New portable system for dental plaque measurement using a digital single-lens reflex camera and image analysis: Study of reliability and validation.

BACKGROUND: The quantification of the dental plaque (DP) by indices has limitations: They depend on the subjective operator's evaluation and are measured in an ordinal scale. The purpose of this study was to develop and evaluate a method to measure DP in a proportional scale.
MATERIALS AND METHODS: A portable photographic positioning device (PPPD) was designed and added to a photographic digital single-lens reflex camera. Seventeen subjects participated in this study, after DP disclosure with the erythrosine, their incisors, and a calibration scale ware photographed by two operators in duplicate, re-positioning the PPPD among each acquisition. A third operator registered the Quigley-Hein modified by Turesky DP index (Q-H/TPI). After tooth brushing, the same operators repeated the photographs and the Q-H/TPI. The image analysis system (IAS) technique allowed the measurement in mm(2) of the vestibular total tooth area and the area with DP.
RESULTS: The reliability was determined with the intra-class correlation coefficient that was 0.9936 (P < 0.05) for the intra-operator repeatability and 0.9931 (P < 0.05) for inter-operator reproducibility. The validity was assessed using the Spearman's correlation coefficient that indicated a strong positive correlation with the Q-H/TPI r s = 0.84 (P < 0.01). The sensitivity of the IAS was evaluated with two sample sizes, only the IAS was able to detect significant differences (P < 0.05) with the sample of smaller size (n = 8).
CONCLUSIONS: Image analysis system showed to be a reliable and valid method to measure the quantity of DP in a proportional scale, allowing a more powerful statistical analysis, thus facilitating trials with a smaller sample size.

INTRODUCTION

The quantification of dental plaque (DP) in clinical trials has been performed for many years using different indices of DP. While these scores dramatically contributed in several areas of research, have some limitations: They rely on the subjective assessment of a clinical operator and they are measured on an ordinal scale.[123] Several authors proposed other methods such as: The delimitation by sight of colored areas plaque,[4] the use of grids and planimetric systems that depend on general observation of an operator,[1] while others include complex apparatus and techniques using fluorescent light.[3] As long as oral hygiene systems improve their effectiveness, it is necessary to have tools with increasingly accurate and sensitive to measure the amount of plaque on the surface of the teeth.[2]

There are several intra-oral imaging systems based on charge-coupled device and complementary metal-oxide-semiconductor sensors, but these systems do not achieve the quality and image resolution required for accurate calibration and measurement. On the contrary, this can be achieved with digital single-lens reflex (DSLR) cameras 35 mm format and a macro lens system.[1]

The purposes of this study were to develop and to evaluate a method to measure the quantity of DP deposited on dental surfaces in a proportional measurement scale (of high level) by using digital intra-oral photographs; and to validate the new system comparing it with an established DP index.

MATERIALS AND METHODS

Experimental subjects

The experimental subjects were randomly selected among patients attending for routine dental consultation at Graduate School Foundation, Resistencia, Chaco, Argentina, who were invited to join this study. Seventeen subjects with a mean age of 33 years old participated, in which 136 incisors were evaluated. Two operators took 85 images and made 680 measurements with the image analysis system (IAS), a third operator registered 272 times, Quigley-Hein modified by Turesky DP index (Q-H/TPI).[56] This research project was approved by the Research and Human Right Committee at Graduate School Foundation, Resistencia, Chaco, Argentina. It was conducted in accordance with the Helsinki Declaration of 1975, as revised in 2000. Every step of the study was explained to all the subjects, and it was clearly stated that their participation was voluntary. Written informed consent was obtained from all the subjects to participate and to have their photographs examined as part of the study.

Experimental design

After disclosing DP, the incisors of each subject were photographed in duplicate by two operators (operators A and B) independently and alternately, re-positioning between each acquisition a custom made device (described later) attached to a photographic camera. A third operator (operator C) registered the Q-H/TPI. Then, each subject performed manual tooth brushing unsupervised for a minute with his personal toothbrush and with his usual technique without receiving instructions. After that, the same operators repeated the photographs and plaque index record. The images of each operator were analyzed independently by each operator.

