Shape-constrained reconstruction in diffuse optical tomography by simulated annealing

When the inverse problem of diffuse optical tomography (DOT) is solved with the Born or Rytov approximation, the size of the matrix of the linear inverse problem becomes large if the volume (or area) of the domain in biological tissue used for reconstruction is large. The number of unknown parameters in DOT is reduced when constraints about the shape of a target are imposed for the inverse problem. Due to such constraints, the inverse problem becomes nonlinear even when the (first) Born or Rytov approximation is employed. We solve this nonlinear inverse problem by the simulated annealing, which is not trapped by local minima of the cost function.