Optimal DSI reconstruction parameter recommendations: Better ODFs and better connectivity

- DSI as implemented by public software is not optimal.
- Optimal DSI parameters are given depending on the SNR of your dataset.
- Better ODF and better connectivity can be obtained from existing datasets.
- Theoretical limitations of DSI are discussed and shown mathematically.

Diffusion Spectrum Imaging (DSI) has been used for tractography in several publicly available software and a number of recent high impact publications. However, there are several important theoretical, numerical and practical considerations that are often ignored. We revisit the theoretical and state-of-the-art processing steps necessary to go from the DSI signal to the diffusion orientation distribution function (dODF) used by tractography. We show that the parameters in the reconstruction have huge impact on the reconstruction quality and that, while there is no consensus about what they should be, the parameters we most often see in the literature are not optimal. We provide applicable recommendations that improve the accuracy of extracted local orientations and improve accuracy of global connectivity as measured by the Tractometer, a tractography online evaluation system. These recommendations come for “free” as they are applicable to all existing DSI data and do not require a significant increase in computation time. Hence, this paper highlights the do's and dont's of DSI reconstruction.