Creating a statistical shape model to aid in the estimation of incomplete soft tissue segments of the surface of South African faces

Abstract

Introduction: A critical gap exists in population-specific data for facial morphology of black South Africans which are essential for the accurate reconstruction of facial features in fields such as aesthetic and reconstructive surgery, prosthodontics and extra-oral facial prosthetics, as well as forensic facial approximations. The objectives of this research were to generate normative reference values of black South African faces for various inter-landmark distances, and to derive a statistical shape model (SSM) of 3D facial shape variation which can be applied to estimate missing soft tissue segments on simulated defective faces. Methods: The study included of 235 computed tomography (CT) and cone-beam computed tomography (CBCT) scans from black South African individuals between the ages of 18 and 87 years. The scans were collected from retrospective records of three medical institutions and excluded individuals that showed conditions potentially affecting facial morphology, including orthodontic treatments, pathological conditions, facial asymmetry, or any history of facial reconstructive surgery. The scans were processed to obtain 3D facial meshes and landmarks were placed at anatomically important loci. For the first objective, inter-landmark distances were calculated, statistically analysed, and compared to published literature on other populations. For the second objective, correspondence of the 3D meshes utilising the landmarks were achieved, and generalised Procrustes analysis and principal component analysis conducted. These steps are crucial in obtaining an SSM comprising the modes of variation and the normal range variance along each mode, which together defines multinormal parameterisation of shape variation. Defect estimations were done by using the SSM to estimate the linear combination of the modes of variation that most closely approximates the intact regions of each face, and estimate the missing regions using a weighted projection onto the modes of variation. Results: Chapter 3 reports on normative facial capulometric measurements specific to the black South African population. It highlights significant differences in facial parameters between sexes and between different populations. The data reveal notable similarities with other African populations, especially in oral features, but significant disparities with non-African groups. Chapter 4 introduces the innovative SSM for extra-oral prosthetic design. This model accurately estimates missing soft tissues, demonstrating a high degree of precision with root mean square errors consistently below 2.58 mm for various facial defects.

Conclusion: The normative measurements highlight the unique facial characteristics of the black South African population, demonstrating the necessity of population-specific data in clinical and forensic applications. The development of the SSM represents a novel advancement in digital reconstruction methodologies. It offers a more objective and patient-specific approach in prosthetic design, especially in addressing complex facial defects such as bi-orbital defects or those crossing the facial midline in a demographic that has been largely overlooked in previous research. This model, by reducing the subjectivity and artistic skill previously required in prosthetic design, aligns with the evolving digital trends in medical technology and aims to address specific local needs and challenges in South Africa, and also have potential for global application.