2.1.

Clinical Protocol

The study population consisted of a control group of 15 healthy volunteers with no clinical signs of gingival inflammation, a test group I of 25 patients with overt signs of gingival inflammation, and a test group II of six patients with a previous history of periodontitis but no overt signs of gingival inflammation (Table 1). The study was carried out at the out-patient clinic of the Government Dental College (GDC), Thiruvananthapuram, Kerala, India. The study protocol was approved by the GDC Institutional Ethical Committee (No. IEC/C/42-A/2011/DCT/dated 18-01-2011). Every volunteer was informed about the nature of the study, and informed consent was obtained prior to initiation of any study-related measurements. The healthy volunteers and patients enrolled in the study were selected by an experienced periodontist. The patients enrolled in the test groups were 25 to 65 years old, whereas the healthy volunteers were 20 to 35 years old. The clinical study was conducted from April 2011 to November 2011.

Table 1

Characteristics of the study group.

The clinical parameters and treatment history of all participants were recorded. Healthy sites were defined as those with pocket depth (PD) $<3\mathrm{mm}$ and no BOP. The healthy volunteers selected for participation had no clinical signs of gingival inflammation and maintained good oral hygiene. Spectral images of the gingival papillae from canine to canine in both the upper and lower arches (10 sites in total) were acquired from all participants.

All patients were categorized based on the GI, which was introduced by Loe and Silness in 1963. The GI scores of 0, 1, 2, and 3 were assigned as normal gingiva, mild inflammation (with slight change in color and slight edema) in 96 sites, moderate inflammation (redness, edema, glazing, and BOP) in 54 sites, and severe inflammation (marked redness, edema, ulceration, and spontaneous bleeding) in 17 sites, respectively. Test group I consisted of 16 patients with a GI score of 1 (96 sites), six patients with a GI score of 2 (54 sites), and three patients with a GI score of 3 (17 sites). Test group II had six patients (eight sites) with previous history of periodontitis but no overt signs of inflammation (Table 1).

The periodontal charts of these patients showed that they had $\mathrm{PD}\ge 5\mathrm{mm}$ with BOP.²² IOPA X ray of these sites was taken to check the presence of bone loss. In order to avoid overlap between healthy gingiva and mild disease conditions, patients with a PD of 3 to 4 mm were excluded. Other exclusion criteria applied in patient selection were:

2.2.

Instrumentation

The diffuse reflectance imaging system shown in Fig. 1 consists of an EMCCD camera (Model: LUCA-R, DL-604M-OEM, Andor Technology, UK) with $1004(H)\times 1002(V)$ pixels, a Nikkon AF 35-70 zoom camera lens, and a liquid crystal tunable filter (LCTF) with a bandwidth of 7 nm (Model:VIS-07-20-STD, CRI Inc., USA) that can be tuned electronically to any wavelength in the range of 400 to 720 nm. In this study, the LCTF was tuned to coincide with the oxy-Hb absorption peak at 575 nm and the DR spectral intensity maximum at 620 nm. The specifications of the EMCCD camera used for imaging is given in Table 2. Suitable adapters and roller assemblies were built to couple the camera with the focusing lens and LCTF and to facilitate camera lens movement during focusing and zooming. The tungsten halogen lamp (12 V, 55 W) of the dental chair (Trichur Surgicals, Kerala, India) was used for white light illumination of the oral cavity during imaging. Sequential recording of the DR images at 575 and 620 nm and computation of the image ratio R620/R575 were controlled by the solutions for imaging and spectroscopy (SOLIS) program (Andor Technology, United Kingdom, Ver. 1.0) and LCTF control software.

Fig. 1

Schematic of the experimental arrangement for in vivo DR imaging measurements.

Table 2

Specifications of EMCCD camera used for imaging of diffuse reflected light from periodontal region.

2.3.

