2.1.

The hardware

The rPPG setup was developed for contactless acquisition of gingival perfusion. The setup comprises of a white LED light source (100W electric power), industrial camera (Ximea-xiQ USB-3.0, ADC 8-12-bits, resolution 648x488 pix.) with mounted lens (Edmund Optics, C-mount f=25mm) and 540nm CWL 10nm FWHM narrow-band filter (Edmund Optics), see Fig.1.A. The light source was powered by a stabilized power supply and provided uniform illumination of the measurement site. To reduce glare of gingiva, crossed polarizers were placed in front of the camera lens and light source. Camera exposure was adjusted optimally to fit camera dynamic range and avoid pixel saturation of area of interest. Video acquisition was performed at 25 frames per second and data stored in AVI container as sequence of uncompressed grayscale frames.

Figure 1.

A-The major components of remote photoplethysmography system, B- Measurement setup during rPPG registering in Dental clinics

2.2.

Examination protocol

The study comprised 12 subjects undergoing routine dental procedures in the Dental clinics of the Medical faculty. Before all procedures, subjects were informed about possible risks of local anesthesia and gave their written informed consent. All procedures were approved by University local Ethics committee, and were in accordance to the Declaration of Helsinki.

The subject was comfortably reclined in the dental chair in supine position and the head was fixed in a vacuum support pillow. A soft dental disposable rubber dam cheek retractor was placed so that the anterior maxillary gingiva was exposed, Fig.1.B. The light source and camera were focused on the recording site, and a video was continuously recorded for 12 minutes consisting of a 6-minute baseline and a 6-minute post administration of local anesthesia. In order to produce alterations of gingival perfusion, buccal infiltration of 0.85 ml (half a cartridge) of 4% articainehydrochloride with 1/100.000 adrenaline (Septodont UK Ltd, Units R & S - Orchard Business Centre - St Barnabas Close -Allington, Maidstone, KentME16 0JZ - UK) was performed on each subject. The injection was administered with a 30 gauge 12 millimeter short needle. The buccal infiltration was executed by placing the needle sub-gingival into the alveolar mucosa close to the mucogingival margin over the estimated root of the right lateral incisor (D12). The local anesthetic solution was infiltrated after negative aspiration.

2.3.

The algorithm of rPPG processing

The video file was processed offline using custom-developed dedicated Matlab software^10,11. The processing consisted of several stages (Fig. 2): 1) the video was stabilized, to compensate motion artifacts, and the region of interest (RoI) was manually selected; 2) the rPPG signal was computed from video and the fast-varied (AC 0.7-4 Hz) and slow-varied (DC 0-0.3 Hz) components were obtained, 3) the amplitudes of AC and DC components were calculated from rPPG signal.

Figure 2.

Remote PPG signal processing

To obtain perfusion parameters, a single-period rPPG waveform was computed in beat-per-beat manner. The feet of AC signal were detected in every n-th beat, and the corresponding values of pulse rate, DC and AC amplitudes were calculated. Perfusion index (PI) is related to both AC and DC signals, and can be foud in following way:

To obtain perfusion map, the PI was calculated as time-averaged value (30 sec time window) in every pixel of the video, using algorithms described in our previous work^10,11. The PI map was normalized, using an equation: PI_norm=(PI_i-PI_min)/(PI_max-PI_min), where PI_min is obtained from baseline (where perfusion is not obtained), PI_max is obtained from the site having maximum circulation.

2.4.

Statistical analyses

In order to assess anesthesia evoked perfusion changes in the subject group, each subjects individual PPG perfusion index values were selected during baseline (30 sec time period), first, second, third and fourth minute following administration and at the end of recording, where values did not change (plato phase). Arithmetic mean and standard deviation was calculated and group mean perfusion values were compared in different stages using Kruskal-Wallis One Way ANOVA with Tukey post-hoc test for multiple comparison at the significance level p<0.05. The data is expressed as mean±standart deviation.

3.1.

Gingiva baseline perfusion

Relatively high amplitude rPPG waveform was observed in all subjects during baseline conditions, indicating adequate basal perfusion, which substantially exceed the so called “biological zero” signal (Fig. 3a). As depicted in rPPG PI maps, perfusion in the gum anterior region exhibited symmetrical bilateral distribution of perfusion.

Figure 3.

The image of gingival, perfusion map and waveforms, obtained from region of interest (RoI) during the baseline (A) and 5 minutes following administration of adrenaline containing local anesthetics (B).

