4. Discussion
Most newly discovered caries lesions are in the pits and fissures of occlusal surfaces. Such surfaces are more challenging for lesion activity assessment using time-resolved SWIR reflectance imaging due to the highly convoluted topography. In the first clinical study to assess lesion activity in vivo using time-resolved SWIR reflectance imaging, lesions located on the occlusal surfaces of primary teeth were chosen to ensure that most of the lesions would be active [26]. The study showed the potential of using the delay to discriminate between active and arrested lesions however there was considerable difficulty in sufficiently dehydrating lesions in these areas in 30 seconds. In addition, many of the acquired curves were noisy or incomplete preventing the calculation of kinetic information from the curves such as rates or %Ifin [26]. Values for ΔI% were calculated, however there was no significant difference in ΔI% between active and arrested lesions. In this study and in a recent in vitro study [25], ΔI% was not significantly higher for active lesions compared to arrested lesions. In earlier studies that utilized simulated lesions or lesions that were confined to the outer half of enamel, ΔI% was significantly higher for active lesions [8, 9]. In those early studies, optical coherence tomography was used to assess lesion severity with deeper lesions being avoided due to the limited penetration depth of OCT. However, for this study and a recent study [25], microCT was used to image the proximal and occlusal lesions many of which were of greater severity.
The newly designed probe for the clinical handpiece with the angled and focused air nozzle was as effective as the benchtop system that had the air nozzle pointed almost directly at the tooth occlusal surfaces. A higher air pressure of 25 psi was used instead of the 10-15 psi that was used in previous studies, however this is still markedly lower than the pressures used by dental air syringes that can be as high as 80 psi. Complete dehydration of active lesions should occur within 30 seconds for practical clinical implementation. This was achieved for all the active lesions using both the benchtop system and the handpiece within 30 seconds. In contrast, complete dehydration can take hundreds of seconds for arrested lesions, however the entire curve is not needed to calculate suitable values of %Ifin and rate as they are sufficiently different from active lesions due to the much slower rate of dehydration for arrested lesions.
Based on this study and recent studies [25, 26], it appears that ΔI is not a reliable parameter to discriminate between active and arrested lesions and is not suitable for in vivo measurements since the depth and severity of the lesions is not known ahead of time. The other three parameters that reflect the kinetics of the dehydration process; delay, %Ifin, and rate are less dependent on the lesion depth and severity and are better suited for clinical use.
A previous in vitro study showed that 1950 nm was best suited for lesion activity assessment using time-resolved reflectance imaging [25], however compact SWIR cameras small enough for clinical use are limited to wavelengths less than 1700 nm, thus we chose to use an SLD operating at 1470 nm which overlaps the water absorption band centered at 1450 nm.
In this study, we choose to use the Hill equation to model the kinetics of dehydration as opposed to the sigmoid function that was used previously [8, 9]. Both functions are frequently used to fit sigmoidal shaped curves, however the Hill equation performed better for many of the arrested curves for which the sigmoidal shape was not as well developed, such as the arrested curve in Fig. 5. 3D scatter plots for both the benchtop system and the handpiece using delay, %Ifin, and rate show that the two groups are well separated for clear discrimination between active and arrested lesions. Clinical assessments are more challenging, and it is valuable to have multiple parameters available for use. The calculation of delay, %Ifin, and rate can be easily automated for rapid calculation and 2D projection maps can be created for those parameters in a similar fashion to what has been done previously for optical coherence tomography [6, 28].