Photographic technique

In order to obtain high quality digital images in a standardized and reproducible way, a portable photographic positioning device (PPPD) [Figure 1] was designed and built by one of the authors (Dr. G.M. R.). For high stiffness and low weight (420 g), it was constructed of tubular aluminum. At one end, two areas of support were placed, so that tripod support was achieved on the patient's head, two points on the forehead and one on the chin. The opposite end allowed setting the standard tripod mount for a 35 mm camera and a system for focal length adjustment. This device allowed maintaining a fixed focal length and geometry of the image in the three planes of space. It's low weight design also allowed for use in portable way for trials outside the lab. For this study, a DSLR camera (Canon EOS D 300, Canon Inc., Ōta, Tokyo, Japan) with a high quality 100 mm macro lens (Canon EF-100 mm f 2.8 Macro, Canon Inc., Ōta, Tokyo, Japan) and a macro ring flash (Vivitar, Macroflash 5000, Vivitar Corporation, Oxnard, California, USA) were used. The following exposure parameters were set, sensitivity: ISO 200; shutter speed: 1/200 s; aperture: F 18; white balance: Custom setup calibrated with a 18% gray card (Kodak Grey Cards Cat N° E152 7795, Eastman Kodak Co., Rochester, New York), focus: Automatic.

Figure 1

The portable photographic positioning device attached to a digital single-lens reflex camera. Front view (a) and in use taking a photograph (b)

A stainless steel intra-oral sterilizable calibration scale was designed and built; it included a side bite bar covered with disposable rubber tubes.[1] A 5-mm-scale was fixed by a system of tubes, which allowed the anteroposterior movement of the scale to be placed on the same focal plane of tooth surfaces to be photographed.

Before each picture was taken, DP was disclosed with 2.7% erythrosine carefully applied with saturated cotton pellets and then asked the patient to rinse for 30 s.[7] Then, photographic lip retractors were placed upside down. PPPD placement was performed in a standardized way, and fine positioning of the camera was performed using the autofocus boxes of camera's viewfinder as a guide.

Digital image analysis

The image obtained with this system included central incisors with stained DP and intra-oral calibration scale located in the same focal plane. The images obtained were 3072 × 2048 pixels (6.3 megapixels) of spatial resolution and 24-bit of contrast resolution. They were stored on optical DVD of 4.7 GB capacity as compressed JPG files without quality loss. For the image analysis, a system consisting a computer and an input device, a pressure sensitive graphic tablet (5 × 4 Wizardpen Genius, Genius Inc., Taipéi, Taiwan) was used.

The IAS measurement technique involved two phases that can be summarized in Figure 2:

Figure 2

Scheme of the sequence of the two phases, segmentation and measurement, of the image analysis technique of the image analysis system. It can be seen as an example the examination of the image of tooth number 11

Phase 1: The images were visualized on a 15-inch monitor with a magnification of ×20. A photographic editing software (Adobe Photoshop 7.0, Adobe Systems Inc., San Jose, California, USA) was used for the segmentation of each tooth by its outline, using a pressure sensitive graphic tablet (5 × 4 Wizardpen Genius, Genius Inc., Taipéi, Taiwan) and the semiautomatic outline selection tool. In that way, each tooth was isolated from the rest of the image. After that, the areas with DP stained in red were automatically detected with the color range command, set with a range of 130, and then separated from the rest of the image

Phase 2: Using an image processing and analysis program (UTHSCSA ImageTool 3.0 program [developed at the University of Texas Health Science Center at San Antonio, Texas and available from the Internet by anonymous FTP from ftp://maxrad6.uthscsa.edu]), which was calibrated with the intra-oral scale in each image, the entire visible tooth area was automatically measured in mm2. From the images of isolated DP and applying a threshold of 0–190 shades of gray, the DP area in each tooth was automatically measured in mm2.

With these values, the percentage of area with DP in relation to the total area of the tooth[3] was automatically calculated.

Statistical analysis

To assess the reliability of the measurements, paired t-test, intra-class correlation coefficient (ICC) of reliability and Pearson correlation coefficient were used; also scatter plots were drawn. The validity of the measurements was assessed by the Spearman's rank correlation coefficient. To compare the two methods of DP measurement, a paired t-test and Wilcoxon test were performed taken the experimental subject as a statistical unit of analysis. The level of statistical significance was established as P < 0.05. The data were analyzed using statistical software (SPSS Statistics, SPSS Inc., Chicago, USA).

RESULTS

Reliability of measurements

Table 1 shows the P values of the paired t-test, the values of ICC and Pearson's correlation coefficient for the three variables measured by IAS in duplicate by two operators. No statistical differences were found by the paired t-test. As can be seen, all values of ICC ranged from 0.9916 to 0.9960 (that corresponds to an excellent reliability range between 0.81 and 1.00), being within a confidence interval of 95% (P < 0.05). Table 1 also shows the values of the Pearson's correlation coefficient found between measurements of each operator and operators. As can be seen, a very high positive statistically significant (P < 0.01) correlation in the three variables for both intra-operator and inter-operator was found.

