<b><i>Introduction:</i></b> Family caregivers of patients with dementia suffer a high burden of depression and reduced positive emotions. Mentalizing imagery therapy (MIT) provides mindfulness and guided imagery skills training to improve balanced mentalizing and emotion regulation. <b><i>Objective:</i></b> Our aims were to test the hypotheses that MIT for family caregivers would reduce depression symptoms and improve positive psychological traits more than a support group (SG), and would increase dorsolateral prefrontal cortex (DLPFC) connectivity and reduce subgenual anterior cingulate cortex (sgACC) connectivity. <b><i>Methods:</i></b> Forty-six caregivers participated in a randomized controlled trial comparing a 4-week MIT group (<i>n</i> = 24) versus an SG (<i>n</i> = 22). Resting state neuroimaging was obtained at baseline and post-group in 28 caregivers, and questionnaires completed by all participants. The primary outcome was change in depression; secondary measures included anxiety, mindfulness, self-compassion, and well-being. Brain networks with participation of DLPFC and sgACC were identified. Connectivity strengths of DLPFC and sgACC with respective networks were determined with dual regression. DLPFC connectivity was correlated with mindfulness and depression outcomes. <b><i>Results:</i></b> MIT significantly outperformed SG in improving depression, anxiety, mindfulness, self-compassion, and well-being, with moderate to large effect sizes. Relative to SG, participants in MIT showed significant increases in DLPFC connectivity – exactly replicating pilot study results – but no change in sgACC. DLPFC connectivity change correlated positively with mindfulness and negatively with depression change. <b><i>Conclusions:</i></b> In this trial, MIT was superior to SG for reducing depression and anxiety symptoms and improving positive psychological traits. Neuroimaging results suggested that strengthening DLPFC connectivity with an emotion regulation network might be mechanistically related to MIT effects.