2.1.

Clinical Protocol

The study population consisted of 30 volunteers with healthy gingiva, 37 patients with gingivitis, and 23 patients with periodontitis. The study protocol was approved by the Institutional Ethical Committee of Government Dental College (GDC), Thiruvananthapuram (Approval No. IEC/C/42-A/2011/DCT/dated 18-01-2011). The study was carried out at the outpatient clinic of GDC and informed, written consent was obtained from each participant prior to enrollment. A diagram showing participant flow is given in Fig. 1. Patient recruitment and recording of clinical parameters were done by a periodontist. Healthy sites were defined as those with pocket depth $(\mathrm{PD})<3\mathrm{nmm}$ and no bleeding on probing (BOP), and gingivitis sites were those with $\mathrm{PD}<3\mathrm{mm}$ and BOP. Periodontitis sites were those with severe involvement having $\mathrm{PD}\ge 5\mathrm{mm}$, clinical attachment loss $\ge 5\mathrm{mm}$, and a gingival index (GI) (Loe and Silness) score of 2.1 to 3.0, which indicates severe inflammation, making these sites clinically distinct. Intraoral periapical (IOPA) radiographs were also taken to confirm alveolar bone loss as bone loss can occur as a result of the extension of inflammation from gingiva to underlying tissues. Healthy volunteers included in the study were in the age group of 20 to 35 years and maintained good oral hygiene. Patients between 30 and 65 years of age with gingivitis and periodontitis were categorized based on the classification by the American Academy of Periodontology. This study was conducted during March and April, 2011. Exclusion criteria followed were:

Stringent selection criteria were followed throughout this study to overcome the limitations of conventional diagnosis methods. Healthy sites used in this study for comparison were from completely healthy volunteers, and healthy sites of patients were not considered. The gingivitis sites in severe periodontitis patients were omitted as well. Severe periodontitis sites were only taken for obtaining periodontitis spectra so as to reduce overlap, and DR measurements were taken from papillary gingiva since periodontal breakdown is known to begin from the col, which is nonkeratinized. Measurement of reflectance spectra was always performed before clinical assessment of disease stage as clinical examinations may disturb the sites, causing bleeding and thereby interfering with the DR measurements. A convenience sampling method was used with a sampling size of at least 50 in each group so that Gaussian assumptions can be satisfied during the statistical analysis.

Fig. 1

Participant flow diagram.

2.2.

Instrumentation

A schematic representation of the portable DR spectroscopic system (DRSS) used in this clinical study is shown in Fig. 2, with the cross-sectional view of the probe tip given as inset. A tungsten halogen lamp (model LS1-LL, Ocean Optics, Dundlin, Florida) was used as the white-light source to illuminate the gingival tissue through the 200-µm-diameter fiber of a specially designed stainless steel hand piece with a diameter of 1 mm at its tip. The scattered light from the tissue was collected using another fiber of 400 µm diameter positioned at a distance of 260 µm from the illumination fiber. The DR light emanating from the fiber tip was coupled to the entrance slit of a miniature fiber-optic spectrometer (model USB 2000 FL VIS-NIR, Ocean Optics, Dunedin, Florida), which was connected to the USB port of a laptop computer. This spectrometer was fitted with a 500-nm blazed, 600-lines/mm grating and a 2048-element linear silicon CCD array for recording of spectra in the 360 to 1000 nm wavelength range with 7 nm resolution. In order to maximize the collection of scattered light from tissue, a black PVC sleeve was inserted at the probe tip, which maintained an optimum distance between the tissue and probe tip. This black sleeve also prevented the room light from entering into the detection system and, being disposable, provided extra hygiene. Furthermore, the small footprint of the probe tip helped in identification of the different levels of erythema and location of the diseased area with precision.

Fig. 2

Schematic of the experimental arrangement for in vivo DR spectral measurements. A cross-sectional view of the probe tip is given as inset.

2.3.

Data Acquisition

After placing the PVC sleeve of the probe tip in contact with the gingiva, the DR spectra were recorded using the OOI Base32 software of Ocean Optics, Inc., configured to record the spectra, averaged for 40 scans, with a boxcar width of 10 nm and an integration time of 100 ms. DR spectral data were categorized as normal/healthy (recorded from 70 sites in healthy volunteers), gingivitis (from 63 sites in patients), and periodontitis (from 58 sites in patients). Fifteen sets of spectral measurements were taken from each site, and the mean spectral data was used for analysis. The hand piece was sterilized everyday and before use on a new patient.

2.4.

