GitHubMechanical Neuroimaging Lab · Univ. of Delaware

Sentinel.jlGPU-Accelerated MRE Inverse Solver

Reconstruct tissue mechanical properties from MRI displacement data using finite elements and iterative optimization.

7 Material Models

Isotropic, orthotropic, poroelastic, transverse isotropic, and more — all with complex-valued viscoelastic formulations.

2.1x speedup with CUDA (cuDSS batched solver + KA gradient kernel). Metal support for Apple Silicon.

80,000+ tests. 100-iteration brain MRE results match Fortran MPI within 2% across 171K material nodes.

Overlapping subproblem domain decomposition for robust brain MRE inversion with parallel execution.