Accuracy of Facial Soft Tissue Measurements Using Photographic Analysis Method

Introduction

Clinical records in the form of facial photographs are an essential diagnostic aid in the field of orthodontics.^1,2 With the advent of technology, digital cameras have been widely used for clinical photography. Initially, the field of anthropology used clinical photography and conventional photographs of the face. Presently, there are many facial measurement studies^3,4 based on standardized photographic techniques.

Stoner after Sheldon in 1940 introduced photogrammetry in the field of Orthodontics and concluded that standardized photographs are useful in recording accurate anthropometric measurements.⁵ The term "poor man's cephalometric analysis" has been used to refer to the facial profile analysis technique.⁶ It was described as the "art, science, and technology of obtaining reliable information about physical objects through processes of recording, measuring, and interpreting photographic images" by Chadwick and the American Society of Photogrammetry.^7,8

Facial photographic analysis offers valuable information regarding diagnosis and treatment planning, marking various landmarks, reference lines and deriving angles, ratios and distances that characterize the face. The advantages of photography include, long-term storage, reliability, availability, no radiation hazard. Comprehensive knowledge of the interactions between the teeth and skeleton, face asymmetries, and other pertinent characteristics deemed necessary for an accurate diagnosis and treatment planning. Other advantages include monitoring treatment progress and any modification needed to evaluate facial proportion and diagnose asymmetric facial features. The quantification of facial form has evolved to prioritize aesthetic outcomes in orthodontics.

Taking measurements directly introduces instrumentation errors due to contact with the subject and requires good patient compliance, which is difficult for young children. There are many studies done on photographs by studying various angles and ratios and in deciding the treatment as well as predicting the treatment by relating the patient's information to parental data through photographs.^9-11 Chang et al. (2011) used frontal photography to study how parents' face traits are passed down to their children.¹²

Numerous morphometric techniques exist, such as tensor analysis and Euclidean distance matrix analysis 23 Faisal A et al., RJDS 2024;16(3):22-30 (EDMA), which examine the soft tissue outline and assess lateral cephalograms.¹³ While two-dimensional digital photography being the old norm is well known for orthodontic treatment planning, the contemporary methods include computerized tomography, stereo-photogrammetry and laser scanning with reliable results. Errors can arise from a variety of sources while obtaining facial photos. Photographs of insufficient calibre & quality and improper standardization techniques can inaccurately reflect the defects that may be present, as well can be a source of error in photographic measurements.

Facial photographic analysis of the soft tissues has been proven to be a reliable and accurate method in the literature.¹⁴ The goal of the current study was to evaluate the intra- and inter-examiner reliability of landmark recognition and measurement using standardized frontal and profile patient pictures using a customized software.

Material and Methods

This study received ethical approval from the IRB of DAPMRV Dental College & Hospital, Bengaluru, Karnataka, India (IRB No: 336/VOL-2/2019).

Sample size

The calculated sample size using the sample size calculators by Wan Nor Arifin was 50 pictures. Based on the results of pilot study, the following input parameters were used: Minimum acceptable reliability (Intra class coefficient) - 0.60, Expected reliability (Intra class coefficient) - 0.95, α-two tailed- 0.05, β (1-power)- 99%, number of raters/repetition per subject-2.

Inclusion criteria

Patients in age range of 12-18 years, soft tissue ANB 0-6 Degrees, no facial or spinal abnormalities and standardized photographs were included in the study.

Exclusion criteria

Patients with genetic syndrome, any craniofacial and dental trauma, gross facial asymmetry, cleft lip/ palate defects (treated / untreated), and malformed faces were excluded.

Photographic technique

Extraoral photographs (frontal & profile photographs) of patients were taken with a DSLR camera (D-52, Nikon Corporation). The ala-tragus plane of the soft tissue was parallel to the floor for frontal, interpupillary, and profile photos, with the teeth in centric occlusion and the facial muscles relaxed. To make sure that the region between the patient's ear and tip of the nose is within the depth of field, the lower eyelid of the patient was the main focus during the frontal photo shoot. The patient's face and neck were photographed at a suitable distance of about 4-4.5 feet, with a fair margin of space surrounding the camera lens in a vertical position, in order to provide high-quality and consistent images.

