Patients with deep brain stimulation (DBS) devices have limited access to magnetic resonance imaging (MRI) due to safety concerns associated with RF heating generated around the implant. The problem of predicting RF heating of conductive leads is complex with a large parameter space and several interplaying factors. Recently however, off-label use of MRI in patients with DBS devices has been reported based on limited safety assessments, raising the concern that potentially dangerous scenarios may have been overlooked. In this work, we present results of a systematic assessment of RF heating of a commercial DBS device during MRI at 1.5T and 3T, taking into account the effect of device configuration, imaging landmark, and patient's body composition. Ninety-six (96) RF heating measurements were performed using anthropomorphic phantoms implanted with a full DBS system. We evaluated eight clinically relevant device configurations, implanted in phantoms with different material compositions, and imaged at three different landmarks (head, shoulder, and lower chest) in 1.5 T and 3T scanners. We observed a substantial fluctuation in the RF heating depending on phantom's composition and device configuration. RF heating in the brain-mimicking gel varied from 0.1°C to 12°C during 1.5 T MRI and from <0.1°C to 4.5°C during 3T MRI. We also observed that certain device configurations consistently reduced RF heating across different phantom compositions, imaging landmarks, and MRI transmit frequencies.