Philippe Garteiser1, Sabrina Doblas1,
Bernard E. VanBeers1,2, Valrie Vilgrain2, Ralph Sinkus1
1INSERM UMR 773, Centre de
Recherche Biomdicale Bichat-Beaujon, Clichy, France; 2Department
of Radiology, Beaujon University Hospital, Paris Diderot University, Clichy,
France
The conventional reconstruction of viscoelastic parameters in MR-Elastography (MRE) is challenging in the presence of the signal noise levels typically encountered in clinical datasets. A novel algorithm is proposed, which stabilizes the noisy wave equation inversion problem by injecting constraints arising from the underlying physics into the system. As a result, the viscoelastic parameters accurately reflect the input data while also obeying necessary basic physical constraints, gaining stability in the process. Proof of concept is provided on simulated noisy waves and on clinically relevant experimental data from a cohort of patients with diffuse liver disease.