniques and ANOVA.

Results

The measurement variations in vitro were between 1 and 2 units of ΔWIO and the average colour differences were less than 1 ΔE * ab unit. For the in vivo study, analysis of the CIELAB parameters and WIO showed that subject variability accounted for between 82 and 99% of the observed variability in the measurement process. The operator variability was less than 0.5% and the overall measurement error was found to be only 0.3% for WIO. Across assessment times the variability was less than 0.5%.

Conclusions

The dental imaging system V-DIS was shown to be a highly reproducible means for tooth colour and whiteness measurement.

Clinical significance

Digital imaging based techniques gives a highly reproducible approach to measuring tooth colour.

Introduction

Tooth colour is important to patients and consumers who wish to enhance their smile and also to professionals who want to match tooth colour for aesthetic restorations and whitening procedures . The colour of teeth is influenced by a combined effect of their intrinsic and extrinsic colourations , and is frequently quantified by visual assessment using a commercial tooth shade guide, or more objectively, by colour measuring instruments . There are a number of instruments that have been used for measuring tooth colour in vitro and in vivo , including colorimeters, spectrophotometers, spectroradiometers and digital cameras . Colorimeters and spectrophotometers have been shown to be reliable, have good repeatability and are accurate for colour matching . However, since they are contact-measurement devices, measurement errors may occur due to factors such as the curvature of the tooth surface, light loss caused by tooth translucency , ambient light and fogging of the optical lens during in vivo measurement .

Non-contact colour measurement systems, such as spectroradiometers and digital cameras, which use external light sources and do not need to directly attach apertures onto the tooth surface , may minimise the systematic error due to translucency and surface curvature . From comparison studies between digital imaging and contact-measurement methods, both spectrophotometric and digital image methods presented sufficient and validated objective evaluation of tooth bleaching efficacy . In another study, it was found that digital camera imaging is reliable in tooth colour quantification, whereas spectrophotometry (colorimetry) gave inaccurate absolute values for tooth colours but gave the same ranking order as the digital-imaging method . A meta-analysis of tooth whitening studies over a 4-year time frame confirmed the suitability of the approach and reliability of digital image analysis for long-term tooth whitening studies . In addition, digital imaging gives further advantages by providing a permanent database of images that can be analysed and re-investigated at a later date; it is relatively quick and simple in terms of training and operation, and does not require a clinician .

Dental-imaging systems for tooth colour measurement usually consist of a digital camera and a light source as the two key elements. Commercial single-lens reflex (SLR) cameras and industrial cameras (e.g. 3CCD cameras) have been used as the image-acquisition devices. Dual daylight lamps , halogen lamps with UV fluorescent tubes and ring light sources have been used as the illuminant for taking tooth images in vivo . The captured images are usually analysed by converting the camera RGB values (device-dependent colour space) into CIE XYZ or CIELAB values (device-independent colour spaces.) for tooth colour measurement. The three-dimensional colour coordinates are transformed into a single scale whiteness index in some tooth whitening studies , e.g. the tooth whiteness index (WIO) that was proposed based on the CIE whiteness index specifically for quantifying tooth whiteness perception .

The objectives of this paper is to evaluate the reproducibility of a new custom built video-based digital imaging system (VDIS) for measuring tooth colour and whiteness under in vitro and in vivo conditions.

Materials and methods

System development

A video-based digital imaging system (VDIS) has been developed for measuring tooth colour in vitro and in vivo . The key elements of the hardware are a digital video camera, a polarised and diffused white LED light source and a custom-built system frame. A digital camera (QImaging, Canada) provides high-speed live video during measurements and can capture still images. The camera has a cooling system to help to maintain a constant operating temperature and minimise thermal noise. A ring light (CCS Inc, Japan) is mounted on the camera lens and a diffusion filter (CCS Inc, Japan) is attached to the light source to provide diffused uniform illumination. Two polarising filters (CCS Inc, Japan) are placed, one in front of the camera lens and one in front of the light source, to provide cross-polarisation for excluding specular reflection in the teeth images. The system is connected via a USB connector to a laptop computer (Dell Inc., USA) from where the camera and imaging procedures are operated. A custom-built system frame was made to hold the camera/lighting set with adjustable distance to the teeth of the subject. A subject chin holder and a forehead bar were made to hold the subject’s head, and a white ceramic tile (Mt. Baker Research L.L.C., USA) is attached to the chin holder to enable monitoring of the lighting variation. The VDIS system is designed to disassemble into easily transportable pieces, and has been fully engineered to meet the requirements of the European Union Declaration of Conformity for safety under the Laboratory Directives, with fully Conformité Européene (CE) safety marking.

