I already found the solution. You can use the c2d code itself in your embedded matlab file:
[m,n] = size(F); [m,nb] = size(G); s = expm([[F G]*Ts; zeros(nb,n+nb)]); Phi = s(1:n,1:n); Gamma = s(1:n,n+1:n+nb);
Alternatively an approximation can be used as presented in the paper below (page 23, equation 2.85).
ode45 can be replaced by coding your own Runge-Kutta solver in embedded matlab.