MR-thermometry allows monitoring of the local temperature evolution during minimally invasive interventional therapies. However, for the particular case of MR-thermometry in the human breast, magnetic field variations induced by the respiratory cycle lead to phase fluctuations requiring a suitable correction strategy to prevent thermometry errors. For this purpose a look-up-table-based multibaseline approach as well as a model-based correction algorithm were applied to MR-thermometry to correct for the periodic magnetic field changes. The proposed correction method is compatible with a variety of sensors monitoring the current respiratory state. The ability to remove phase artefacts during MR-thermometry of the human breast was demonstrated experimentally in five healthy volunteers during 3 min of free-breathing using pencil-beam navigators for respiratory control. An increase of 170-530% in temperature precision was observed for the look-up-table-based approach, whereas a further improvement by 16-36% could be achieved by applying the extended model-based correction.