A new technique, based on in-vivo energy dispersive X-ray fluorescence (EDXRF), has been developed to gadolinium (Gd) concentrations identification in planar X-ray fluorescence (XRF) images. Higher signal-to-noise (SNR) ratios while keeping a low radiation dose were achieved. Experimental validation was performed using tissue equivalent phantoms under two different data acquisition criteria. The first criteria consisted on acquiring the energy spectra from different experimental narrow spectrum beam (FWHM = 2.5 keV) with peak central energy above the L edge energy and determining the spectrum which producing Lowest-Limit-of-Detection (Lowest-LoD) for a specific acquisition time. This also provided the minimum dose expected under the condition of minimum irradiation time. The second criteria consisted on measuring the surface dose required to obtain a specific LoD by different narrow spectrum beam, providing the Lowest-Dose setting. Surface (2D) Gd-doped tissue-equivalent phantoms scanning were performed according to optimized scenarios: Lowest-LoD setting (obtaining to central energy of 10.9 keV) and Lowest-Dose setting (obtaining to central energy 12.9 keV). 625 pixel images were acquired in two different conditions: a pre-defined time (5 s) per pixel was set in the first approach, whereas a pre-defined total surface dose (4 mGy) was set to the second approach. According to the results obtained for the first approach, a 58 times reduction was observed when comparing SNR between the Lowest-LoD and Lowest-Dose settings. On the other hand, for the second approach pre-defining total dose during the whole examination, the best SNR was obtained for the Lowest-Dose configuration exhibiting a 42% of increment respecting to the Lowest-LoD configuration and 47 times higher when compared with the limit case of no optimization.