What you're going to need to do is build an pair of operators that correspond to creating and annihilating your proposed ground state, and then look at the two point function that operator generates.
For example, you can build a bound state hydrogen atom creation operator as
$$\Psi_{H,n\ell m}(\mathbf{R}) = \int \mathrm{d}^3r\ \psi_{n\ell m}(\mathbf{r})\, \Psi_e\left(\mathbf{R} + \frac{m_p}{m_e+m_p}\mathbf{r}\right)\, \Psi_p\left(\mathbf{R} - \frac{m_e}{m_e+m_p}\mathbf{r}\right)$$
using the proton and electron creation operators, $\Psi_p$ and $\Psi_e$, respectively, and $\psi_{n\ell m}$ the standard "excited state" of the hydrogen atom.
Now, critically, when you look at the long term time behavior of the two point function of your compound operator, if it is a single bound state, you should find the same sort of $\mathrm{e}^{-mT}$ behavior you have with the simple two point function.
If you do not have the ability to build such a perfect compound operator you can always just look for the spectrum in the higher order operators. Notice how the compound operator above is built from a particular superposition of two more primitive operators (and $\Psi_p$ is, itself, a compound operator…). That tells you that the bound state behavior is to be found in the higher order correlators of the lower order theory. So, if you examine the correlator
$$\left\langle\phi(0+\mathbf{r}_1,0)\,\phi(0-\mathbf{r}_1,0)\,\phi(\mathbf{R}-\mathbf{r}_0,T)\, \phi(\mathbf{R} + \mathbf{r}_0,T)\right\rangle$$
you should find a whole spectrum of comopnents that go like $\mathrm{e}^{-mT}$ for multiple values of $m$ that correspond to the net potential energy of the bound states (i.e. $m = 2m_0 c^2 - E_B$, where $E_B$ is the binding enrgy of the state).
In other words, you want to do an inverse Laplace transform on the time coordinate of the 4-point function in Euclidean space-time and examine the peaks. I don't know how important it is to rearrange the spatial coordinates into center of mass and separation the way I have done above, but since $\mathbf{R}$ is the position coordinate of any putative bound state particle, it seems like a good idea.
In Minkowski space-time, the exponential decay from the cluster decomposition theorem gets replaced with oscillations. Specifically, if you look at the real space Feynman propagator the decay is actually $e^{mr}/r$ in the spatial direction, and $e^{imT}/T$ in the time direction. Which suggests looking for peaks in the Fourier transform of $T$ times the 4-point correlator for $mT\gg1$ should yield the desired results.
Thinking further, though, the frequency spectrum won't have discrete peaks. It will instead be written as a sum of functions that all have a lower limit in their support (e.g. $\sqrt{1+x^2}$). Each of those lower limits will correspond to the net rest mass energy of each bound eigenstate.
Sorry I can't provide more details - I don't know them all, myself.
