Skip to main content
. Author manuscript; available in PMC: 2015 Jul 15.
Published in final edited form as: IEEE Trans Med Imaging. 2012 May 16;31(10):1837–1848. doi: 10.1109/TMI.2012.2199763

Table I.

Outline of the alternating maximization used by the KCR Algorithm

μ[0,,0] = Initial reconstruction (FBP)
Λ[0,0,·] = Initial guess for registration parameters
H[0,0,·] = σI, initial guess for inverse Hessian
for k = 0 to max_iterations-1,
% Registration Update Block
for r = 1 to P (number of registration updates)
   Compute ΛL(μ[k,,0],Λ[k,p1,])
  H[k, p,·] = BFGS update using new gradient
   γ^ = line search in Λ[k,p1,]+γH[k,p,]ΛL(μ[k,,0],Λ[k,p1,])
   Λ[k,p,]=Λ[k,p1,]+γ^H[k,p,]ΛL(μ[k,,0],Λ[k,p1,])
end
Compute di[k,p,](Λ[k,p,])
% Image Update Block
for m = 1 to M (number of subsets)
   Compute curvatures, ci[k,,m](μ[k,,m1]Λ[k,p,])
   Compute modified line integrals, ti[k,,m](μ[k,,m1],Λ[k,P,])
  
μj[k,,m]=[μj[k,,m1]+MiΩmbijg˙i(ti[k,,m1];di[k,P,])βR˙j(μ[k,,m1])MiΩmbij(jbij)ci[k,,m1]+2βR¨j(μ[k,,m1])]+
end
μ[k+1,,0]=μ[k,,M],Λ[k+1,0,]=Λ[k,P,],H[k+1,0,]=H[k,P,]
end