Assessment of full frame and crop frame DSLR cameras for dental shade selection: in vitro comparative study

Introduction

In recent years, there has been a substantial improvement and interest growth in the sense of attractiveness and esthetics by both dental professionals and patients. It is mandated that dentists replace lost teeth with restorations, which closely resemble the patient’s original teeth [1, 2]. The literature document the advantages of instrumental approaches over visual methods for shade selection, such as spectrophotometers, intra- oral scanners, and digital cameras [3]. Regarding color matching in dentistry, spectrophotometers and spectro­radiometers are extremely dependable and precise instruments. In many comparative studies on shade selection, they were regularly used as reference devices. Nevertheless, these instruments are not usually available in standard dentistry practices due to their exorbitant cost and specialized application [4-6]. Since digital cameras are very accessible and affordable, and because of their tremendous advancements in producing, transferring, processing, and analyzing visual data, they have become indispensable tools in dental workflows for a wide range of tasks, from diagnosis to record keeping. They are therefore a good choice for lab communication and shade choosing [2]. In dentistry, DSLR (digital single lens reflex) cameras are among the most utilized types of camera systems. DSLR cameras are categorized by sensor size into two different groups: units with 35 mm or full frame FX sensors (36 × 24 mm), and a smaller crop frame (24 × 16 mm), commonly referred to as DX or APS-C format (advanced photography system, cropped). Bigger pixel size and better resolution with superior image quality of full frame sensors are their most evident advantages. Because there is more surface area accessible for light gathering, less amplification of each pixel’s output signal is required for image processing, meaning less digital noise is generated, which might compromise the accuracy of color and details [7, 8]. Compared with smaller crop frame cameras, full frame sensors may generate a broader dynamic range and richer colors courtesy to their larger pixels [7, 9, 10]. In tooth-color research, CIE-Lab color system is frequently employed to calculate and express color differences, created in 1976 by the International Commission on Illumination (Commission internationale de l’eclairage (CIE)). Lightness is represented by L*, red to green by a*, and yellow to blue by b* [11-13]. Disparities between L*, a*, and b* values of the standard or target color are denoted by color difference or ΔE, which is taken from the German term “Empfindung” for “feeling” [12, 13]. The formula for calculating ΔE is: ∆E = (∆L2 + ∆a2 + ∆b2)1/2 [6, 14]. In dental color research, the amount of color variation that is apparent and/ or acceptable to human observers is still not fully defined, nor is it optimally assessed [11, 14-16].

Objectives

This study examined the hypothesis that the spectro­radiometer (utilized as a reference) and digital photos from DSLR cameras with full frame and crop frame sensors would yield similar results in terms of shade selection accuracy.

Material and methods

Image and spectral radiance acquisition

In this study, six cameras were used, including two full frame (Nikon D610 [Nikon Corporation, Japan]; and Canon 5D MIII [Canon Corporation, Japan]), and four APS-C cameras (Nikon D7200 and Nikon APS-C D 5600 [Nikon Corporation, Japan]; and Canon 80D and Ca­non 750D [Canon Corporation, Japan]), along with two Tamron 90 mm macro-lens one with Nikon F mount and Canon EF mount. These photographic equipment was used to capture images of 26 shade tabs from a new VITA 3D-Master shade guide (1M1, 1M2, 2L1.5, 2L2.5, 2M1, 2M2, 2M3, 2R1.5, 2R2.5, 3L1.5, 3L2.5, 3M1, 3M2, 3M3, 3R1.5, 3R2.5, 4L1.5, 4L2.5, 4M1, 4M2, 4M3, 4R1.5, 4R2.5, 5M1, 5M2, 5M3). Cameras were secured into a tripod, and camera mode was set to manual with shutter speed of 1/200 sec, ISO of 100 F/stop of 22, as shown in Figure 1. Metering was set to average and picture control to neutral. Raw format was employed to save the images. Macro-lenses were set to manual focus and magnification ratio of ½ distance was adjusted until shade tab images were in sharp focus to the photographer, and the distance indicator in the viewfinder stopped blinking as the object being imaged was going to cover the same area on the sensor regarding the crop factor. Illumination consisted of two TTL Speedlights (Viltrox, China) mounted on their respective table stands, one speedlight connected to the camera through a wireless trigger (Godox Global, China) and the other set to slave mode to the first speedlight. A custom-made linear polarizing filter consisting of one lens polarizing filter screwed to the macro-lens (NiSi, China), and two circular polarizing filters (one attached to each speedlight by metal clips), were used to provide cross-polarized illumination that was devoid of white burned areas, reflection, and enhanced saturation, revealing the real multiple hues and translucent nature of shade tabs. The cross- polarizing filter effect was tested by rotating the filter attached to the lens, making 90° with the filters attached to the speedlights, and blocking any glare or no reflection from objects being imaged [17]. White balance was set to manual, and calibrated using a grey card with CIE-Lab values of L = 79, a = 0, and b = 0, imaged with each shade tabs to alleviate the impact of cross-polarization on color temperature [18]. The cameras with 2 speed lights were arranged to provide 0° angle of observation and an angle of 45° illumination to the object. The power of both speedlights were set for full frame cameras to ½ power, while for APS-C were set to ¼ using an external light meter (TTArtisan Light Meter, China). Each shade tab was secured into a custom-made cast to simu­late clinical conditions as possible. Each shade tab was imaged 3 times and later during image processing, and average value was used (Figure 2).
For spectral radiance measurement, the shade tabs were fastened to a drop in spring clamp with a distance of 8 cm away from the device lens, with 0°-degree observation angle. Shade tabs were lit with CIE standard illuminant D65 (300W, Newport Corp-Oriel Instruments, Stratford, Conn) attached to a fiber optic cord at 45° degrees to the specimen. The lens of spectroradiometer were focused to the middle third of each shade tabs. Readings were transferred through USB cable to personal computer. SpectraWin 2.3 software (Photo Research Inc.) was used to process data, and showed as CIE L*a*b* system coordinates. All reading were within the visible light spectrum (range, 380-780 nanometers).

