3 Bayes theorem 1 2 1 2 3

3 Bayes' theorem 1 … 2 1 2 3 Partial discreteness: a new type of prior knowledge for MRI reconstruction Gabriel Ramos-Llordén 1, Hilde Segers 1, Willem J. Palenstijn 1, Arnold J. den Dekker 1, 2 and Jan Sijbers 1 1 i. Minds-Vision Lab, University of Antwerp, Belgium. 2 Delft Center for Systems and Control, Delft University of Technology, Delft, The Netherlands.

Introduction Breast implant Dental MRI 1 FLAIR sequences 2 Angiography 3 4 Some regions are approximately constant in intensity Partial discrete images: piece-wise constant part + texture part Partial discreteness as a prior for ill-posed reconstruction problems 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 1/12

Partial discreteness model Phase Variant intensity class 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 5 2/12

Partial discreteness model Phase Variant intensity class 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 6 3/12

Penalized iterative reconstruction image k-space data Fourier matrix 3417 Regularization parameter Discreteness error Partial discreteness: a new type of prior knowledge for MRI reconstruction 4/12

Bayesian segmentation operator ch ara Past cte riz a tio n 2 Bayes' theorem 1 K-Gaussian mixture model fitting … 3 1 3 2 A posteriori probability maps [Caballero J. , MICCAI 2014] 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 5/12

Bayesian segmentation operator ch 2 3417 1 tio n … 3 Temporal regularization ara Past cte riz a 1 3 2 Partial discreteness: a new type of prior knowledge for MRI reconstruction 5/12

Bayesian segmentation operator Otsu thresholding 3417 Estimated partially discrete image Partial discreteness: a new type of prior knowledge for MRI reconstruction 6/12

Bayesian segmentation operator Discreteness error: 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 7/12

Experiments • Simulations with breast implant and angiography data • Single coil radial k-space sampling with varying number of spokes, • Smoothly varying phase added • Comparison against Conjugate Gradient (CG) with smoothness prior and Total Variation (TV) [Gai. J. et al. (Impatient Toolbox), ISMRM 2012] 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 8/12

Results Breast implant experiment Recovered images and implant contour detection (a) CG + smoothness 3417 (b) CG + TV (c) Proposed Partial discreteness: a new type of prior knowledge for MRI reconstruction 9/12

Results Breast implant experiment: segmentation metrics 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 10/12

Results Angiography experiment (a)Original 3417 (b)CG + smoothness (c)CG+TV (d)Proposed Partial discreteness: a new type of prior knowledge for MRI reconstruction 11/12

Conclusions Partial discreteness prior More detailed reconstructed images Segmentation benefits from partial discreteness Thanks for your attention! Contact: http: //visielab. uantwerpen. be/people/gabriel-ramos-llorden 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction 12/12

Image references 1. 2. 3. 4. 5. Radiopedia. org http: //www. drbicuspid. com/ www. reviewofoptometry. com https: //www. healthcare. siemens. com/magnetic-resonance-imaging/ options-and-upgrades/clinical-applications/advanced-angio 6. M Maijers, Ph. D Thesis 3417 Partial discreteness: a new type of prior knowledge for MRI reconstruction