Real-time simulation of B-mode ultrasound images for medical training

  • Echtzeit Simulation von B-Mode Ultraschall Bilder zwecks Medizinisches Training

Law, Yuen; Kuhlen, Torsten (Thesis advisor); Cotin, Stéphane (Thesis advisor)

Aachen (2016, 2017)
Dissertation / PhD Thesis

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2016


Ultrasound (US) imaging is a low-cost, non-invasive and non-radioactive technique, often preferred as a mean to explore the body's internal structures during diagnosis procedures. Furthermore, it allows the physicians to view these structures interactively, making it ideal in the guidance of invasive procedures such as Regional Anesthesia (RA) and biopsies, where a needle must be inserted into the patient's body and guided to a specific point. However, obtaining enough training for these procedures is not easy due to the low availability of patients with whom to practice the necessary skills. Moreover, patient safety and comfort are major concerns here. Other traditional training methods include the use of phantoms made of gels or meat and practicing on fellow trainees. Ideally, a training phantom for US procedures would offer variety of scenarios, repeatability and anatomical correctness. In these sense, Virtual Reality (VR) can fulfill these requirements and play an important role in addressing the current challenges in US training.This work presents an US simulation framework aimed to contribute in improving the current training situation by providing the necessary tools to build the aforementioned VR-based training phantoms. In the design of the simulation methodology, attention was focused on reproducing the characteristic features that identify an US image, without compromising too much of the output framerate required for real-time interaction, essential for interactive applications and training purposes. It was also important to provide flexible software interfaces to support the creation of multiple training scenarios, ranging from changing the transducer properties, through creating new patient anatomies, to developing different training tools.The major contribution in this work is the US simulation method and corresponding software framework. The simulation method emulates the functionality of real ultrasound machines to produce the ultrasonic beam, detect echoes and construct the final images. The framework facilitates rapid integration of US simulation in a variety of use cases thanks to its flexible design. An additional contribution, which represents a prerequisite for the simulation, is the review and analysis of the requirements for soft-tissue modeling. This work resulted not only in a list of guidelines to improve existing anatomical models, but also in a set of tested acoustic properties for the most common types of tissue. Furthermore, a description, an implementation and the respective results of the validation and verification process are provided. The inherent complexity of this process, caused by the lack of ground-truth data against to which compare the results, is further discussed. Finally, the work discusses two use case examples in which the simulation framework was used to integrate US synthetic images to develop different training tools.