Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) are the two most commonly used non-invasive methods for studying brain function, having different but complementary strengths: high temporal resolution of the former and high spatial resolution of the latter. Crucially, fMRI is vital for studying subcortical areas, as those are practically out of reach for EEG. At the same time, EEG is cost-effective and, thus, preferable to fMRI if comparable information can be extracted. Here we present an EEG-to-fMRI neural network with an interpretable module for feature extraction. Using the EEG-fMRI dataset, we show that our model allows us to predict the detailed resting state Blood Oxygenation Level Dependent (BOLD) activity of seven bilaterally symmetric subcortical structures solely from multichannel EEG data. Preliminary results reported here show a performance level significantly above chance and exceeding the state-of-the-art accuracy typically reported for a single structure such as the amygdala or striatum. These findings pave the road toward the creation of low-cost mobile scanners of subcortical activity with improved usability, EEG-based fMRI digital twin technology, with a broad range of applications – from fundamental neuroscience through diagnostics to neurorehabilitation and affective neurointerfaces. The demo video is presented in unmapped: uri https://youtu.be/IOOwb7Wt2sY.