Création de modèles surfaciques personnalisés et sans contact pour l’estimation précise des surfaces corporelles saines et brûlées en contexte clinique

Abstract

The care of a severely burned patient is a complex medical challenge. It requires expertise to accurately assess the %TBSA, which refers to the ratio of burned surface area to the total body surface area (BSA). This percentage is crucial for determining the appropriate treatments to be provided to the patient (e.g., fluid resuscitation, medication, and surgical procedures). Current clinical methods, which are in 2D and standardized, lack precision and show inter- and intra-evaluator variability depending on the clinician’s level of expertise. Alternative methods have emerged, particularly with advances in 3D modeling tools, allowing for three-dimensional modeling of the whole-body surface. However, global 3D model adjustment alone is insufficient to ensure the precision required for clinical applications. It is necessary to personalize the morphology, body composition (e.g., fat and muscle mass), and local dimensions to obtain a locally personalized 3D model and accurately estimate the %TBSA for tailored patient care. Furthermore, 3D tools and the required data are not always suited to the clinical context of burn patient management. In this context, the objective of this thesis is to develop automated methods for contactless, personalized 3D modeling of the entire human body, with accurate BSA and local body areas to estimate the %TBSA. To achieve this, two complementary methods were developed: the STAP and S3D methods, each incorporating a statistical model—specifically, a body composition statistical model and an anthropometric statistical model. The STAP method personalizes global and local surface areas using a few parameters already available in clinical settings, by integrating body composition, global and local fat and muscle mass. The S3D method refines the 3D model generated by the STAP method using a partial 3D scan of the patient to extract a few anthropometric measurements and personalize specific body regions. The Mean Absolute Percentage Error (MAPE) of the BSA is approximately 2% for both methods, which is twice as accurate as existing clinical and 3D methods. Another contribution of this work is the validation of the accuracy of regional body surface areas. The S3D method achieves MAPE values ranging from 4% to 6% for the torso, arm, and leg regions, allowing for %TBSA estimations with MAPEs between 2% and 6%, lower than the 10% error observed in existing 3D modeling tools such as BurnCase 3D. In conclusion, this thesis has contributed to the development of two whole-body 3D modeling methods, capable of representing the diversity of body types, morphologies, and local body shapes. This level of personalization supports tailored treatment for each patient, improves their chances of survival, and reduces the risks of medical complications and long-term sequelae.

Date	27 Jun 2025
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Carlos Vázquez (Supervisor) & Jacques A. de Guise (Co-supervisor)