The optimization of tumor laser ablation requires the evaluation of the temperature distribution in the tumor volume, but minimally invasive sensors can only provide information in one dimension, and often with consistent errors. Therefore, a suitable prediction algorithm, combined with accurate measurements, are required to reconstruct the temperature map in the tumor mass. This work provides preliminary results on the temperature mapping in an agar-gel phantom, using a quasi-distributed temperature sensor made of a fiber Bragg grating array with improved accuracy, and an algorithm of estimation of the temperature spatial distribution based on the thermal Green's function. Details on the fabrication and packaging of the sensor are provided along with an experimental evaluation of the thermal diffusivity in the phantom. Furthermore, it is shown how the accuracy on the evaluation of diffusivity is influenced by the synchronization error, which is the delay between the firing of the laser and the temperature acquisition.