The image analysis software was written in Matlab (MathWorks Inc., USA). The core algorithm of the software is the camera characterisationmodel for predicting CIE XYZ values (and CIELAB values) from the camera RGB values. A polynomial regression model was used for this conversion . A Digitizer colourchecker chart (VeriVide, UK) was used as the standard reference to build the model. It contains 240 patches in a 12 cm x 20 cm grid. The colours on this chart give a good coverage of colour space which allows the characterisation model of the camera to suitably predict any colours inside of this range. The CIE XYZ values of each colour patch under D65 illuminant and 2° standard observer provided with the chart were considered as the ‘true’ values. The transform between the camera RGB and XYZ values can be expressed as:

<SPAN role=presentation tabIndex=0 id=MathJax-Element-1-Frame class=MathJax style="POSITION: relative" data-mathml='X=AV’>X=AVX=AV
X = A V

X represents the XYZ matrix, A is the transform matrix and V is the RGB matrix. For different order polynomials, the transform matrix A is a different size. Consider the 2nd-order polynomial model as an example, the polynomial transform equations are extended as below.

<SPAN role=presentation tabIndex=0 id=MathJax-Element-2-Frame class=MathJax style="POSITION: relative" data-mathml='X=a11R+a12G+a13B+a14RG+a15RB+a16GB+a17R2+a18G2+a19B2Y=a21R+a22G+a23B+a24RG+a25RB+a26GB+a27R2+a28G2+a29B2Z=a31R+a32G+a33B+a34RG+a35RB+a36GB+a37R2+a38G2+a39B2′>X=a11R+a12G+a13B+a14RG+a15RB+a16GB+a17R2+a18G2+a19B2Y=a21R+a22G+a23B+a24RG+a25RB+a26GB+a27R2+a28G2+a29B2Z=a31R+a32G+a33B+a34RG+a35RB+a36GB+a37R2+a38G2+a39B2X=a11R+a12G+a13B+a14RG+a15RB+a16GB+a17R2+a18G2+a19B2Y=a21R+a22G+a23B+a24RG+a25RB+a26GB+a27R2+a28G2+a29B2Z=a31R+a32G+a33B+a34RG+a35RB+a36GB+a37R2+a38G2+a39B2
X = a 11 R + a 12 G + a 13 B + a 14 R G + a 15 R B + a 16 G B + a 17 R 2 + a 18 G 2 + a 19 B 2 Y = a 21 R + a 22 G + a 23 B + a 24 R G + a 25 R B + a 26 G B + a 27 R 2 + a 28 G 2 + a 29 B 2 Z = a 31 R + a 32 G + a 33 B + a 34 R G + a 35 R B + a 36 G B + a 37 R 2 + a 38 G 2 + a 39 B 2

The best-fit regression should minimise the sum of residual square error, then the matrix A can be calculated by Eq. (3) , V’ is the transpose of the matrix V.

<SPAN role=presentation tabIndex=0 id=MathJax-Element-3-Frame class=MathJax style="POSITION: relative" data-mathml="A=XV'(VV’)−1″>A=XV(VV)1A=XV'(VV’)−1
A = X V ‘ ( V V ‘ ) − 1

In general, different types of camera have different colour rendering characteristics, so that the polynomial transform suitable for one camera may not fit other cameras. Several orders of polynomial should be tested to find the best polynomial transform matrix for a certain camera . In this study, three polynomial transforms (1st-order, 2nd-order and 3rd-order) were tested to find the best mapping between RGB values and CIE XYZ values. Considering the overall colour-rendering ability, the 2nd −order polynomial regression based on the colourchecker chart was chosen for the characterisation model of the V-DIS. Then XYZ values were converted into CIELAB values by Eq. (4) , and tooth whiteness index (WIO) values were calculated by Eq. (5) .

<SPAN role=presentation tabIndex=0 id=MathJax-Element-4-Frame class=MathJax style="POSITION: relative" data-mathml='L*=116(Y/Yn)1/3−16a*=500[f(X/Xn)−f(Y/Yn)]b*=200[f(Y/Yn)−f(Z/Zn)]’>L*=116(Y/Yn)1/316a*=500[f(X/Xn)f(Y/Yn)]b*=200[f(Y/Yn)f(Z/Zn)]L*=116(Y/Yn)1/3−16a*=500[f(X/Xn)−f(Y/Yn)]b*=200[f(Y/Yn)−f(Z/Zn)]
L * = 116 ( Y / Y n ) 1 / 3 − 16 a * = 500 [ f ( X / X n ) − f ( Y / Y n ) ] b * = 200 [ f ( Y / Y n ) − f ( Z / Z n ) ]
Only gold members can continue reading. Log In or Register to continue

Jun 17, 2018 | Posted by in General Dentistry | Comments Off on The in vitroand in vivoreproducibility of a video-based digital imaging system for tooth colour measurement
Premium Wordpress Themes by UFO Themes