Table 1

Reliability of measurements calculated for the three variables measured by image analysis in duplicate by two operators

Figure 3 shows a scatter plot diagram of the variable area with DP (%) conducted among the registered values by the operators A and B. The high reproducibility of the new IAS can be seen. The Pearson's correlation coefficient was 0.993 (P < 0.01), and the calculated regression line was also plotted.

Figure 3

Scatterplot diagram of the variable area with dental plaque (%) conducted among the registered values by the operators A and operator B

Validation

In order to evaluate the validity of the measurements of IAS, the Spearman correlation coefficient was calculated between the Q-H/TPI registered by operator C and the two variables registered with the IAS independently by operators A and B. The Spearman's rank correlation coefficient for the variable area with DP in mm2 was 0.811 and 0.807 (P < 0.01) for operators A and B, respectively. The Spearman's rank correlation coefficient for the variable area with DP in % was 0.849 and 0.843 (P < 0.01) for operators A and B, respectively. Higher values of Spearman's rank correlation coefficient were found for the variable percentage of plaque area which would indicate that this variable had a greater similarity to the quantification of DP by Q-H/TPI. In all cases, a strong positive correlation was found between the Q-H/TPI and the DP measurements by IAS.

Figure 4 shows a scatterplot made between the values of Q-H/TPI and the variable area with DP in % recorded with the new IAS by two operators independently. The strong correlation between the two measuring instruments can be observed. Furthermore, the excellent reproducibility between operators of IAS can be observed by noticing the superposition of independent records of the operator A in blue and B in red.

Figure 4

Scatterplot made between the values of Quigley-Hein plaque index modified by Turesky and the variable area with dental plaque in % recorded with the new image analysis system by two operators independently. The records of operator A were drawn in blue and operator B in red

In Table 2, the two methods of DP measuring are compared in terms of power to detect significant changes in the recorded amount of DP in the pre versus postbrushing moments. The IAS was measured on an interval scale and paired t-test was used. The plaque index method was measured on an ordinal scale, and both a paired t-test and the Wilcoxon test were used. The P values found by performing the respective hypotheses tests are presented for two sample sizes. As can be seen with a sample of 17 subjects, both methods showed significant differences, but with a smaller sample size of 8 subjects, only the IAS was able to detect significant differences. This indicates the greater sensitivity and discriminative power to detect changes in the quantity of DP by IAS method compared to the Q-H/TPI.

Table 2

Comparison between the two DP measurement methods and the statistics analysis recommended in each case. The hypothesis tests were performed with the data obtained from pre- and post-brushing records

DISCUSSION

The presented results showed that the PPPD with the IAS is a reliable, highly reproducible, and valid method to measure the quantity of DP on the teeth surfaces. The reliability of IAS was r = 0.99 for both intra-operator repeatability and the inter-operator reproducibility, being these values higher than published data for DP indexes that reported r = 0.44–0.89 for intra-operator repeatability, and r = 0.15–0.45 for inter-operator reproducibility.[8] By comparing the accumulated DP records between the IAS and a well-established and widely used DP index, the presented data clearly validated the measurement of the tooth surface covered with DP by the new proposed IAS.

The new IAS evaluated has some advantages over methods that use indexes: IAS does not depend on the observation and subjective interpretation of an operator, the registered variables are measured on a proportional scale of measurement, thus that allowing the analysis of data by parametric statistical tools that are generally more powerful.[29] Methods using DP indices measured on an ordinal scale, the mean values of the samples can only be compared by parametric hypothesis tests as long as they have a normal distribution.[210] The greater sensitivity and discriminative power to detect changes in the quantity of DP by IAS method compared to the DP indexes would reduce the number of subjects in clinical trials thus helping reduce the cost of the clinical trials. With the IAS, the clinical time is reduced compared to traditional indices, as it takes a few minutes to prepare the patient's mouth and take the pictures. The stored images can be analyzed without the presence of the patient, simplifying the double-blind studies.[2] This also enables the creation of databases for later re-use as well as sending electronically the images to facilitate its analysis during multicenter studies.

