Jonathan Immanuel Sperl1, Ek Tsoon Tan2, Kedar Khare2, Kevin F. King3, Xiaodong Tao2, Christopher J. Hardy2, Luca Marinelli2, Marion I. Menzel1
1GE Global Research, Garching, Germany; 2GE Global Research, Niskayuna, NY, United States; 3GE Healthcare, Waukesha, WI, United States
Diffusion spectrum imaging provides radial information of diffusivity in the brain, such as diffusional kurtosis. This work presents a two step quadratic programming framework for fitting the diffusion and kurtosis tensors. Furthermore, the effects of using undersampled data and compressed sensing recon-structions are investigated. Various derived scalar measures for kurtosis are compared for a human brain data set. The results show the robustness of the fitting procedure as compared to standard linear fitting. Compressed sensing allows for either faster acquisitions or improved image quality, while providing some degree of denoising.