Model-based ultrasound characterization of the interface between bone tissue and a dental implant – DynImplant

Implants are widely used in oral surgery. However, there remain risks of failure, which are difficult to anticipate. The main determinant of the implant surgical success is the implant stability at insertion and healing stages. The causes for implant failures, which depend on the biomechanical properties of the bone-implant interface (BII), remain unclear due to the lack of information on the BII properties.

A prototype of quantitative ultrasound (QUS) device has recently been developed to evaluate implant stability. This device aims to assess both primary and secondary stabilities of dental implants by employing QUS to monitor the evolution of the BII biomechanical properties. However, these techniques are still immature because the multiscale behavior of the BII remains unclear, especially, when dynamic effects become significant. Moreover, only empirical relationships between the BII properties and its ultrasound response could be established. Rigorous mathematical methods and low-cost forward/inverse algorithms allowing access to the BII properties using a limited number of in vivo measurements are needed to go beyond this empirical stage.

The general objective of DynImplant is to formulate (1) enhanced physical models describing the BII’s multiscale dynamic behavior; (2) adapted and efficient numerical methods for the direct simulation of the interaction between ultrasonic waves and the BII; (3) a robust and accurate full-waveform inversion procedure based on an original data misfit functional, which will be optimized by employing adapted machine learning-based strategies; (4) we aim at validating these model experimentally in a preliminary clinical trial. Three mathematical developments will be performed:

First, we aim to derive biomechanical models describing the dynamic multiscale behavior of the BII as function of the implant environment at different frequency ranges. The BII model must describe (i) the micro-structured and evolutive nature of bone tissue and (ii) the dynamic friction/adhesion phenomena at the complex BII in the context of a multiscale roughness. Dynamic homogenization techniques will be employed.

Second, we aim to develop advanced computing solvers for the forward simulation of the multiscale scattering phenomena of ultrasonic waves at the BII. An explicit high order finite element method using an isogeometric analysis will be employed. The multiscale surface roughness as well as the physical models of BII (see the first objective) will be introduced.

Third, we aim to find a robust model-based inversion procedure allowing retrieve in real-time the status of the bone-implant contact. The inversion scheme needs to be able to efficiently analyze and classify the signals recorded by experiments with respect to multi-parameter and big dataset generated from forward simulations. An inverse surrogate-based optimization algorithm using machine learning technique will be developed.

The performance of the developed physical and numerical models will be validated by comparing with data obtained in vitro and in a preliminary clinical trial.

In conclusion, DynImplant proposes to tackle several unsolved problems with open fundamental questions. The originality lies in the synergy between acoustical modeling, high performance computing technology and clinical experiments. DynImplant will lead to a better understanding of the causes of implant failure as well as to the development of new characterization methods of the bone-implant interface. The success of DynImplant will be obtained thanks to the complementarity of the consortium in: multiscale modeling (MSME) and ultrasonic wave propagation simulation (MSME, LAGA), inversion problems (LAGA), implant surgery (CHU-Nantes) and medical device development (WaveImplant).