Technical validation of the Di3D stereophotogrammetry surface imaging system

https://doi.org/10.1016/j.bjoms.2007.09.005Get rights and content

Abstract

The purpose of this work was to assess the technical performance of a three-dimensional surface imaging system for geometric accuracy and maximum field of view. The system was designed for stereophotogrammetry capture of digital images from three-dimensional surfaces of the head, face, and neck. A mannequin head was prepared for imaging by adding texture in the form of red paint, and facial landmarks as black ink dots. The mannequin was imaged at the manufacturer's recommended settings for human studies. Colour-coded surface difference images among repeated exposures were computed. We compared measurements of physical linear distance with digital measurements. The three-dimensional stereophotogrammetry system had a mean error in the three-dimensional surfaces of 0.057 mm, a repeatability error (variance) of 0.0016 mm, a mean error of 0.6 mm in linear measurements compared with manual measurements, and a field of view of 170° horizontally and 102° vertically.

Introduction

There has been considerable technological development and application of three-dimensional surface imaging in a range of clinical areas. Clinically, this technique has been used by maxillofacial surgeons,1 orthodontists,2 and psychiatrists.3 Detailed surface models of the human face have been used to record facial morphology,4 monitor changes in growth over time5 and for the characterisation of genetic abnormalities.6 It is clear that these non-invasive, fast and patient-friendly technologies have a part to play in clinical decision-making. In clinical practice three-dimensional images are useful in planning treatment, follow up of operations, and they have potential research applications. Facial deformity, whether developmental (cleft lip) or acquired (post-traumatic), presents challenges in diagnosis and planning treatment because of the complex three-dimensional nature of the defects. The ability to analyse such defects in terms of asymmetry and volume gives the clinician additional information that can be used to plan corrective operations. Assessment of cleft lips preoperatively is reliable and readily applied to potentially uncooperative infants.7 There is also potential for three-dimensional images to be used in research into areas such as postoperative swelling and changes in facial morphology after orthognathic surgery.

Cephalometry has used direct anthropometry (linear measurements of the subject), conventional photography, radiographic imaging, and more recently three-dimensional surface imaging. There are various products available to capture digital information and reconstruct a detailed surface contour of the head, neck, and face. The four categories are; laser scanning, structured light, moiré fringes and three-dimensional stereophotogrammetry.8 Three-dimensional stereophotogrammetry has emerged as a new technique for capturing fully-textured (colour information) three-dimensional surfaces and has the potential to be applied easily clinically. It is based on the acquisition of two stereoscopic views that can calculate distance using the geometry of the camera systems and triangulation.8 Three-dimensional stereophotogrammetry relies on the system being able to match corresponding points from two separate images. This means that the source images should be of high quality to enable well-defined features from each image to be identified and matched.

A range of studies have been described that have assessed the technical validity of surface scanning systems. One variable used to define performance is accuracy. This is normally quoted as a mean error measurement when measurements of a real object are compared to the three-dimensional digital model. Laser scanning system studies have a mean (SD) accuracy of 1.9 (0.8) mm and 1.1 (0.3) mm for the Minolta 7009 and 90010 laser scanning systems. A further study showed that the accuracy may be as high as 0.56 (0.25) mm with a left/right registration error of 0.13 (0.18) mm.11 Three-dimensional stereophotogrammetry has been compared with direct anthropometry using a mannequin head and found to produce errors of under 1.0 mm.12 One of the important conclusions from this work was that, although there were significant differences in measurements between each of the two camera systems, as well as differences between caliper-based measurements made directly on the mannequin, they were so small as to be clinically irrelevant (defined as <1.0 mm).

The three-dimensional surface capture system (product name Di3D), by Dimensional Imaging, Glasgow (http://www.di3d.com), uses three-dimensional stereophotogrammetry to produce fully textured three-dimensional surface contour maps of the head and face (180° ear to ear view). The system captures two stereo pairs of images (4 cameras in total) and specialist software is used to create a three-dimensional surface using triangulation. The purpose of the present work was to assess the geometric accuracy of a new three-dimensional surface imaging system.

Section snippets

Materials and methods

This three-dimensional stereophotogrammetry system, capture software V3.1.22 and reconstruction software V3.1.6, uses 8 megapixel Canon EOS 350D digital cameras with 50 mm lenses. A mannequin head made of white polystyrene was imaged, with red paint applied to provide visual texture to the surface. Eighteen anatomical landmarks were identified with black dots using a fine-point ink pen on the mannequin head. All dots were less than 0.5 mm in size. Fig. 1 shows the mannequin head with its textured

Experiment 1

Fig. 3 shows the mean difference of the 10 repeated exposures compared with the mean surface. The colour coding (in mm) is shown at the bottom of the figure. Fig. 3(a) shows there was a mean error of 0.057 mm with a maximum error of 1.06 mm (in red) observed along the midline of the mannequin head and at the extreme edges. Fig. 3(b) shows the same data rotated 45° to the left. This illustrates that the error increases at the extreme edges of the head. The mean variance of the errors was 0.003 mm,

Discussion

There is a range of methods for assessing the performance of surface scanning systems. We have adopted an approach that used a static mannequin head, which allowed us to assess repeatability and geometric accuracy of the system under normal operating conditions. Results from experiment 1 showed that the system is capable of measuring the same object to a high degree of repeatability, making the system suitable for longitudinal clinical studies. There was no evidence of non-uniformity or

References (13)

There are more references available in the full text version of this article.

Cited by (72)

  • Reliability of optical devices for three-dimensional facial anatomy description: a systematic review and meta-analysis

    2020, International Journal of Oral and Maxillofacial Surgery
    Citation Excerpt :

    Therefore, the applicability of the current results to people other than collaborative adult volunteers is still to be verified24,61. The sample size ranged from one mannequin head13,22,54,58,59,61 to 18 mannequin heads56, and from five subjects52 to 80 subjects49. In most cases, there were fewer than 10 subjects or inanimate specimens8,10,13,19,21,48,52,60 or between 10 and 304,5,12,15,26,50,53,55–57.

View all citing articles on Scopus
View full text