Accuracy and Reliability of Kinect Motion Sensing Input Device’s 3D Models: A Comparison to Direct Anthropometry and 2D Photogrammetry

Abstract

AIM: This study aims to evaluate the accuracy and reliability of Kinect motion sensing input device’s three-dimensional (3D) models by comparing it with direct anthropometry and digital 2D photogrammetry. MATERIALS AND METHODS: Six profiles and four frontal parameters were directly measured on the faces of 80 participants. The same measurements were repeated using two-dimensional (2D) photogrammetry and (3D) images obtained from Kinect device. Another observer made the same measurements for 30% of the images obtained with 3D technique, and interobserver reproducibility was evaluated for 3D images. Intraobserver reproducibility was evaluated. Statistical analysis was conducted using the paired samples t-test, interclass correlation coefficient, and Bland-Altman limits of agreement. RESULTS: The highest mean difference was 0.0084 mm between direct measurement and photogrammetry, 0.027 mm between direct measurement and 3D Kinect’s models, and 0.018 mm between photogrammetry and 3D Kinect’s. The lowest agreement value was 0.016 in the all parameter between the photogrammetry and 3D Kinect’s methods. Agreement between the two observers varied from 0.999 Sn-Me to 1 with the rest of linear measurements. CONCLUSION: Measurements done using 3D Images obtained from Kinect device indicate that it may be an accurate and reliable imaging method for use in orthodontics. It also provides an easy low-cost 3D imaging technique that has become increasingly popular in clinical settings, offering advantages for surgical planning and outcome evaluation.

Facial scanning technologies in the era of digital workflow: A systematic review and network meta-analysis

“…Among the 29 studies included in the systematic review, different study groups were identified: the most common control group was direct anthropometry. Among the studied groups, static-stereophotogrammetry was the most represented with 14 study groups [9,10,19,[23][24][25][26]29,31,36,38,43]; eight studies investigated infrared structured light [9,12,35,38,39,[42][43][44][45]; six studies considered portable-stereophotogrammetry [32][33][34][39][40][41] and structured light [9,20,28,30,37,38], while laser scanners [18,21,22,27,40] were investigated in five groups. Only one study reported the use of LED structured light [9].…”

“…The measurement technique and number and type of landmarks used for the assessment were collected. Seventeen studies [12,[18][19][20][22][23][24][25][26][27][28]31,32,[36][37][38][39] investigated only linear measures using a range of 6-48 landmarks, four studies [21,40,43,44] investigated both linear and angular measures, and four [34,42,45] investi-gated both linear measures and surface-to-surface deviation. Three types of measurements were employed in two articles [29,33], and two [9,30] additionally investigated surface-to-surface deviation only.…”

“…Structured light requires 20 s and 10 s [28,30,37,38], portable-stereophotogrammetry [32][33][34][39][40][41][42], and new laser scanners [22,27,40]. In the case of infrared structured light, scanning times varied between 5 and 30s [9,12,30,35,38,39,[42][43][44][45] because they were influenced by the presence of a single unit or several units, which considerably reduced the time. It also depends on the software used for scanning and the type of device that supports the application.…”

“…SL and p-SP fall into the affordable category, and ISL, used in the most recently published articles, were the cheaper technologies. However, when ISL technology is used by the implementation of multiple units, the cost increases slightly and the technology falls into the affordable category [44].…”

“…In other studies, the Domain 3 was considered "Unclear" because the DA measures, considered the gold standard technique, were not reported [28,33,43] and these were essential to have a reference to compare the data of the test group. Four studies [19,22,26,44] were judged "Unclear" on Domain 4 because the possibility of unusable facial scans was not resolved clearly. Some concerns regarding applicability were met in one study [27] regarding the patient selection domain.…”