Nonlinear dipole inversion (NDI) enables robust quantitative susceptibility mapping (QSM)

• Verfasst von: Polak, Daniel [VerfasserIn]
• Verfasst von: Chatnuntawech, Itthi [VerfasserIn]
• Verfasst von: Yoon, Jaeyeon [VerfasserIn]
• Verfasst von: Iyer, Siddharth Srinivasan [VerfasserIn]
• Verfasst von: Milovic, Carlos [VerfasserIn]
• Verfasst von: Lee, Jongho [VerfasserIn]
• Verfasst von: Bachert, Peter [VerfasserIn]
• Verfasst von: Adalsteinsson, Elfar [VerfasserIn]
• Verfasst von: Setsompop, Kawin [VerfasserIn]
• Verfasst von: Bilgic, Berkin [VerfasserIn]
• Titel: Nonlinear dipole inversion (NDI) enables robust quantitative susceptibility mapping (QSM)
• Verf.angabe: Daniel Polak, Itthi Chatnuntawech, Jaeyeon Yoon, Siddharth Srinivasan Iyer, Carlos Milovic, Jongho Lee, Peter Bachert, Elfar Adalsteinsson, Kawin Setsompop, Berkin Bilgic
• E-Jahr: 2020
• Jahr: 20 February 2020
• Umfang: 13 S.
• Fussnoten: Gesehen am 07.01.2021
• Titel Quelle: Enthalten in: NMR in biomedicine
• Ort Quelle: New York, NY : Wiley, 1988
• Jahr Quelle: 2020
• Band/Heft Quelle: 33(2020,12) Artikel-Nummer e4271, 13 Seiten
• ISSN Quelle: 1099-1492
• DOI: doi:10.1002/nbm.4271
• Datenträger: Online-Ressource
• Sprache: eng
• Sach-SW: deep learning
• Sach-SW: nonlinear inversion
• Sach-SW: quantitative susceptibility mapping
• K10plus-PPN: 1744153590

Abstract

High-quality Quantitative Susceptibility Mapping (QSM) with Nonlinear Dipole Inversion (NDI) is developed with pre-determined regularization while matching the image quality of state-of-the-art reconstruction techniques and avoiding over-smoothing that these techniques often suffer from. NDI is flexible enough to allow for reconstruction from an arbitrary number of head orientations and outperforms COSMOS even when using as few as 1-direction data. This is made possible by a nonlinear forward-model that uses the magnitude as an effective prior, for which we derived a simple gradient descent update rule. We synergistically combine this physics-model with a Variational Network (VN) to leverage the power of deep learning in the VaNDI algorithm. This technique adopts the simple gradient descent rule from NDI and learns the network parameters during training, hence requires no additional parameter tuning. Further, we evaluate NDI at 7 T using highly accelerated Wave-CAIPI acquisitions at 0.5 mm isotropic resolution and demonstrate high-quality QSM from as few as 2-direction data.