Spectral Data Analysis

In order to determine the extent of periodontal destruction in each patient, PD, CAL, and GI were measured and calibrated by a trained examiner across the study population. The DR spectra of gingival sites were normalized to the peak intensity at 620 nm in order to identify spectral regions of maximum variance during tissue transformation. Intensity ratios from spectral regions of interest (SROI) thus identified were compared with the periodontal parameters to determine the diagnostic accuracies for discrimination of gingivitis and periodontitis from healthy gingiva. ANOVA was performed using Origin 7 software to comprehend whether the means of DR spectral intensity ratios show significant differences between the groups.

3.1.

Spectral Features

The averaged DR spectrum in the 400 to 750 nm region normalized to the maximum intensity at 620 nm (Fig. 3) shows marked differences between healthy/normal, gingivitis, and periodontitis sites. The oxygenated hemoglobin absorption dips at 545 and 575 nm are prominent in the DR spectra of periodontitis and not in the spectra of healthy gingiva. The ratio values of R620/R575, R620/R545, and R615/R460 show an increasing trend from healthy to periodontitis (Table 1).

Fig. 3

Averaged normalized DR intensity spectra of healthy, gingivitis, and periodontitis tissues. Error bar gives the standard deviation of spectral intensities at 545 and 575 nm. Each spectrum is the average of $70\times 15$ healthy, $63\times 15$ gingivitis, and $58\times 15$ periodontitis sites.

Table 1

Averaged DR intensity ratios (R620/R575, R620/R545, and R615/R460) for different grades of gingival inflammation.

3.2.

Scatterplot Analysis

Marked differences observed in the normalized spectra between diseased and normal tissues are mainly associated with oxygenated hemoglobin absorption. The DR intensity ratios with respect to the intensity maximum at 620 nm and the intensity at 575 nm (R620/R575) were calculated, and the scatterplot is given in Fig. 4. Similarly the scatterplot of the ratio R620/R545 is given in Fig. 5. The absorption coefficient crossovers of oxygenated hemoglobin and deoxygenated hemoglobin (isobestic points) at 460 and 615 nm, where the absorption is invariant to changes in the ratio of two chromophores, were employed to discriminate the diseased and healthy gingival tissues. The scatterplot for the DR ratio R615/R460 is not shown. The cutoff lines are drawn in the scatterplot diagrams at the mean value of the data points of the two classifying groups. The cutoff lines drawn at 1.378 and 1.850 in the scatterplot of R620/R575 (Fig. 4) discriminate the tissue groups healthy-gingivitis and gingivitis-periodontitis, respectively, whereas the cutoff lines at 1.727 and 2.302, respectively, discriminate the same groups in the scatterplot of the R620/R545 ratio (Fig. 5). The sensitivity and specificity obtained from the scatter plots for the three ratios R620/R575, R620/R545, and R615/R460 are given in Table 2.

Fig. 4

Scatterplot of DR spectral intensity ratio R620/R575; (a) gingivitis-periodontitis (G–P) and (b) healthy-gingivitis (H–G).

Fig. 5

Scatterplot of DR spectral intensity ratio R620/R545; (a) gingivitis-periodontitis (G–P) and (b) healthy-gingivitis (H–G).

Table 2

Diagnostic accuracies (sensitivity and specificity) obtained from the scatterplot of intensity ratios R620/R575, R620/R545, and R615/R460.

In Fig. 4(b), 6 out of the 63 gingivitis cases were misclassified as healthy, giving a sensitivity of 90%, and out of 70 healthy cases, 4 cases were misclassified as gingivitis, with a specificity of 94%. Also in Fig. 4(a), 5 periodontitis cases were misclassified as gingivitis, giving a sensitivity of 91%, and no gingivitis cases were misclassified as periodontitis (100% specificity). The scatterplot of the R620/R545 ratio (Fig. 5) shows that out of the 63 cases of gingivitis, 16 were classified as healthy (sensitivity of 75%), and out of 70 healthy cases, 12 were classified as gingivitis (specificity of 83%). In the case of discrimination between gingivitis and periodontitis, 11 out of 58 periodontitis cases were classified as gingivitis, with a sensitivity of 81%, and no gingivitis cases got classified as periodontitis (specificity of 100%). In the case of the R615/R460 ratio (figure not shown) 16 cases of healthy tissues were misclassified as gingivitis, and 23 gingivitis cases were misclassified as healthy, leading to a lower sensitivity of 64% and a specificity of 77%. Using this DR ratio 8 gingivitis cases were misclassified as periodontitis, and 21 periodontitis cases were misclassified as gingivitis, which further reduces the sensitivity to 64% and specificity to 87%.

3.3.

