Antropometria tridimensional da face: novas alternativas de captura e processamento de imagens 3D

Abstract

Cosmetic facial surgeries are procedures with a high degree of complexity, with Rhinoplasty being one of the five most performed surgeries in the world. The preoperative clinical evaluation, together with the aesthetic and anthropometric facial analysis, are essential for a satisfactory result and allow the alignment of expectations between the patient and the surgeon. Three-dimensional (3D) anthropometry has proven to be an effective and safe method for measuring facial measurements. However, the high cost of the most widespread and validated 3D capture systems is one of the causes that depart and inhibit surgeons from incorporating this type of technology into clinical practice. In order for the method to spread, it is necessary to use a low-cost, portable and easy-to-use form of capture. The objective of this work was to establish and execute a protocol for epidemiological evaluation and capture of three-dimensional (3D) images of the face for potential use in 3D anthropometric analysis, and also to verify the reliability of the captured images, the quality of the different capture modes and the relevance in clinical practice. For this work, 3D images of the faces of selected patients were generated through the use of different technologies: photogrammetry in the open source software Blender and its add-on OrtogOnBlender (OOB), 3D reconstruction of images of Computed Tomography (CT) of breasts of the face and the infrared scanning available on the smartphone through the Scandy Pro application. Three sequential photographic sessions (Session A, B and C) and three facial scanning sessions (Session D – Phase 1, 2 and 3) were carried out, with the production of seven 3D facial meshes of the participants, with the reference mesh being that of CT soft tissue surface. The comparative analysis of the 3D meshes was based on the ICP (Iterative Closest Point) algorithm implemented through the CloudCompare open source software. These results were analyzed by four specialists with experience in the subject and by one of the OOB developers. The sample comprised 19 patients, mostly female, whose most common symptom was nasal obstruction and aesthetic dissatisfaction with the profile of the nose. The Blender software photogrammetry with the highest number of photos (session C) had better resolution, greater similarity with the reference 3D mesh and more satisfactory according to the experts. The OOB developer recommended taking the captures according to the tool's official protocol, noting that constant lighting, adequate focal length and the number of photographs are some of the essential aspects for a good photogrammetry result. Facial scanning with the Scandy Pro application generated meshes with a higher density of points, but without significant difference with the other pairings. Most of the mean distances between the captured point clouds and the CT reference were smaller than 1.5mm, thus providing compatibility between the meshes. Therefore, photogrammetry and facial scanning methods have great potential for clinical use, but further studies are needed to define precision, accuracy and scientific validation of the methods.