Cropped images in a 4 × 6 portrait layout using Microsoft Image Editor were saved at optimal quality without any compression as TIFF files with a resolution of 300 DPI. All guidelines of the American Board of Orthodontics were adhered to while taking photographs.15 Photographs of 50 patients were selected, which included their profile and frontal images, totalling 100 photographs.

Photographic analysis

Frontal and profile photographs (Tables 1) were analyzed using the software IC Measure, The Imaging Source Europe, Version 2.0.0.133, Type EXE. Numerous choices are available for identifying landmarks and measuring angular and linear dimensions with this software. It is user-friendly, easy to understand, does not slow down processing performance, and has few storage problems. The TIFF format of the patients' standardized extraoral photos was imported into the software. To create intra and inter-examiner correlations, the examiners marked and measured the linear and angular dimensions using a variety of software tools. Using a mouse-driven pointer, landmark detection was done manually on face photos, and the software automatically reflected the measurements (Figure 1). Instant readings were recorded using software, tabulated in excel sheet and sent for statistical analysis.

Two examiners were selected who were qualified Orthodontists with same years of residential training and five years of clinical experience after post-graduation. Both the examiners were made to mark the landmarks and perform the analysis in standard conditions (i.e, morning time) and were given 10 minutes for profile and frontal photographic analysis, respectively. The analysis was done two times, at T0 and after 24 hrs (T24) by the same examiners on all the 50 frontal and profile pictures (Figure 2).¹⁶

Statistical analysis

The statistical tests were carried out using IBM's SPSS (Statistical Package for Social Sciences) version 25.0, located in Chicago, and all measurements were recorded in Excel sheet. To evaluate the probability distribution, the Kolmogorov-Smirnov test was employed. The descriptive statistics were used. The intra-class correlation coefficient was used to evaluate the correlation between and among examiners. A statistically significant P-value was defined as less than 0.05.

Results

Table 3a shows the correlation coefficient value and P values for different set of linear parameters. The inter observer initial reading and after 24 hr reading correlation coefficient for two observers are presented along with the intra observer initial reading and 24 hr reading correlation coefficient for both the observers.

In case of inter observer readings, initial readings (T0) of the linear parameters Sn-Me, Sn-St, St-Me, Tr-Me, ChR-ChL showed significant correlation (P <0.05) whereas parameter Tr-N showed non-significant correlation (P >0.05). In case of inter observer post 24 hr readings (T24), the linear parameters Sn-Me, Sn-St, St-Me, Tr-Me, ChR-ChL showed significant correlation (P <0.05) whereas parameter Tr-N showed non-significant association (P >0.05).

In the case of intra-observer readings (Observer 1), both for initial readings (T0), and after 24 hours readings (T24), the linear parameters Sn-Me, Sn-St, St-Me, Tr-Me, and ChR-ChL showed significant correlations (P<0.05), whereas Tr-N demonstrated a non-significant correlation (P>0.05). In the case of Observer 2, in both initial readings (T0) and after 24 hr readings (T24), the linear parameters Sn-Me, Sn-St, St-Me, Tr-Me showed significant correlation (P<0.05) whereas parameter Tr-N, and ChR-ChL showed non-significant correlation (P>0.05).

The correlation coefficient is displayed in Table 3b along with P-values for different sets of angular parameter values. The inter-observer initial reading shows the correlation coefficient for the two observers presented along with the intra observer 1st reading and 2nd reading correlation coefficient for both the observers. In the case of inter-observer readings, the initial reading (T0), the angular parameters N-Pn-Cm, N-Pn/N-Pog, Li-B-Pog, G-N-Nd, and N-Pn-Pog showed a significant correlation (P<0.05*), whereas parameter Cm-SnLs demonstrated a non-relevant correlation (P>0.05).

