UNITE: a generic, rapid, and robust platform for personalized functional circuit-guided neuromodulation targeting

Release notesunite.25.1.1: NVIDIA Driver Version: >= 520.61.05 / CUDA Version: >= 11.8 (GPU optional)unite.25.1.1.cuda129: NVIDIA Driver Version: >= 570.124.06 / CUDA Version: >= 12.9 (GPU optional)If you are using an NVIDIA 50 series GPU, please download the image ending with cuda129.Neuromodulation techniques—such as transcranial magnetic stimulation (TMS), focused ultrasound stimulation (FUS), deep brain stimulation (DBS), and intracranial electrical stimulation (iES)—target specific brain functional circuits to treat various brain disorders. Resting-state fMRI (rsfMRI) has been increasingly used to map functional circuits in individual patients for target localization. However, this strategy faces three major barriers to clinical translation: time-intensive rsfMRI processing, insufficient robustness in clinical cases, and a lack of automated targeting solutions. Here, we introduce a generic, rapid, robust, and automated platform for personalized functional circuit-guided target planning across different neuromodulation techniques, termed UNITE(Unified platform for Neuromodulation with Individualized Target Estimation). The platform’s high computational efficiency and robustness are empowered by DeepPrep, a recently developed end-to-end preprocessing pipeline that achieves a 10-fold speedup and improved robustness in clinical cases compared to the state-of-the-art pipeline. Using DeepPrep’s preprocessed data, the platform maps personalized functional circuits and localizes personalized targets through automated targeting algorithms. Currently, the platform offers three built-in TMS targeting algorithms for post-stroke functional rehabilitation and one FUS targeting algorithm for Parkinson’s disease (PD).Notably, it is designed to be extendable to other brain disorders and neuromodulation applications. To ensure quality and usability, the platform generates a data quality control report and a visual verification report of the identified targets, with target coordinates exported in multiple formats for compatibility with different neuronavigation systems. Overall, this platform enables personalized neuromodulation target localization within 30 minutes on a local GPU-equipped workstation. A basic level of computer science proficiency, along with neuroimaging experience, is sufficient for its use.