I don't know exactly where it appears in the literature, but a coordinate-free interpretation is certainly known, based on the idea of Penrose's twistor construction.  The main points can be summarized as follows:

First, everything is equivariant under the group $\mathrm{SO}(2m{+}1)$:  One has $S^{2m} = \mathrm{SO}(2m{+}1)/\mathrm{SO}(2m)$ as the space of unit vectors $\mathbb{R}^{2m+1}$, and $\mathrm{Gr}^+_2(\mathbb{R}^{2m+1})$, the Grassmannian of oriented $2$-planes in $\mathbb{R}^{2m+1}$ is $\mathrm{SO}(2m{+}1)/\bigl(\mathrm{SO}(2){\times}\mathrm{SO}(2m)\bigr)$.  This latter space can be realized as the complex quadric $Q_{2m-1}\subset\mathbb{CP}_{2m}$ by identifying a unit bi-vector $e_1\wedge e_2$ (where $(e_1,e_2)$ is an oriented orthonormal basis of the $2$-plane $P\in\mathrm{Gr}^+_2(\mathbb{R}^{2m+1})$) with the 'isotropic' complex line $[e_1{-}ie_2]\in\mathbb{CP}_{2m}$. (It is called 'isotropic' because it is spanned by the complex vector $v = e_1{-}ie_2$ that satisfies $v\cdot v = 0$, i.e., $v$ is 'null', aka 'isotropic'.)  It is easy to see that the correspondence $P\leftrightarrow [e_1{-}ie_2]$ is one-to-one and onto.

It's also worthwhile to note that a maximal torus $T^m$ in $\mathrm{SO}(2m)$ is also a maximal torus in $\mathrm{SO}(2m{+}1)$ and can be understood as the stabilizer of a splitting of $\mathbb{R}^{2m}$ as an orthogonal direct sum of $m$ oriented $2$-planes.  In particular, the space $\mathrm{SO}(2m{+}1)/T^m$ can be thought of as the space of orthogonal splittings $\mathbb{R}^{2m+1} = L \oplus P_1\oplus\cdots\oplus P_m$, where $L$ is an oriented line (i.e., an element of $S^{2m}$) and each $P_i$ is an oriented $2$-plane (i.e., an element of $\mathrm{Gr}^+_2(\mathbb{R}^{2m+1})$).  Note that, since $\mathrm{SO}(2m{+}1)/T^m$ is a compact simple group modulo a maximal torus, it has a natural complex structure.  In fact, it has more than one, but there is a particularly nice one to choose for applications to this problem, one such that the mapping $\pi_m: \mathrm{SO}(2m{+}1)/T^m\to Q_{2m-1}\subset\mathbb{CP}_{2m}$ given by 
$$
\pi_m\bigl(L \oplus P_1\oplus\cdots\oplus P_m) = \overline{P_m}
$$
is holomorphic. 

Next, back in the 60s, Calabi had shown that, if $x:S^2\to S^{2m}$ is a (branched) minimal immersion of the $2$-sphere that is *non-degenerate* (i.e., its image does not lie in a great $(2m{-}1)$-sphere), then $x$ has a canonical lifting $\hat x:S^2\to\mathrm{SO}(2m{+}1)/T^m$ as a holomorphic rational curve of the form
$$
\hat x =  \mathbb{R}\,x \oplus \pi_1(x)\oplus \cdots \oplus \pi_m(x)
$$
Basically, the idea is that you choose a splitting of the $x$-pullback of the tangent bundle of $S^{2m}$ into oriented $2$-planes corresponding to the osculating spaces of the immersion.  For example, $\pi_1(x)$ represents the tangent space, $\pi_2(x)$ represents the first normal, etc.

What Barbosa observed is that the map 
$$
\overline{\pi_m(\hat x)}:S^2\to \mathrm{Gr}^+_2(\mathbb{R}^{2m+1}) = Q_{2m-1}\subset \mathbb{CP}_{2m}
$$
is holomorphic, and he showed that there is a way to recover $x$ from $\overline{\pi_m(\hat x)}$.  

Moreover, he showed that, if one starts with a holomorphic rational curve $\xi:S^2\to Q_{2m-1}\subset \mathbb{CP}_{2m}$ that is nondegenerate in the sense that its image does not lie in a hyperplane in $\mathbb{CP}_{2m}$, then it is of the form $\xi = \overline{\pi_m(\hat x)}$ for some unique branched minimal immersion $x:S^2\to S^{2m}$.