Abstract

Diffusion tensor imaging (DTI) aims to non-invasively characterize the anatomy and integrity of the brains white matter fibers. To establish individual-specific precision approaches for DTI, we defined its reliability and accuracy as a function of data quantity and analysis method, using both simulations and highly sampled individual-specific data (927-1442 diffusion weighted images [DWIs] per individual). DTI methods that allow for crossing fibers (BedpostX [BPX], Q-Ball Imaging [QBI]) estimated excess fibers when insufficient data was present and when the data did not match the model priors. To reduce such overfitting, we developed a novel crossing-fiber diffusion imaging method, Bayesian Multi-tensor Model-selection (BaMM), that is designed for high-quality repeated sampling data sets. BaMM was robust to overfitting, showing high reliability and the relatively best crossing-fiber accuracy with increasing amounts of diffusion data. Thus, the choice of diffusion imaging analysis method is important for the success of individual-specific diffusion imaging. Importantly, for potential clinical applications of individual-specific precision DTI, such as deep brain stimulation (DBS), other forms of neuromodulation or neurosurgical planning, the data quantities required to achieve DTI reliability are lower than for functional MRI measures.