Modélisation et compensation des déformations mécaniques des tissus dues à la pression d’une sonde échographique

Abstract

The acquisition of good quality ultrasound images requires good acoustic coupling between the ultrasound probe and the patient's skin. In practice, this good coupling is achieved by the operator by applying a force to the skin through the probe, thus removing any possible layer of air between the probe and the skin. This force has the effect of deforming the tissues underlying the probe. The ultrasound images thus acquired represent deformed tissues. In the case of certain applications, such as a diagnosis carried out on a 2D image or a simple check of the fetus as in gynecology, this deformation of the tissues does not affect the relevant information used by the operator. However, for other applications, these deformations can hinder the quality of the relevant information for the operator. This is the case, for example, in free-hand 3D ultrasound, which consists of moving a conventional probe through an area of interest, then concatenating all the images acquired to reconstruct an ultrasound volume. Indeed, throughout the movement of the probe, the different images were acquired with a variable probe/skin force, inducing different deformations from one image to another. This difference in tissue deformation from one location to another generates discontinuities from one image to another during the 3D reconstruction, thus creating artifacts. These artifacts deteriorate the quality of the ultrasound image, limiting its spectrum of use. The main objective of this doctorate is to compensate for the deformations induced by the contact force between the probe and the skin during the acquisition of ultrasound images. This correction would therefore allow to obtain an ultrasound volume free of deformations, and which would have been acquired with zero force, either without contact between the probe and the skin, or with a uniform force. To meet this objective, we have developed two deformation compensation methods based on the construction of a biomechanical model. The general principle of these two methods consists in building a 2D biomechanical model representing the mechanics of the soft tissues involved in the image, then in estimating a field of realistic displacement undergone by the tissues following the indentation of an ultrasound probe. The inverse displacement field is then applied to the distorted image to compensate for the image distortion. Both methods consist in optimizing two key parameters of a finite element biomechanical model, which are the indentation of the probe, and the elasticity ratio. For the first method guided by the similarity score, the aim is to optimize these two parameters to maximize the normalized intercorrelation between the deformed image and a reference image. The second method consists in optimizing these two parameters to minimize the difference between the displacement field estimated by the model and that estimated by an image-based registration method. These two methods have been validated on simulated images, as well as experimental images. The performance of these methods was evaluated and compared to an image-based deformation correction method. Another line of research arising directly from the need for experimental validation concerns the design of an ultrasonic phantom. Indeed, this work provide a model to predict the elastic modulus of a soft tissue-mimicking phantom based on two very easily controllable parameters: gelatin concentration and refrigeration duration. The tissue-mimicking phantom is made following a low-cost and simple fabrication procedure using commercial household ingredients. A large range of elastic properties can be obtained (15-100kPa) with the proposed recipe. In conclusion, this work allowed the evaluation of three methods of to correct deformations due to the pressure of the ultrasound probe, but also provided a simple, fast and inexpensive recipe to fabricate phantoms mimicking biological soft tissues reproducing a targeted modulus of elasticity with a high level of confidence.

Date	30 Jan 2023
Original language	French
Awarding Institution	École de technologie supérieure
Supervisor	Catherine Laporte (Supervisor) & Yvan Petit (Co-supervisor)