We develop methods needed to perform loop calculations in light cone perturbation theory using a helicity basis, refining the method introduced in our earlier work. In particular this includes implementing a consistent way to contract the four-dimensional tensor structures from the helicity vectors with d-dimensional tensors arising from loop integrals, in a way that can be fully automatized. We demonstrate this explicitly by calculating the one-loop correction to the virtual photon to quarkantiquark dipole light cone wave function. This allows us to calculate the deep inelastic scattering cross section in the dipole formalism to next-to-leading order accuracy. Our results, obtained using the four dimensional helicity scheme, agree with the recent calculation by Beuf using conventional dimensional regularization, confirming the regularization scheme independence of this cross section. PACS numbers: 24.85.+p,25.75.-q,12.38.Mh p, h, α p ≡ p − k, h , β k, λ, a; k + = zp + q ≡ k − zp p, h, α p ≡ p − k, h , β k, λ, a; k + = zp + q ≡ k − zp FIG. 1: Left: Gluon emission vertex from a quark V α;β,a h;h ,λ (q, z) Eq. (6), where α, β are quark colors, h, h the quark helicities before and after the emission, a the gluon color and λ the gluon helicity. Right: Gluon absorption vertex into quark V β,a;α h ,λ;h (q, z) Eq. (8).Any factor of dimension arising from the numerator Lorentz and Dirac algebra should be labeled as d s , and should be distinct from the dimension d. Once the spin and tensor algebra is done one analytically continues the result to d < 4 and takes the limit d s → 4 for the spins of the internal particles.Typically the dimensionality is parametrized as d = 4 − 2ε. We will also use the notation d ⊥ ≡ d − 2 = 2 − 2ε for the number of transverse dimensions in light cone coordinates. Within the DR and FDH schemes one can still choose the momentum of observed particles to be either d-dimensional or 4-dimensional. At one-loop order, however, these choices lead to difference of O(ε) and thus one can set the observed particles momenta to be 4-dimensional.The question of which regularization scheme is most efficient for a given calculation is of course very subjective. We would like to argue in this paper that for one-loop LCPT calculations the helicity basis supplemented with the FDH regularization scheme is in fact the most efficient one. However, as we will show below, the helicity basis approach can also be combined with other dimensional regularization scheme choices, and in particular with the CDR scheme. Our overall motivation for using the FDH scheme is the following. The one-loop results for physical observables arise from a product of a one-loop tensorial loop integral and another tensor from the spin/helicity structure of the vertices. The resulting contributions can be classified into three kinds of terms. The most divergent part is obtained by taking the divergent 1/ε-term from the integral, and evaluating the helicity structure in 4 spacetime dimensions. This part has no scheme dependence. The scheme de...