The perfusion values moderately varied (3.7-1.4 a.u.) across the subjects (see Fig. 4b), possibly due to other influencing factors such as different levels of emotional stress and arterial blood pressure differences during registering procedure, although similar variance of gingival perfusion has been previously observed in studies of Ahn et.al and Ketabi et al. during Laser Doppler measurements^12,13. The comparison of obtained gingiva baseline perfusion is rather difficult as several principal differences exist between Laser Doppler Imager obtained perfusion, which is a product of red blood cell velocity and concentration and photoplethysmography derived perfusion index which is determined mainly by magnitude of AC component.

Figure 4.

Characteristic trend of gingiva perfusion index at the baseline and following anesthesia, (A), data from one subject. Group mean perfusion values during baseline, and following anesthesia (mean±std) (B).

Still debatable is interpretation of rPPG AC component as there is no unified consensus regarding reflection mode photoplethysmogramme genesis and its provided information. It has been proposed that rPPG AC signal registered at green light illumination, partly reflect complex dynamics of red blood cell orientation and changes in RBC aggregation in small caliber vessels¹⁴, partly on viscoelastic properties of the dermis (dermal deformation) as the dominant mechanism of rPPG signal formation¹⁵, and depends on pulsatile transmural pressure of the arteries, which mechanically compress capillaries in the dermis, thus modulating the blood volume in the capillary bed, which in turn modulates the power of remitted light¹⁶. Thus photoplethysmography signal largely depends on tissue structure and may exhibit differences between gingiva and skin.

However, present observations (unpublished materials) indicate some similarities between cutaneous rPPG and rPPG signal acquired from gingiva despite tissue differences. However, in contrast to cutaneous, gingival rPPG has lower AC amplitude, and larger distortions, probably due to non-uniform illumination of gum surface and due to artefacts, produced by patient motions, which cannot be reduced by motion-compensated algorithms.

3.2.

Perfusion changes during anesthesia

In the moment of anesthetic administration, gums slightly moved, producing sharp and short artefact in the signal for 1-4 seconds, therefore this fragment was excluded from offline analyses. Rapid and statistically significant (p<0.05) decrease of perfusion (approx. 50% from baseline) was observed adjacent to injection site one to three minutes post administration. Low perfusion region gradually spread in lateral direction reaching the midline and remained low for the entire recording time, as seen in the map acquired at the end of recording during plato phase (Fig. 3b). The same trend is depicted in RoI-averaged perfusion curve recorded from one subject (Fig. 4a).

In the first minute following administration, group mean perfusion decreased by 36.66±16%, at second 51.88±11.56%, at third 56.80±12.39% and at fourth minute 59.57±11.06% from the baseline. While the largest perfusion was observed during plato phase 62.16±12.96%, five to six minutes after injection, although due to the variance in subject group statistically significant differences were observed only between baseline perfusion and second, third and fourth minute (Fig. 4b).

Substantial alterations were noticed in rPPG waveform, the amplitude diminished, and distortions increased during plato phase, where gum perfusion in the anesthetized region was substantially reduced reaching nearly “biological zero” level as hemoglobin absorption decreased due to eliminated functional capillary density, produced by arteriolar vasoconstriction at the anesthetic affected site (Fig. 3b).

Similar perfusion alterations were observed by Laser Doppler Imaging, after administration of adrenaline containing anesthetics^12,13. Nevertheless, in this study, the effect of lidocaine and lignocaine containing adrenaline on gingival blood flow has been evaluated but effect of articaine with adrenaline has not been studied. The comparison of the present data is challenging as to the best of our knowledge, this is the first pilot study exploring use of remote photoplethysmography for assessment of gingiva perfusion in human subjects during local infiltration anesthesia and in particularly during infiltration anesthesia administrating articaine with adrenaline.

Previous remote PPG studies on cutaneous circulation demonstrate close similarities between rPPG perfusion and Laser Doppler obtained perfusion trend¹¹, advocating perfusion measurement similarities, which is rather speculation. As direct comparison of perfusion obtained by these two photonic techniques is problematic in the framework of the present study because mucosal and cutaneous structure differs.

We could conclude that in the light of present knowledge, the observed decrease of perfusion during infiltration anesthesia seems consistent, as adrenaline produces substantial vasoconstriction at the injection site which prevents further spread of anesthetics prolonging its action¹⁵.

Taken together, the present pilot results provide an initial insight in the use of remote photoplethysmography technique for oral mucosa assessment however further studies are required to assess reliability of this technique using different provocation tests.