In the case of inter observer readings, after 24 hours (T24), all the angular parameters N-Pn-Cm, N-Pn/ N-Pog, Cm-SnLs, Li-B-Pog, G-N-Nd, N-Pn-Pog showed significant correlation (P<0.05*). In case of intra observer readings (Observer 1), in both T0 & T24, the angular parameters N-Pn-Cm, N-Pn/ N-Pog, Li-B-Pog, G-N-Nd, N-Pn-Pog showed significant correlation (P<0.05) whereas, parameter Cm-SnLs showed non-significant correlation (P>0.05). In case of intra observer readings (Observer 2), in T0 and T24, the angular parameters N-Pn-Cm, N-Pn/ N-Pog, Li-B-Pog, G-N-Nd, and N-Pn-Pog showed a significant correlation (P<0.05), whereas parameter Cm-SnLs showed a non-significant correlation (P>0.05).

Table 4 lists the mean and standard deviation of the difference between the linear and angular parameters. In case of all linear parameters, in the initial reading, for inter observer 1 & 2, the overall mean value of difference was 0.1±0.24 and for after 24 hr reading, the overall mean value of differences was 0.09±0.23.

In case of all linear parameters, for intra observer 1 (initial and after 24 hr), the overall mean value of difference was 0.1±0.25 and for inter observer 2 (initial and after 24 hr reading), the overall mean value of difference was 0.1±0.23.

In case of all angular parameters, in the initial reading, for inter observer 1 & 2, the overall mean value of difference was 0.5±0.64 and for after 24 hrs reading, the overall mean value of difference was 0.5±0.61.

In case of all angular parameters, for intra observer 1 (initial and after 24 hr reading), the overall mean value of difference was 0.54±0.65 and for inter observer 2 (initial and after 24 hr reading), the overall mean value of difference was 0.46±0.51.

Discussion

The present study was conducted to evaluate whether performing facial analysis on standardized photographs can achieve the same precision for landmark identification by checking the inter- and intra-examiner correlation coefficients. In this study, measurements were preferred to landmark identification for facial analysis. This is so that treatment planning is done using the measurements, which are the final product of the analysis. Differences in landmark locations can generate measurements with magnitude discrepancies. Therefore, the use of measurements is always given preference over landmarks, especially when used in combination with other landmarks.^17,18

Reliability, which measures agreement between two measurements of the same object, and reproducibility, which verifies repeatability using two methodologies, are the two assessment tools used to evaluate the accuracy of data. In this study, intra-examiner evaluation provided reliability, and inter-examiner evaluation provided reproducibility of the measurements on standardized photographs.

Many studies which performed facial analyses on photographs as well as cephalograms have obtained varying results.^19,20 Rest position of the lips has the highest reproducibility which is in partial consonance with the results of this study where parameters related to upper lip i.e. Nasolabial angle (Cm-Sn-Ls) showed positive correlation on inter examiner evaluation only and the lower lip Mento labial angle (Li-B-Pog) was found to be statistically significant and in good agreement both in intra and inter examiner evaluation.²¹

The research points out that the area around the ears and cheekbones has sparse characteristics that makes it difficult to locate landmarks with accuracy.²² For instance, soft tissue Gonion always shows less reproducibility in landmark identification on photographic analysis or cephalometric tracing which can be attributed to the height and depth in vertical directions of the soft tissue gonion point.²³

Two studies found that the mean total deviation of linear distances in the upper and lower lips was 0.25±0.12 mm for 3D stereophotogrammetric images of the same participant taken at two different time points.^22,23 Another study that used 3dMD scans on the same person at different time points revealed a reliability of 0.2 mm. Maal (±0.39 mm) later validated this in the context of oral and maxillofacial surgery.²⁴ For mannequin head measurements, a mean worldwide error of 0.2 mm was observed and same was confirmed by Aldridge et al. and Weinberg et al.^25,26 In the current investigation, among the six linear calculated measurements, the overall inter observer differences of mean value at T0 was 0.1±0.24 mm and for inter observer reading at T24, the overall difference of mean value was 0.09±0.23 mm. At T0, for intra observer reading, overall mean value of difference was 0.1±0.25 mm and for intra observer reading at T24, the overall difference between the mean values was 0.1±0.23mm, which was well within the acceptable limits.

