Compressive sensing methods for multi-contrast magnetic resonance imaging

---

Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Bilgisayar Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2017

Öğrenci: ALPER GÜNGÖR

Danışman: FATOŞ TUNAY YARMAN VURAL

Özet:

Compressive sensing (CS) is a signal processing tool that allows reconstruction of sparse signals from highly undersampled data. This study investigates application of CS to magnetic resonance imaging (MRI). In this study, first, an optimization framework for single contrast CS MRI is presented. The method relies on an augmented Lagrangian based method, specifically alternating direction method of multipliers (ADMM). The ADMM framework is used to solve a constrained optimization problem with an objective function consisting of a linear combination of the total variation on the magnitude image and the l1-norm. Then, a fast implementation is derived for MRI, which requires only two FFT operations per iteration. Second, for better exploitation of sparsity, a joint reconstruction method for multi-contrast CS MRI is presented. This method uses non-convex group-lp-sparsity as well as joint total variation as objective functions. Finally, a joint dictionary learning based method for finding the sparsifying transformation along with the image is presented. The sparsifying transformation reconstructed by the method enforces group sparse representation on all contrast images. All the proposed methods are compared quantitatively and qualitatively with previous methods that exist in the literature using both experimental in-vivo and simulated datasets. The effectiveness of the ADMM for single contrast reconstruction is demonstrated over other single contrast methods. Then, the advantages of using joint reconstruction is discussed and demonstrated. Although dictionary learning based method require high computational cost, it presents benefits in terms of image quality is shown.