Data Acquisition

The DRIS assembly was kept in the instrument tray of the dental chair at a distance of one foot from the patient for the easy imaging of the anterior sextant of the gingiva. A sterilized cheek retractor was used to pull the cheeks away from the mouth and hold them in place during camera exposure. The DR imaging was carried out in a dark room to keep external light from entering the camera. Before the tungsten halogen lamp was turned on, the patient was given an eye protective goggle to ensure comfort. The patients were asked to remain stationary for 20 s during the sequential acquisition of images at 620 and 575 nm with the LCTF control software. DR imaging was always performed before clinical assessments, because an injury of the blood vessels of the crevicular plexus by paper strips and a periodontal probe could affect the reflectance spectra of gingiva. Before taking DR measurements, saliva was removed from the gingiva with an absorbing tissue to reduce specular reflection and increase light penetration, thereby enhancing the diffuse reflectance from underlying tissues.

The Andor SOLIS image capturing software that recorded the grayscale images of gingiva has a provision for carrying out automatic image alignment with pixel accuracy to compensate for slight patient/camera movement during imaging. Arithmetic functions associated with the software were used to compute the ratio images. A pseudo color map (PCM) of the image intensity ratio (R620/R575) variation was generated using the same software to distinguish clearly between healthy and diseased sites. Differences in the diffuse reflectance were calculated for every image pixel, approximately corresponding to the area occupied by a single cell. Properties such as the diffuse reflectance were calculated from the pixel intensity, and their spatial distribution (PCM) was overlaid on to the image of the diseased site. The PCM of the ratio image classified the gingiva as blue (healthy tissue), green (mild inflammation), red (moderate inflammation), or yellow/white (severe inflammation).

2.4.

Spectral Data Processing

Region of interest (ROI) is an important post-acquisition tool for quantitative analysis of data using the Andor SOLIS program. Thirty ROIs with a maximum size of 100 pixels were selected for each site in the ratio images of patients and healthy subjects. Mean pixel intensity was estimated as the average of 30 such ROIs from each site. These intensity values are presented in a scatterplot diagram for tissue classification as healthy, mild, moderate, and severe.

2.5.

Statistical Analysis

Summary statistics (proportion, mean, and SD) were used to describe the collected data. The study group was divided into four different categories (healthy gingiva, mild inflammation, moderate inflammation, and severe inflammation) based on GI value. Differences in mean age were compared across these three groups using analysis of variance (ANOVA). The medians of mean pixel intensities were compared across the study groups using Kruskal-Wallis one-way ANOVA ($K$ samples) to discriminate healthy tissue from mild inflammation, healthy tissue from moderate inflammation, healthy tissue from severe inflammation, and moderate inflammation from severe inflammation. We used Statistical Package for Social Sciences (Ver. 17, SPSS, Inc., Chicago, Illinois) for all analyses.

3.1.

Image Analysis Features

Figures 2Fig. 3–4 show clinical photos of diseased periodontium, along with the monochrome and processed ratio images for healthy/mild, moderate, and severe inflammatory conditions, respectively. Figure 2(c) is the PCM R620/R575 ratio image of a healthy volunteer, shown in Fig. 2(a); the corresponding grayscale ratio image is shown in Fig. 2(b). Figure 2(d) shows the photo of the gingiva of a patient with mild inflammation. The grayscale ratio (R620/R575) image for that patient, shown in Fig. 2(e), appears in green color after PCM in regions of mild inflammation, as shown in Fig. 2(f). Pseudo color code is also shown to denote pixel intensity variations.

Fig. 2

(a) Healthy gingiva, (b) grayscale image ratio R620/R575 of healthy gingiva, (c) PCM ratio image of healthy gingiva, (d) gingiva with mild inflammation, (e) grayscale ratio image of gingiva with mild inflammation, and (f) PCM ratio image of gingiva with mild inflammation.

Fig. 3

(a) Gingiva with moderate inflammation highlighted inside yellow rectangle, (b) grayscale image ratio R620/R575, (c) PCM image ratio, and (d) enlarged view showing disease-active sites.

Fig. 4

(a) Gingiva with severe inflammation highlighted inside yellow rectangle, (b) grayscale image ratio R620/R575, (c) PCM image ratio, and (d) enlarged view of disease-active sites.

