Teeth Fluorescence
Melanie Ren
Introduction
Reflectance and fluorescence both affect how we perceive the appearance of an object when light falls upon it. Reflectance is the absorption of visible light (400-700nm) and spontaneous emission in the same wavelength region. People usually see objects based on the light that is reflected from the surface of the object. Fluorescence on the other hand, is the absorption of UV light (<=400nm) and spontaneous emission in a longer wavelength region such as 430-450nm. This "shift" in wavelength is known as the Stokes shift. Human teeth exhibit a property known as autofluorescence, which is natural fluorescence by biological structures without the presence of any artificial fluorophores. This fluorescent property gives the teeth a look of vitality that we have grown accustomed to seeing. In the dentistry setting especially, it is important to understand more about teeth fluorescence to be able to properly "patch up" natural teeth with artificial components.
Background
Cavities and other forms of teeth damage and breakdown are common in our lives but people place high value in a perfect smile. Dentists use a variety of artificial materials to amend damaged teeth but must take into account many factors in order to provide the best restoration for their patient. The composites need to appear to be the same color under various lightings (as was researched by a previous project) and also have similar fluorescence properties. These two phenomena are uncorrelated so the study discussed here is focused on finding the location of and peak intensities of teeth fluorescence so that we can get a better sense of what kind of materials to consider when picking a composite, since they must fluoresce in a similar manner with natural teeth. This is especially relevant in recent years since more artificial lighting with blue or UV components are now in use.
Some previous studies have been done on teeth fluorescence as well but most are studies in vitro, while ours is in vivo. In addition, most previous studies are performed with teeth that are heavily damaged whereas our study focuses on the average subject with generally healthy teeth. Previous studies also suggest that teeth fluorescence varies with age and that older teeth fluoresce at a longer wavelength (red-shift) and fluorescence peaks at the age of 26.5 and slowly declines thereafter. To see if we can validate this observation, we recorded the subjects' age information as well to see if there are any age-related trends.
Methods
For this study a Point Grey Camera (see image) was set up with two panels of LED lights. A total of 14 different LEDs (range 365nm - 940nm) were flashed onto the target surface in sequence and then captured through 8 different monochrome filters that limited the wavelength of the light detected by the camera. The first filter is a broadband filter that lets through the entire range of the LED wavelengths. Filters 2 through 8 are ranged between 437nm - 813nm.
We began each image capture session with white chalk as our target for calibration purposes and used the acquired images to normalize for the camera gain. After the white calibration, we captured images of teeth from 10 volunteers and then performed analysis using Matlab.
For our analysis we looked at the average intensity of the pixels after processing the image to account for camera gain, and filter and LED sensitivity differences. The processed image were displayed as a visual Donaldson Matrix where the rows correspond to the filter wavelengths (emissions) and columns correspond to LED channel wavelengths (excitation/absorption). The diagonal in the diagram is where the reflectances are and the region left of the diagonal is where the expected fluorescence signals will be. In the case where there is no fluorescence, the images will appear completely black except in the broadband filter and in the reflectance diagonal.
LED Channels/Light Absorbed
365nm 395nm 447nm 470nm 505nm 530nm 590nm 627nm 655nm 680nm 780nm 850nm 880nm 940nm
Filters/Light Emitted
broadband, 437 to 463nm, 487 to 513nm, 537 to 563nm, 587 to 613nm, 637 to 663nm, 687 to 713nm, 787 to 813nm
Results
Based on inspection of our visual Donaldson Matrix for all subjects, we decided to pursue data in the first 4 LED channels (365nm - 470nm) in the UV and blue range.
Fluorescence by Subject
This leads to the following analysis below:
We plotted the average intensity of the pixels for the selected patch (each image) as a function of the emission wavelength and excitation LED. Here for subject 7, each curve represents one input LED channel (column on the matrix image above), and the average pixel intensity at each of the rows within that channel (x-axis of plot). It can be seen that the orange band representing illumination at 395nm produces the highest pixel intensities. The peak emission wavelength is also seen to be the first range of wavelengths 437-463nm. Note that even though we used four of the LED channels in this portion of the analysis, absorption in fluorescence is generally limited to the UV range (the first two channels).
We then did this plot for all subjects and the trend was generally the same.
Fluorescence by Age
One interesting thing to note from the previous figure is that with older subjects (age 50+), the second point on the orange line is at around the same or higher than the intensity for the first point, indicating the peak fluorescence wavelength is shifted to a higher range from 437-463nm to 487-513nm even though the absorption wavelength is the same (395nm).
