A 3D facial prediction methodology for patients undertaking dental malocclusion correction surgeries

In recent years, the development of 3D technology has found large applications in the medical field, from new diagnostic 3D imaging techniques to surgery. The use of 3D technologies plays an essential role in pre-operative planning because it allows you to obtain high-precision reconstructions of anatomical parts and plan interventions in a targeted and specific way. 3D reconstructions in medicine deriving from CT images can be Volume Rendering type, useful for surgical planning and Surface Rendering type, meshes obtained through segmentation, which can be post-processed for finite element analyses or to perform displacement simulations, as in the case of osteotomies for bone reconstructions. The 3D modelling of the facial compartment is essential in the orthodontic field and maxillofacial surgery. In addition to clinical planning, it allows a post-operative face reconstruction in advance. This paper aims to define a procedure to create a predictive model of what the future face will be and therefore let surgeons know in advance the changes in the tissues involved to perform targeted and customized interventions and, consequently, to show the patients what the post-operative face will look like, a critical aspect for the communication between doctor and patient. The subject of the study is five patients of “San Giovanni Battista” Hospital in Turin with dental malocclusion who have undergone BSSO or LeFort I operations. Often accompanied by disorders of mandibular retrusion and protrusion, the face has flat cheekbones, making it necessary to treat zygomatic valgus further. The pre-and post-operative images of the subjects analyzed were acquired with the CBCT 3D technology, which allows obtaining a remarkable biometric precision with low doses of radiation transferred to the patient compared to traditional CT. After several steps of processing the 3D images in DICOM format and after a transformation into a square grid of the pre-and post-operative surfaces, it was possible to calculate the displacements undergone by the soft tissues by analyzing not only the parts involved in the intervention but also the involuntary adaptive shifts of the whole face. From the manual positioning of 19 landmarks on the soft tissues according to the Swennen, the surroundings of these were identified, and the displacements suffered after the surgery were calculated. The search for the similar patient was implemented through two methods: starting from a subject as a test sample, for the first method, a non-parametric statistical test was carried out, analyzing the characteristics of the surfaces through three geometric descriptors, atanF, Fden2 and Sfond1; for the second method, the Euclidean distance between the points on the surface was calculated. Having identified the similarity for the entire face and particularly for the malar area, the displacements suffered by those who had a similar result were applied to the pre-operative surface of the test sample. Once the 3D point cloud of the new face was obtained, a linear interpolation was carried out to get the surface of the predicted face. Subsequently, we passed to the verification of the model in which the predicted face was compared with the real post. It was possible to obtain a predictive model highly faithful to the real face. In addition, thanks to 3D surfaces, a method that can reconstruct a missing part of the face due to unclear CT images has been developed. All steps of the methods described were developed in the MatLab environment.