To determine the best compromise between low radiation dose and suitable image quality for the detection of lytic and sclerotic bone lesions of the lumbar spine and pelvis. Methods: A phantom was scanned using the routine protocol (STD, 13 mGy) and six decreasing dose levels. Raw data were reconstructed using level 3 of iterative reconstruction (IR3) with 1-mm slice thickness for the STD protocol and highest IR levels with 3-mm slice thickness for the others. CTDIvol was used for radiation dose assessment. Quantitative criteria (noise power spectrum [NPS], task-based transfer function [TTF], and the detectability index [d']), as well as qualitative analysis, were used to compare protocols. NPS and TTF were computed using specific software (imQuest). d' was computed for two imaging tasks: lytic and sclerotic bone lesions. A subjective analysis was performed to validate the image quality obtained on the anthropomorphic phantom with the different dose values. Results: Similar d' values were found for CTDIvol from 3 to 4 mGy with IR4 and from 1 to 2 mGy for IR5 compared with d' values using the STD protocol. Image quality was validated subjectively for IR4 but rejected for IR5 (image smoothing). Finally, for the same d', the dose was reduced by 74% compared with the STD protocol, with the CTDIvol being 3.4 mGy for the lumbar spine and for the pelvis. Conclusion: A dose level as low as 3.4 mGy, in association with high levels of IR, provides suitable image quality for the detection of lytic and sclerotic bone lesions of the lumbar spine and pelvis.