In frontal photographic analysis, the mid-sagittal plane landmarks in the oro-nasal area (Sn-St, St-Me, Sn-Me) had the greatest accuracy with high correlation, which was statistically significant (P≤0.05) between repeated measurements, indicating that the investigator correctly repeated the measurements; however, Nasion was difficult to locate. According to Sekiguchi and Savara, non-visualization of the landmark may make it challenging to correctly identify the Nasion (N).²⁷ A study by Chang et al. on CBCT derived images suggested that identification errors increased specifically at the Nasion and Orbitale.24 However, Chien et al.²⁶ presented no variation in the nasion, which is contrary to the outcomes of the current investigation where statistically insignificant results were noted in the identification of Nasion both by intra- and inter-examiner evaluations at T0 and T24.

Bilateral landmarks, ChR-ChL, demonstrated a statistically significant correlation (P<0.05) on inter and intra-examiner evaluations, which is contradictory to the study findings, in which the right Chelion showed poor agreement.⁴ Various research using both automated and conventional techniques revealed challenges in identifying the Menton (Me).^27,18,28,29 This is contradictory to the findings of present study as Menton was easily identified in both inter and intra examiner evaluation (P<0.003).

Among the six angular measurements on profile photographic analysis, five measurements showed moderate to good agreement indicating that landmarks were easily identifiable including, N-Pn-Cm, N-Pn/N-Pog, Li-B-Pog , G-N-Nd, N-Pn-Pog (P<0.05). This is in agreement with a study that also reported accuracy in identification of same landmark.³⁰ In case of all angular parameters, the inter-examiner evaluation at T0 showed that the overall mean value of differences was 0.5±0.64 degrees and for inter examiner evaluation at T24, the overall difference of mean values was 0.5±0.61 degrees. Intra examiner evaluation at T0 showed that the overall mean value difference was 0.54±0.65 degrees and at T24, the difference in overall mean value was 0.46±0.51 degrees. The measurement with the highest level of imprecision in the present study was Cm-Sn-Ls on profile analysis and Tr-N on frontal analysis.

Further scope for using such facial photographic analysis in treatment of facial deformities, facial asymmetry, cleft lip and palate, different types of orthognathic surgeries, patients requiring longer monitoring periods should be explored and can be an area of research.

The limitation of the study was that more number of landmarks resulting in more measurements could have been incorporated for analysis. However the reason for not doing so was to avoid examiner fatiguability in locating multiple landmarks in limited time. Thus it was preferred to include 10 landmarks in profile photograph and seven in frontal photograph, to be identified in 10 minutes resulting in six measurements in each photograph.

The findings of this study resonate well with those of a previous study that exhibited high reliability of the measurements.^31-34 The present study revealed that the analysis performed on standardized photographs can provide reliable values both when performed by inter and intra-examination at two-time intervals, T0 and T24. However, inter-examiner error is generally considered greater than intra-examiner error.35 In the current study, both inter- and intra-examiner measurements were statistically significant in terms of correlation.

Conclusion

The present study concluded that the accuracy of facial photographic analysis of standardized photographs provided reliable and reproducible measurements. The error in landmark identification and measurement was observed in profile photographs for Cm-Sn-Ls on profile analysis in inter examiner evaluation at T0 and intra examiner evaluation at T0 and T24.

In frontal photographs, Tr-N showed error in landmark identification and measurement in inter examiner evaluation at T0 and T24 and intra examiner evaluation at T0 and T24. Very high reliability was found for both frontal linear parameters and profile angular parameters, as assessed by the study examiners. The overall mean value of differences in linear parameters was 0.1±0.23 mm and the overall mean value of differences in angular parameters was 0.46±0.51 degrees.

Photographic analysis is an effective and reliable method for identifying and measuring soft-tissue landmarks. Photogrammetry is an upcoming area in the field of orthodontics, cosmetics, and corrective surgical branches for diagnosis, analysis, treatment comparison, outcome, and prediction.

Conflict of Interest

Nil