Image analysis

In total, 468 digital images were acquired for 26 shade tabs transferred through SD card reader to a desktop computer, and assessed with Adobe Photoshop CC 22 (Adobe Systems Inc., San Jose, CA, USA) image software. Raw image profile was set to neutral camera to avoid software interference in original colors, and image mode was set to CIE-Lab. Eyedropper tool was used to select color values in CIE-Lab from the center of middle third of each shade taps (Figure 3). Readings were repeated 3 times, and the average reading was used. ∆E was measured from CIE-Lab values obtained from digital images with CIE-Lab control values measured with a non-contact spectroradiometer (SpectraScan PR 670, Photo Research Inc., and Chatsworth, CA, USA) using Microsoft Excel software (Microsoft, USA) function: ((L1 – L2)2 + (a1 – a2)2 + (b1 – b2)2)^(1/2).
Data were recorded in an Excel sheet, and analyzed using IBM SPSS Statistics 21 (SPSS, Chicago, IL, USA). Kappa measure of agreement was employed to evaluate a wither, and one-way ANOVA teat was utilized to evaluate statistically significant differences among 3 DSLR cameras regarding ∆E values. Post-hoc Tukey’s HSD test (p ≤ 0.05) was applied to assess significantly different means.

Results

Intra-class correlation (ICC) reliability test was performed for the reliability of CIE-Lab color coordinates determination using spectroradiometer and the six DSLR cameras. The ICC for L* was 0.9787, for a* = 0.9788, and for b* = 0.9910. Full frame cameras showed the lowest mean of ΔEab followed by crop frame units (Table 1). One-way ANOVA test showed that there were statistically significant differences among the means of ΔEab of the three tested cameras (p < 0.01) (Table 2). Tukey’s post-hoc test indicated highly statistically significant differences between FX cameras and APS-C DSLR cameras, and between semi-professional and entry level APS-C DSLRs (p < 0.01) (Table 2). The L*, a*, and b* parameters were evaluated statistically (Table 3). One-way ANOVA test demonstrated that means of L* and b* were not statistically different among the tested groups (p ≤ 0.05), while statistically significant differences (p < 0.01) were found in a* parameter (Table 4). Tukey’s post-hoc test showed highly statistically significant differences between spectroradiometer and tested cameras regarding the means of a* parameter of CIE-Lab system. Moreover, statistically significant differences were observed between full frame and crop frame, and between high- end and low-end crop frame cameras regarding a* parameter (Table 4).

