Bayesian Inference on MRI Diffusion Tensor Imaging Data

The dataset we used in this project comes from the Stanford HARDI dataset that includes MRI data of two healthy male participants. The goal of this project is to estimate the diffusion tensor (D) and baseline signal (S0). For simplicity, we focus on a single voxel from white matter by exploring different Bayesian inference techniques (view repo). This was held as part of the master’s level course “Advanced Probabilistic Machine Learning - 1RT705” at Uppsala University.

Personally, worked on implementing the Metropolis-Hastings algorithm from scratch, as a Stochastic Approximate Inference technique.

The implementation was done in SciPy (Python).