There is still no solid theoretical understanding of the mechanisms behind decohernce in neutrino oscillations: in literature it is possible to find many different approaches, but they all must rely on several strong approximations, often offering contradicting predictions. This could be relevant for the next generations of neutrino experiments: due to their high precision, even a small modification of the oscillation probability could affect the result. We will present some of the results obtained using a model where all the particles are described by fields, which are evolved dynamically and consistently with Quantum Field Theory. We have found that environmental interactions are crucial for decoherence, since they localize the moment of production of the neutrino; otherwise the time evolution of the state would be a coherent integral over all the possibilities and there would be no maximal coherent length: this is the case, for example, when neutrinos are produced and detected in vacuum. However this is not the only way to get the decoherence: if the final states are localized as well, the position in time and space of the neutrino production would be constrained by kinematic. We will show that if this is the case, there is a maximal coherence length again, in agreement with the other models available in literature.