Discussion

In this study, Vita 3D-Master shade guide was chosen because it is widely used in dental clinics and labs, generating reproducible results and a wider spectrum of shades [3]. While SpectraScan spectroradiometer (PR-705, Photo Research Inc., Chatsworth, California) has been utilized as a standard in multiple prior studies on shade selection [6, 25]. Due to its widespread use in scientific color research, CIE-Lab was selected as the testing color space [13, 14]. Extensive literature on dental shade is dedicated to the selection of thresholds for perceptibility (PT; the difference that is perceivable to an observer) and acceptability (AT; the color difference clinically acceptable to equate the outcomes) [2, 11, 14]. A wide array of color threshold values for perceptibility and acceptability have been proposed for human teeth and restorative materials. Acceptability threshold values are generally higher than perceptibility threshold values. Khashayar et al. [14] reported PT as ΔEab = 1 and AT of ΔEab = 3.7. Ghinea et al. [12] observed a PT of 1.8 ΔEab units and AT of 3.46 ΔEab units. Alghazali et al. [11] in clinical study showed a PT value of 1.9 ΔEab units and AT value of 4.2 ΔEab units. A border line of PT/AT was shown in several studies, and Johnston and Kao reported a PT of 3.7 and AT of 6.8 [2, 24]. Paravina et al. [13] in their comprehensive study on perceptibility and acceptability thresholds in dentistry proposed that for a tooth-colored materials, an excellent match could be achieved with ΔEab ≤ 1.2, while an acceptable match was reported for ΔE > 1.2 and ΔE ≤ 2.7. However, ΔEab > 2.7 and ΔEab ≤ 5.4 produced mode­rately unacceptable match, and higher ΔEab values resulted in clearly unacceptable match [11, 14-16].
The results of the current study were evaluated with perceptibility of ΔEab = 1.2, and acceptability of ΔEab = 2.7 color thresholds [12]. The null hypothesis was rejected, as the analysis of ΔEab values showed that full frame cameras tested in this study were more truthful to color replication, and produced ΔEab values closer to that of the spectroradiometer. While APS-C DSLR cameras examined in this study were less accurate, and generated ΔEab values distant from the control values.
This is the first study dedicated to evaluate the effect of camera sensor size upon color reliability and accuracy of dental shade selection. To the best of the authors’ knowledge, no study evaluated specifically this aspect, despite professional photographers’ recognition of the difference between full frame and crop frame cameras in the field of color accuracy [26]. From a clinical point of view, the results of this study could be related to the interpretation of color differences proposed by Paravina et al. [13], and based on the PT and AT threshold values described by a study done earlier (Paravina et al. [12]). Both of the full frame cameras (i.e., Nikon D610 and Canon 5D MIII) produced ΔEab means of 1.92 and 1.93, respectively, which is considered above PT but within AT values, showing an acceptable match. The semi-professional APS-C DSLR cameras (i.e., Nikon D7200 and Canon 80D) produced ΔEab means of 2.66 and 2.68, respectively, which is above AT, yielding moderately unacceptable match. The entry level APS-C DX cameras (i.e., Nikon D5600 and Canon 750D) demonstrated ΔEab means of 4.19 and 4.22, respectively, which is above AT, but showing a clearly unacceptable match. These results could be attributed to full frame cameras’ larger sensors and larger pixel area with higher signal-to-noise ratios and wider dynamic range, leading to better ability to capture more light, resulting in superior image quality and color accuracy [7-10]. More accurate white balance algorithm of full frame cameras [17, 24, 27] and higher color depth of FX sensor increase camera’s capability to produce additional colors and creating accurate image [7, 9]. No statistical differences were found between the spectroradiometer and the tested DSLR cameras regarding CIE L* and b* values. While CIE a* values showed statistically significant differences between the control and tested cameras. These differences can be related to the detectors impeded within the sensors of these cameras, the variation in response to red channels of the visible light spectrum that could be associated with the ability of red elements in color filters to transmit red light, and the design of the infra-red filter or in semi-conductors used [28, 29]. The authors’ of this study treated the images with curves adjustment in Adobe Photoshop software by reducing the CIE a* values, which greatly reduced the difference with the control values. However, this requires further investigation, and developing a suitable algorithm to be incorporated within camera firmware or to be developed as a separate addition to image processing software to improve dental images accuracy regarding shade selection.

Conclusions

In the current study, the tested full frame DSLR cameras demonstrated better accuracy in determining dental shade than crop frame cameras. Within the limi­tations of this study, it is highly advised to use full frame cameras for determining dental shade and conveying shade data, as it can improve the success rate of fabricated restorations. In addition to other methods for determining shade, digital photography can be successfully utilized as an adjunct mean of dental shade selection. Further research is needed evaluating DSLRs compared with newly introduced mirrorless cameras and with other camera formats, such as medium format cameras.

Disclosures

1. Institutional review board statement: Not applicable.
2. Assistance with the article: None.
3. Financial support and sponsorship: None.
4. Conflicts of interest: The authors declare no potential conflicts of interest concerning the research, authorship, and/or publication of this article.

References