Nicole Seider, NOUS Imaging Neuroscientist, publishes in NeuroImage on the “Accuracy and reliability of diffusion imaging models” advocating we need more data per subject and to avoid overconfident prior-driven models. 

This pivotal research led by NOUS Imaging co-founder Nico Dosenbach and NOUS Neuroscientist, Nicole Seider, in collaboration with neuroscientists at Washington University in St. Louis (WashU), University California San Diego (UCSD), New York University (NYU), and Johns Hopkins advances our understanding of the limitations and interpretability of popular diffusion imaging methodologies.  

While diffusion imaging models have advanced at an accelerated rate with increasing computer power, the data acquisition improvements are not commiserate with the model data needs. Seider and team tested a subset of the most popular models in the literature to validate their accuracy and reliability with an excess of high-quality data within individuals.  

For each model evaluated, their research demonstrated areas of higher and lower reliability and accuracy. Specifically, brain regions with complex microstructures such as crossing fibers and multiple compartments need more data for model convergence than what is commonly acquired in a DWI dataset.  

We congratulate Nicole and team as this research offers important guidance toward acquisition standards, particularly in the clinical setting. The need for better diffusion imaging data is essential to accurately inform a definitive patient diagnosis and corresponding treatment pathway. 

https://www.sciencedirect.com/science/article/pii/S105381192200266X?via%3Dihub