Hb Concentration and Oxygen Saturation

In order to test the variation in relative concentrations of Hb, oxy Hb, and apparent ${\mathrm{SO}}_2$, DR spectra were recorded from papillary gingiva in clinically healthy, gingivitis, and periodontitis sites of a few sets of patients. The ${\mathrm{HbO}}_2$, Hb, and apparent ${\mathrm{SO}}_2$ concentrations in these cases were calculated by the method described by Hanioka et al.¹⁸ and are plotted in Figs. 6(a)–6(c). Relative concentrations of Hb and ${\mathrm{HbO}}_2$ show an increasing trend whereas apparent ${\mathrm{SO}}_2$, signifying oxygen saturation in tissues, shows a decreasing trend with the inflammatory status of tissue.

Fig. 6

Relative concentration of (a) hemoglobin, (b) oxygenated hemoglobin, and (c) apparent ${\mathrm{SO}}_2$ from healthy, gingivitis, and periodontitis sites.

4.1.

Spectral Features

The mean DR spectrum shown in Fig. 3 for periodontitis represents severe periodontitis cases. This spectrum shows prominent dips at 545 and 575 nm due to the increased oxygenated hemoglobin absorption in periodontitis tissues as compared to gingivitis and healthy/normal tissues. This could be attributed to changes in functional parameters, such as blood concentration, oxygenation, and blood flow, during disease progression. Previous studies have shown that blood concentration and blood flow in gingiva increase and tissue oxygen saturation reduces during gingival inflammation.^8,9 In a study of hemoglobin concentration and oxygen saturation of clinically healthy and inflamed gingiva in humans, Hanioka et al.¹⁸ reported an increase in Hb concentration, ${\mathrm{HbO}}_2$ concentration, and deoxy Hb concentration and a decrease in apparent ${\mathrm{SO}}_2$(Table 3). The increase of Hb concentration in inflamed gingiva shown in Fig. 6(a) mainly reflects increased blood volume in tissues, resulting from the opening of nonfunctioning vessels and formation of new vascular parts with gingival inflammation.²⁰ The increase in oxy Hb concentration shown in Fig. 6(b) suggests that the oxygen supply to inflamed gingiva increases in order to meet metabolic demand, and the decrease in apparent ${\mathrm{SO}}_2$ shown in Fig. 6(c) reflects tissue hypoxia.¹⁸ However, Zili et al.²¹ reported a decrease in concentration of oxy Hb with increase in inflammation. In the earlier studies conducted by Liu et al.¹³ it was reported that oxy Hb concentration increases in gingivitis sites as compared to healthy tissue and decreases in periodontitis sites. This discrepancy can be observed in the case of total hemoglobin concentration as well. This could have been due to the recording of healthy spectra from apparently healthy sites contralateral, or nearest to contralateral, of the diseased site (probably leading to lack of independence of multiple sites within the same patient) and using probe tips of larger diameter.

Table 3

Contradictory clinical and spectral impressions from DR measurements along with the site-specific clinical parameters of patients.

The healthy site in a periodontitis patient has a different cytokine profile with respect to healthy sites in healthy volunteers. Since inflammation is known to start from the papillary region, where interdental col is nonkeratinized and more susceptible to periodontal breakdown, our measurements were taken from the papillary gingiva. The DR spectra of healthy sites show comparatively smaller absorption dips due to oxygenated and deoxygenated hemoglobin and also differ from other anatomical sites of the oral cavity. Hanioka et al.¹⁸ reported a variance in the Hb index between gingiva and alveolar mucosa in healthy subjects. In our study, the DR spectrum of healthy volunteers showed minimal absorption due to oxygenated hemoglobin. Moreover, with our small-diameter (1-mm) probe tip we were able to record the spectrum precisely from regions of erythema.

In another study, it was shown that gingival tissues in the healthy population, like interdental papilla and gingiva, have higher oxygenation content but lower total hemoglobin concentration than alveolar mucosa.^8,22 In a longitudinal study with dogs, Baab and Oberg²³ reported that gingival blood flow increases slightly with inflammation, but significant decrease on resolution of inflammation was not noticed. During a cross-sectional study on humans, Kervonbundit et al.⁸ reported considerably larger blood flow for cases with moderate gingivitis than for healthy cases. Liu et al.¹³ demonstrated that a near-IR optical spectroscopy system can be used to acquire spectra from periodontal tissues, and the data extracted was found to provide detailed, site-specific information on multiple aspects of periodontal inflammation. Both Liu et al. and Hanioka et al.^13,18 found that tissue oxygen saturation was significantly decreased in both gingivitis and periodontitis sites as compared to control sites. These studies correlate well with our findings (Table 3).