The pseudo coloring of the monochromatic ratio image with varying pixel intensity helps to locate points of maximum oxy-Hb absorption. It can be seen that the average pixel intensity ratio value is $4.81\pm 1.04$ for healthy tissue but $8.82\pm 1.79$ for mild inflammation. Figure 3(a) shows the clinical picture of a moderately inflamed gingiva. The inflammatory region is shown inside the yellow rectangular box. The PCM ratio (R620/R575) image [shown in Fig. 3(c)] of an inflammatory condition shows red spots, which can be visualized in Fig. 3(d). The average pixel intensity ratio (R620/R575) obtained for moderately inflamed gingiva is $12.92\pm 2.77$. Figure 3(b) shows the grayscale image of the moderately inflamed gingiva. Figure 4(a) is a clinical picture of severe periodontal inflammation, and Fig. 4(c) is its PCM image with the affected site highlighted inside the yellow rectangle. The highlighted region on enlargement, shown in Fig. 4(d), shows regions of yellow color with an average pixel intensity ratio (R620/R575) value of $16.64\pm 2.18$. Figure 4(b) is the monochromatic R620/R575 ratio image of diseased gingiva before PCM.

Figure 5(a) shows a photo of the gingiva of a patient with moderate bone loss in the 12 region. Figure 5(b) shows its monochrome image R620/R575 ratio, and Fig. 5(c) shows the PCM of the image ratio, with the affected region highlighted inside the yellow rectangular box. This patient had moderate periodontitis with an apparently healthy gingiva in the upper anterior sextant. The DR image correlates well with the absence of inflammation in the papillary region and inflammation in the attached gingiva in region 12, as shown in Fig. 5(d). This case was identified due to the presence of underlying inflammation in the attached gingiva analogous to periodontitis. The affected region clinically had recession and bone loss, which was confirmed through an IOPA X ray.

Fig. 5

Gingiva of a patient with a calculus index score of 1 and moderate bone loss in region 12: (a) Photo of the affected gingiva, (b) grayscale image ratio R620/R575, (c) PCM image ratio, and (d) enlarged view of disease-active site.

3.2.

Scatterplot Analysis

The DR image pixel intensity ratio (R620/R575) for different categories of periodontal inflammation was determined, and a scatterplot diagram was drawn (Fig. 6). For group classification, cutoff lines are drawn in the scatterplot diagram at the mean value of the adjacent groups. For example, the average pixel intensity ratio value is $4.81\pm 1.04$ for healthy tissue and $8.82\pm 1.79$ for mild inflammation, so the cutoff line for discrimination between healthy and mild inflammation is drawn at the mean of 4.81 and 8.82, which is 6.81. Similarly, the cutoff lines are drawn at 10.87 and 14.78 in the scatterplot of R620/R575 ratio (Fig. 6) to discriminate between the tissue groups as mild-moderate inflammation and moderate-severe inflammation, respectively. The sensitivity and specificity obtained from the scatterplot of the R620/R575 image ratio are given in Table 3. In Fig. 6, out of 96 mild inflammatory cases, eight were misclassified as healthy, giving a sensitivity of 91.6%. Out of 77 cases of healthy gingiva, five were misclassified as mild inflammation, giving a specificity of 93%. Out of 54 cases of moderate inflammation, nine were misclassified as mild inflammation, giving a sensitivity of 83%. Out of 96 cases of mild inflammation, four were misclassified as moderate, giving a specificity of 95.8%. Similarly, out of 17 cases of severe inflammation, three were misclassified as moderate, giving a sensitivity of 82.3%. Out of 54 cases of moderately inflamed gingiva, 13 were classified as severe, giving a specificity of 75.9% (Table 3).

Fig. 6

Scatterplot diagram of the DR image intensity ratio R620/R575 for discrimination between healthy gingiva and mild inflammation, mild and moderate inflammation, and moderate and severe inflammation categories.

Table 3

Diagnostic accuracies obtained by DR imaging technique for the classification of gingival sites.

4.1.

DR Imaging for Monitoring of Periodontal Disease

In an earlier clinical study, it was observed that oxy-Hb concentration increases with an increase in periodontal inflammation, and the associated changes could be detected from the R620/R575 ratio with good diagnostic accuracies.²¹ In this study, the gingival erythema was monitored from changes in the intensity of oxy-Hb absorption at 575 nm in the diffusely reflected white light spectra. The results show that sites with BOP correspond well with the PCM image and the averaged pixel intensity ratio (R620/R575) value. Furthermore, it is shown that DR imaging can precisely demarcate areas with increased vascularity, which is known to occur during gingival inflammation (Figs. 2Fig. 3–4).

