Datenbestand vom 03. Mai 2021

Warenkorb Datenschutzhinweis Dissertationsdruck Dissertationsverlag Institutsreihen     Preisrechner

aktualisiert am 03. Mai 2021

ISBN 9783843934398

84,00 € inkl. MwSt, zzgl. Versand


978-3-8439-3439-8, Reihe Medizintechnik

Thomas Küstner
Motion Correction of Positron Emission Tomography Data by a Magnetic Resonance Imaging derived Model

244 Seiten, Dissertation Universität Stuttgart (2017), Softcover, A5

Zusammenfassung / Abstract

In the field of oncology simultaneous Positron-Emission-Tomography/Magnetic Resonance (PET/MR) scanners offer a great potential for improving diagnostic accuracy. PET acquisitions of the body trunk in the range of several minutes are affected by motion originating from the patient, mainly respiration and cardiac motion which manifests as blurring of the lesions and impairs the diagnostic quality.

The proposed method uses the simultaneous MR acquisition to perform an MR-based non-rigid motion correction of the PET data.

The concept of Compressed Sensing smoothly integrates into this purpose of dynamic MR imaging. A 3D high-resolution dynamic MR imaging sequence is developed which samples the Cartesian k-space in a random manner and incorporates self-navigation signals. Since tissue boundary integrity is highly desired for motion correction, a Partial Fourier-like sampling space compression (ESPReSSo) is proposed which samples high frequency areas denser leading to increased edge delineation. Furthermore, signal to noise ratio can be improved if a-priori knowledge about the underlying tissue distribution is integrated into the sampling mask generation. The randomness of this sub-Nyquist sampled MR trajectory allows to retrospectively gate the acquired samples into their respective motion state. Gating is conducted by means of surrogate signals which capture the true underlying motion. In order to provide full PET scan time coverage and decisive motion state estimation, different surrogate signals (MR navigators, camera, respiratory belt, ECG) are combined by a sensor fusion approach. The subsampled and motion-gated k-spaces are reconstructed by a motion-estimated Compressed Sensing algorithm to obtain aliasing-free and motion-resolved MR images. An efficient optical-flow based image registration algorithm (LAP) is proposed which allows to retrieve a motion model during reconstruction from the MR images. The motion model is afterwards applied to a listmode-based and motion-compensated PET reconstruction yielding the final motion-corrected PET image. The whole process is set up online on the scanner to enable a streamlined processing in a clinically feasible manner. The proposed method is publicly available.

A cohort study on 36 patients with suspected liver or lung metastasis revealed an improvement of the motion-corrected PET to an uncorrected PET in terms of lesion quantification by 22%, lesion delineation by 64% and diagnostic confidence by 23%.