Validation of structural brain connectivity networks: The impact of scanning parameters

• DOI: 10.1016/j.neuroimage.2019.116207
• Publication Date: 2019-09-17T23:33:48Z
• Authors: Karen S. Ambrosen; Simon F. Eskildsen; Max Hinne; Kristine Krug; Henrik Lundell; Mikkel N. Schmidt; Marcel A.J. van G...; Morten Mørup; Tim B. Dyrby
• Keywords: Gold standard (test)

Evaluation of the structural connectivity (SC) brain based on tractography has mainly focused choice diffusion model, algorithm, and their respective parameter settings. Here, we systematically validate SC derived from a post mortem monkey brain, while varying key acquisition parameters such as b-value, gradient angular resolution image resolution. As gold standard use matrix obtained invasively with histological tracers by Markov et al. (2014). performance metric, cross entropy measure that enables comparison relative tracer labeled neuron counts to streamline tractography. We find high signal-to-noise ratio are important estimate SC, low (1.03 mm3) in better agreement network, than those (0.53 or at an even lower (2.03 mm3). In contradiction, sensitivity specificity analyses suggest if is sufficient, balanced compromise which identical remains 60–64% regardless other scanning parameters. Interestingly, graph assumed be but thresholding, increases 70–75%. Hence, using metrics binarized graphs, one risks losing information, changing graphs dependence different