4.2.

Identification of Inflammatory Changes in Subclinical Stage in Sites with Disease Activity

Periodontal destruction is site specific and does not occur in all parts of the gingiva and teeth simultaneously. The exact points of inflammation in moderately inflamed and severely inflamed gingiva, as seen in the false colored images (Figs. 3 and 4) are not observed using conventional diagnostic procedures. Therefore, DR imaging can identify areas with early inflammatory changes in the underlying tissues, even when the sites appear clinically healthy (Fig. 5) but are prone to disease. This enhances the prospect for site-specific treatments like photodynamic therapy to selectively target and inactivate periodonto-pathogenic bacteria.

The currently used diagnostic procedures cannot distinguish between disease-active and disease-inactive sites at any given point of time, nor can they reliably identify susceptible individuals based on disease-active and disease-inactive sites.²³ The DR image ratio (R620/R575) of the gingiva of a patient having apparently healthy gingiva in the upper anterior sextant (Fig. 5) corroborated well with the moderate bone loss detected in region 12 in the IOPA X ray (data not shown) and with areas in the papillary region where inflammation was absent. Similar changes were observed in the case of six patients having apparently healthy gingiva with alveolar bone loss that was revealed in the X ray. This implies that the DR imaging technique could be effectively utilized to diagnose underlying erythema in apparently healthy gingiva after scaling.

Conventional clinical examinations depend on alveolar bone and periodontal attachment loss. They cannot be used to predict the susceptibility to future disease progression or to dictate appropriate treatment plans.²³ At present, the only clinically available method for the prediction of progressive attachment loss or disease progression is BOP, which is not a good predictor of progressive attachment loss. However, the absence of BOP is an excellent predictor of periodontal stability.^24,25

4.3.

Early Detection of Inflammation

The early detection of periodontal disease enhances the possibilities for appropriate treatment planning, whether it is prophylactic, medical, or surgical. The DR ratio images obtained in this study reveal the ability of this technique to help physicians detect early symptoms of gingival inflammation, whether due to mild gingivitis or attached bone loss. While radiographs reveal bone loss only after 30% to 50% of the mineral has been lost, the DR imaging technique has the potential to find early stages of mineral loss from the inflammatory conditions in gingival tissues. Also, early diagnosis and consequent treatment of periodontal disease may have a major impact on the control of other systemic infections.^4,26 Therefore, the currently used technique of periodontal probing, which fails to monitor the initiation or progression of periodontal destruction, could be replaced by noninvasive DR imaging, which can diagnose early stages of periodontal breakdown.

4.4.

Reproducibility of the Results

The DR imaging system used in this study assesses the given area of the periodontium as a whole and gives an objective result with uniform criteria for disease diagnosis, thereby increasing its reproducibility and reliability. Usually the clinical assessment of gingival inflammation is often subjective, because the evaluation of visual parameters can vary from person to person. Previous studies have demonstrated that DR spectroscopy can be used to assess the degree of erythema and edema taking into consideration the light absorbed by particular chromophores. Kolias et al.²⁷ reported that erythema correlates well with the relative concentration of oxygenated hemoglobin. Although these findings were reported for skin, gingival tissues have a similar structure, consisting of an external layer of epithelium with an underlying layer and connective tissue, where most of the microvasculature is embedded.²⁸

4.5.

Noninvasive Technique for Real-Time Diagnosis of Periodontal Disease

Image analysis using the scatterplot of the R620/R575 image ratio (Fig. 6) has shown a sensitivity of 91.6% and a specificity of 93% for the discrimination of healthy tissue from mild inflammation, along with a sensitivity of 83% and a specificity of 95.8% for the discrimination between mild and moderate inflammation. In comparison, a sensitivity of 82.3% and a specificity of 75.9% was obtained for the discrimination between moderate and severe inflammation. The misclassified cases cannot be completely attributed to the technique used; they could be due to the limitation in taking the conventional diagnostic technique as our gold standard. Also, in some cases, there could be a possibility for specularly reflected white light from saliva to overshadow the less intense backscattered light from underlying tissues. Since the diagnostic accuracies achieved are promising, studies on a random population would go a long way in establishing the reliability of this technique in the mapping of periodontal diseases.