Abstract

Previous studies have showed only regions with a sensitivity higher that 1% of the maximum value can affect the recovery result for diffuse optical tomography (DOT). Two methods of efficient sensitivity map generation based on Finite Element Models (FEM) are developed based on (1) reduced sensitivity matrix and (2) parallelisation process. Time and memory efficiency of these processes are evaluated and compared with conventional methods. It is shown that the computational time for a full head model containing 200k nodes is reduced from 3 hours to 48 minutes and the required memory is reduced from 5.5 GB to 0.5 GB. For a range of mesh densities upto 320k nodes, the required memory is improved by ~1000% and computational time by ~400% to allow near real-time image recovery.

© 2015 SPIE