To look further into the intensity differences across subjects in different age groups, we plotted the average intensity of the pixels in the patch vs. the subject's age.
Both figures are from illumination at 395nm (lead to peak fluorescence). The left figure represents the average intensity for all subjects' teeth emitting light at 437-463nm, while the figure on the right is for teeth emitting light at 487-513nm. The left cluster and right cluster are different age groups, and the points are colored by gender (red for female, blue for male). There does not appear to be any gender effects based on this plot. The "red-shift" can be somewhat seen again, where the older subjects tend to have more intensity in the higher wavelength compared to younger subjects. However, the error bars are so wide, it is hard to make a judgment based on this alone. Overall, our data does not appear to suggest that older subjects have less fluorescence compared to younger subjects.
Reflectance
Although this study is focused on fluorescence and not reflectance, we briefly took a look at the relative reflectances (diagonal in photo matrix) across the illumination channels just to see if it is comparable with the spectral reflectances found in the previous project about teeth color matching. Since the post-processing to normalize for gain and filter/LED sensitivity effects greatly oversaturated the reflectance diagonal (see earlier figure for subject 7), for this portion of the plots we picked the image in the each subject's set with the highest intensity and scaled all images down equally by it before calculating the average intensity of the pixels in the patch. Therefore the intensity values are not directly comparable with the spectral reflectance data from the color matching project but the relative shape is still the same. Compared with their data, ours has less variance with the exception of the blue line (subject 1). This is most likely due to the fact that subject 1 had a different set of camera capture settings compared with the other subjects in the study. There were issues with the chalk alignment for subject 1's white calibration so from subject 2 onwards, a different set of shutter settings were used.
Conclusions
Previous in vitro studies have shown that when human enamel and dentine is illuminated at 405, 444, and 532nm the emission peaks shifted by ~50-75nm higher than the excitation wavelength. In addition, it has previously been shown that in-vitro teeth are expected to have a peak fluorescence around 440-460nm, the higher range for higher for older, more worn down teeth.
Our findings in vivo show that illumination at 395nm has the highest peak emission between 437-463nm. This is very consistent with the literature. In addition, we also notice the redshift with higher age subjects since almost all (except subject 9) subjects over the age of 50 have a slightly higher peak emission between 487-513nm compared to at 437-463nm. However since the standard deviation in the average intensities is very large, this relationship is still inconclusive. This is mainly due to the image patch selection in which the cropped portion does not contain 100% teeth pixels in order to avoid bias in pixel selection due to subjects not staying completely still. Additionally, we do not have enough evidence in our data to show that older subjects have less fluorescence in their teeth. In fact, our highest peak intensity value is from a subject over age 50, but this may be an outlier.
Unfortunately due errors with camera settings at the start of the project, we were unable to make use of 6 subjects' worth of data because of poor quality. If we had that additional data to make a total of 16 subjects rather than just 10, perhaps the findings would have been a bit more conclusive. Also if we could guarantee better that the subjects can hold completely still for the 5 minutes required for image capture, we could choose a smaller sized patch with only pixels from the teeth and reduce our standard deviation error bars to see trends more clearly. Another issue is that our LED range does not cover enough of the UV spectrum and most of it is in the IR, which isn’t as relevant to the study.
If there is continued interest in this topic, it would be beneficial to partially repeat this study with more subjects and lights covering more of the UV wavelength range (<400nm), and perhaps also tweak the setup so it is easier to keep the subjects' heads still.
Now that we know a bit more about the spectrum where teeth fluoresce in vivo, and other studies have shown that different types of dental caries and calculus have much higher emission wavelengths, we could potentially develop more non-invasive applications to gauge the healthiness or cleanliness of teeth. Perhaps sometime in the near future we will be able to shine LEDs from special flashlights or cellphones and record the image via different wavelength filters in the phone or device to immediately assess the state of our teeth!
References
[1]H. Blasinski, J. Farrell and B. Wandell, "Simultaneous Surface Reflectance and Fluorescence Spectra Estimation", 2016.
[2]I. Lutskaya, N. Novak and V. Kavetsky, "Fluorescence of dental hard tissue and restorative materials", INTERNATIONAL DENTISTRY – AFRICAN EDITION, vol. 2, no. 5, 2016.
[3]Y. Lee, "Fluorescence properties of human teeth and dental calculus for clinical applications", Journal of Biomedical Optics, vol. 20, no. 4, p. 040901, 2015.
Appendix I
All data and script files, along with additional figures, can be found here: [1].