The capability of spectroscopy to identify early signs of inflammation leading to tissue breakdown is well understood.²⁴ In a few clinically healthy sites, spectral changes and variations towards the diseased nature were noticed. This could possibly be due to the presence of subclinical inflammation present in clinically healthy sites. DR spectroscopy has the ability to detect erythema noninvasively when gingivitis is present in any part of the gingiva. The results obtained in this study show the capability of DR ratios to detect underlying inflammation even in the case of severe periodontitis patients having plaque-induced inflammation.

4.2.

Tissue Classification

In this study it was possible to distinguish healthy gingiva, gingivitis, and periodontitis, based on their position in the scatterplot, by correlating with the clinical diagnosis based on the probing PD, CAL, and GI score (Figs. 4 and 5). Analysis of DR intensity ratio R615/R460 using the scatterplot diagram have shown a sensitivity of 64% and specificity of 77% for discriminating healthy tissues from gingivitis, whereas for distinguishing gingivitis from periodontitis a sensitivity of 64% and specificity of 87% was observed. Recently, Zakian et al.¹⁹ carried out spectral imaging of gingival inflammation at 615 and 460 nm wavelengths, which represent the spectral absorption crossovers between oxygenated and deoxygenated hemoglobin, and used the spectral reflectance image ratio R615/R460 to demarcate regions of inflammation. The diseased subjects in their study were healthy volunteers who were induced gingivitis by cessation of oral hygiene for two weeks. ANOVA shows a significance level with $p$-value $<0.05$ while comparing the means of the R620/R575 and R620/R545 ratios for discrimination between healthy and gingival cases, and healthy and periodontitis cases, as compared to the significance level of $p<0.01$ obtained for the R615/R460 ratio (Table 1).

Spectral analysis using the scatterplot ratio R620/R545 (Fig. 5) has shown a sensitivity of 75% and specificity of 83% for the discrimination of healthy tissues from gingivitis and a sensitivity of 81% and specificity of 100% for the discrimination between gingivitis and periodontitis. In comparison, the R620/R575 ratio (Fig. 4) has shown an improved sensitivity of 90% and specificity of 94% for discrimination between healthy and gingivitis tissues and a sensitivity of 91% and a specificity of 100% for discrimination between gingivitis and periodontitis tissues. The increase in sensitivity and selectivity observed in this study can be attributed to the stringent selection criteria followed, in particular the use of a healthy volunteer data set for comparison with inflamed tissues. However, the levels of various mediators of inflammation present in apparently healthy sites of diseased individuals could possibly differ from healthy volunteer sites. For the same reason, gingivitis cases localized in cases of generalized periodontitis and vice versa were omitted. This was also required in order to avoid misclassification. The study results show that the extent of underlying inflammation that may not be visible clinically could be detected using DR spectroscopy, leading to discrimination of gingivitis from periodontitis with higher sensitivity and specificity. Moreover, in comparison with the probe tip diameters of 2 mm and 5.18 mm used, respectively by Hanioka and Liu,^13,18 our probe tip was of 1 mm in diameter, which to a large extent eliminated any chance of diseased sites overlapping with healthy sites (Table 3). Sometimes erythema affects very small areas, and sites even a few millimeters away could also show a completely different level of erythema. Therefore, with our smaller diameter probe tip we were able to locate the diseased area more accurately as compared to earlier reports.

Currently, the diagnosis of periodontitis is primarily based on clinical observations, radiographs, and microbiological findings that are considered to complement each other; together, they help to define the extent and severity of periodontitis. Periodontitis being a multifactorial disease, its manifestation and progression are also influenced by a variety of other risk factors, such as genetic, systemic, social, and behavioral factors. After clinical observation to determine GI based on visual perception of tissue color, the most widely used mode of periodontal examination is periodontal probing that helps to determine critical clinical parameters, such as BOP, PD, and CAL, of periodontal tissue. Nevertheless, there are some unavoidable limitations due to the difficulty in precisely duplicating the insertion force, probe placement, and angulation. Further, the extent of the probe penetration is influenced by the inflammatory status of the tissue. When healthy tissues are examined, the probe tip stops coronally at the apical termination of the junctional epithelium, but at inflamed sites the probe tip usually passes apically to this point. In such cases, inflammation affects the measurement of probing depth in a manner unrelated to the attachment level. Besides, the poor reliability and reproducibility associated with measuring CAL to monitor the progression of periodontal destruction and to evaluate the effect of periodontal treatment limit the practical value of periodontal probing. Table 4 shows specific cases wherein the spectral impression obtained using DR has failed to correlate with standard clinical procedures followed in this study. Therefore, the deviation of spectral impression from clinical parameters could be attributed to the drawback of this conventional diagnostic method. It may also be pointed out that these methods do not have sufficient reliably to identify susceptible individuals or distinguish between disease-active and -inactive sites.

Table 4

Comparison of study criteria, materials, and DR spectral characteristics.