Magnetic resonance thermometry imaging (MRTI) has demonstrated attractive abilities in the thermal therapies scenario. A safe and accurate thermal treatment requires real-time temperature control throughout the procedure. The possibility to obtain multi-dimensional temperature maps with good spatial and temporal resolution makes MRTI a good candidate for thermometry implementation. The gradient-echo pulse sequences currently remain the most used for proton resonant frequency shift MRTI clinical study. However, the limited temporal resolution is motivating the development and the investigation of more performant pulse sequences. In this study, two MRTI protocols were compared, i.e., the 2D and the 3D echo-planar imaging (EPI), and their performances were assessed in two scenarios. As a first step, MRTI temperature values were compared with temperature data acquired with fiber optic sensors based on fiber Bragg grating (FBG) technology during water-based phantom cooling. The MRTI performance was estimated by applying a linear regression giving a root mean square error of 0.30 °C and 0.38 °C for the 2D and 3D protocols, respectively. Then, the Bland-Altman analysis shown 0.14 °C and −0.16 °C for the mean of difference and 0.72/-0.45 °C and 0.92/-1.24 °C for the level of agreement for the 2D and 3D protocols, respectively. As a second step, the thermal maps were extracted during an ex vivo liver thermal therapy. Results showed a better temperature reconstruction for the 3D protocol guaranteed by its spatial resolution, thickness, and ability to monitor a volume with 24 mm thickness, with a time scan of about 3 s. In conclusion, this work encourages further investigations concerning the 3D protocol abilities in the more complex pre-clinical and clinical applications.