Both measurements of the tooth area and plaque area demonstrated improved reliability and reproducibility than a published study conducted with a more complex cephalometric positioning apparatus.[1] Possibly due to the use of a higher resolution DSLR camera with a high quality 100 mm macro lens, which allowed to obtain images of very high resolution that can be displayed with 20-fold magnification on a monitor and so that it allowed seeing greater detail. Previous work of Smith et al.[11] had probed that a DSLR camera permitted greater reliability of DP surface area measurements than recordings on images captured by intraoral cameras. In addition, in the present work, the use of a pressure sensitive graphics tablet with high precision semiautomatic selection tool allowed the segmentation of the contour of the teeth from the images. This simplified the delineation in cervical areas which were reported in the literature as more difficult areas to define.[1]

Carter et al.[12] reported a lower reproducibility and a low relationship between plaque area measurements (by an image analysis technique) and a DP index. However, the photographic technique for image acquisition presented in that paper had only the camera to object distance as spatial reference, lacking of spatial standardization during image acquisition (for example with head's fixed points of reference). The PPPD used in the present work allowed to maintain a fixed focal length and standardize the geometry of the image acquisition in the three planes of space (by tripod support on the patient's head) thus allowing a highly reproducible image acquisition procedure. In the study reported by Carter et al.,[12] the camera white balance was set to auto but since the software automatically detects the DP by analyzing the color values of the pixels, for better color consistency and reproducibility, it would be recommended that white balance was set to custom and calibrated with an 18% gray card as was done in the present work.

Recently, Klukowska et al.[13] published data showing the reproducibility of a digital plaque image analysis to quantitatively assess oral hygiene in orthodontic patients with multi-bracket appliances. However, the hardware used in that study was quite different than the one used in the present work. The camera system despite having a lower resolution was physically bigger needs a bulky power pack and the system need a computer linked to camera for storing the images, so that it was less practical to be portable. Klukowska et al.[13] used ultraviolet (UV) light as a light source, that require less common and more expensive flash units; on the other hand our device used light of visible wavelength (a photographic macro ring flash) that is a more common, cheaper, and healthy safer. Another complication of the use of UV light of high intensity in a clinical scenario is that it would require special protection eyewear for patients and operator. The reproducibility of the measurements in that study was a slightly lower (correlation = 0.97) than the data obtained in the present work (correlation = 0.99) by re-positioning the acquisition device by two operators independently and alternately. This difference may be explained in part because in the camera positioning system proposed by Klukowska et al.[13] the fixation on patient's head has only a chin rest (one point), in contrast, the PPPD proposed in the present work has tree points of rest on the patient's head (one in the chin and two in foreheads).

In our study, we positioned an intra-oral calibration scale in order to spatially calibrate the images as was previously proposed by Smith et al.[1] This allows to register the DP and tooth contour area in mm2 and not only as a proportion (%) between area with DP versus area free of DP as was proposed by others papers.[31213] Having the information of the absolute measurement of DP in mm2 can be useful for some trials of the effectiveness of antimicrobial agents or electric toothbrushes. Another advantage of the technique proposed in the present paper is that since each tooth was isolated in each image, the data could be analyzed by tooth type or dental arch to obtain additional information. On the contrary, other works measured the DP area of teeth together, and this information could not be obtained.[1213]

Previous work has proposed as an alternative to indices of DP, the use of complex, not portable and high-cost apparatus.[1313] On the other hand, the new proposed PPPD and IAS include photographic equipment (a DSLR camera, a 100 mm macro lens and a macro ring flash) that produce high quality images with no geometric distortion (a key point to perform area measurement) and are very common nowadays at laboratories and dental offices. The cost of this equipment and proposed software (one is very common in personal computers, and the other is a public domain program) is quite low thus allowing the use of this technique in trials in the developing countries with a limited budget. Another advantage of the PPPD is that as being very portable, it allows for the possibility to use it for work outside the laboratory.

As a disadvantage of this system in relation to DP indexes, the limitation of its application to buccal surfaces of anterior teeth can be mentioned. By minor modifications of the PPPD, it could be prepared for using it in the posterior region of the oral cavity as well as for the lingual surfaces using photographic mirrors.

CONCLUSIONS

The evaluated new PPPD and IAS allow objective measurement of DP in a proportional measurement scale with excellent reliability and reproducibility of the measurements. The IAS demonstrated a greater sensibility, allowing more powerful statistical tests and a higher discriminative power, permitting clinical trials with a smaller sample size. The greater sensitivity of this tool over the DP index allows greater discrimination power to assess oral hygiene measures such as electric toothbrushes or antimicrobial agents.