Real-time tracking of deformable targets in 3D ultrasound sequences

Abstract : Nowadays, mini-invasive treatments, such as radio-frequency ablation, are increasingly being used because they allow eliminating tumors locally from needle insertion. However, the success of these therapies depends on the accurate positioning of the needle with respect to anatomical structures. To ensure correct placement, ultrasound (US) imaging is often used since this system has the advantage to be real-time, low-cost, and non-invasive. However, during the intervention, US imaging can complicate the visualization of targeted structures due to its poor quality and its limited field of view. Furthermore, the accuracy of these interventions may also be perturbed by both physiological movements and medical tools displacements that introduce motions of anatomical structures. To help the surgeon to better target malignant tissues, many research teams have proposed different method in order to estimate the position of regions of interest in ultrasound imaging. This thesis provides several contributions that allow tracking deformable structures in 3D ultrasound sequences. We first present a method that allows providing robust estimation of target positions by combining an intensity-based approach and mechanical model simulation. In this thesis, we also propose novel ultrasound-specific similarity criterion based on prior step that aims at detecting shadows. The last contribution is related to a hybrid tracking strategy that allows improving quality of ultrasound images. From these contributions, we propose a tracking method that has the advantage to be invariant to speckle noise, shadowing and intensity changes that can occur in US imaging. The performance and limitations of the proposed contributions are evaluated through simulated data, phantom data, and real-data obtained from different volunteers. Simulation and phantom results show that our method is robust to several artefacts of US imaging such as shadows and speckle decorrelation. Furthermore, we demonstrate that our approach outperforms state-of-the-art methods on the 3D public databases provided by MICCAI CLUST'14 and CLUST'15 challenges. In this thesis, we also propose an application that combines ultrasound imaging to Magnetic Resonance lmaging (MRI). This method allows observing anatomical structures that are not visible in US imaging during the intervention. It is based on the combination between US tracking method and multi modal registration obtained from external localization system. This application was evaluated on a volunteer thanks to an MRJ imaging platform locate at the University Hospital of Rennes.
Document type :
Theses
Complete list of metadatas

Cited literature [95 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01730334
Contributor : Abes Star <>
Submitted on : Tuesday, March 13, 2018 - 11:08:06 AM
Last modification on : Friday, January 11, 2019 - 2:27:07 PM
Long-term archiving on : Thursday, June 14, 2018 - 1:30:29 PM

File

These_2016ISAR0017_Royer_Lucas...
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01730334, version 1

Citation

Lucas Royer. Real-time tracking of deformable targets in 3D ultrasound sequences. Medical Imaging. INSA de Rennes, 2016. English. ⟨NNT : 2016ISAR0017⟩. ⟨tel-01730334⟩

Share

Metrics

Record views

254

Files downloads

226