Deep brain stimulation (DBS) is a surgical treatment used to improve motor skills in patients with Parkinson’s disease (PD), but there is a prevalent need to determine whether certain patients would respond to DBS before undergoing the surgery. To fill this need, the study explored quantitative susceptibility mapping (QSM) as an alternative to the levodopa challenge test (LCT). LCT only measures dopamine responsiveness, while QSM utilizes magnetic resonance imaging to map iron distribution in the brain, which accounts for broader changes as iron is involved in dopaminergic, serotonergic, GABAergic, and glutamatergic pathways. These QSM deep gray nuclei spatial features were used to develop a QSM radiomics model that predicts improvements following DBS. The model was evaluated across 67 PD patients undergoing bilateral subthalamic nucleus (STN) DBS by the Universal Parkinson’s Disease Rating Scale III. It was found that the model is a more accurate predictor of symptom improvement following DBS than the LCT; there were Pearson correlations of r = 0.75 (p = 1.1 x 10^-7) and r = 0.71 (p = 1.8 x 10^-5) across the two patient centers, compared to that of the LCT, where r = 0.12 (p = 0.48) and r = -0.14 (p = 0.45). When the feature maps of responder and non-responders to DBS were compared, there was a substantial difference in the STN and substantia nigra (SN). Overall, the study highlighted QSM radiomics as a more accurate method of predicting DBS outcome in PD patients, through iron as a precursor of certain neurotransmitters in the STN and SN. A stand out limitation is that the study focused on preoperative predictions, so the QSM model did not consider operative or post-operative factors such as lead placement and stimulation programming.