Convergence is determined by testing, after each iteration, for image positivity, by checking if f ^ { k-1 } = 0 for the case when f ^ { k-1 } ( Note that f refers to the different x-ray linear attenuation coefficients at different voxels of the patient image ).