Endovascular Brain-Computer Interfaces: Review of Safety and Digital Communication Advances

Introduction Brain‐computer interfaces (BCIs) offer high throughput decoding of neural signals. Current clinical implications include communication and motor rehabilitation. Invasive implants offer high‐fidelity recordings at the cost of craniotomy and higher morbidity. Conversely, non‐invasive modalities suffer from poor spatial resolution. Endovascular BCIs entail minimally invasive procedures via vascular access for cortical activity recording. They have the potential to both decrease morbidity, while still offering high‐fidelity neural encoding. This State‐of‐the‐Art (SotA) review aims to integrate both preclinical and clinical evidence. It seeks to elucidate feasibility, safety, and translational research gaps. Materials and Methods A State‐of‐the‐Art (SotA) methodologic review of Scopus, Web of Science, Ovid MEDLINE, EMBASE, IEEE Xplore, Google Scholar, and ClinicalTrials.gov was conducted. All applicable studies up to August 2025 for endovascular BCIs were screened. These included any devices utilized for recording, stimulation, device design, and/or clinical applications. Aggregate data and quantitative outcomes were extracted, including but not limited to: signal quality, safety, functional results, and/or multivariate factors related to the procedures. Priority was given to preclinical and first‐in‐human studies (NCT03834857 and NCT05035823). Results Preclinical models indicate that stent‐like electrode arrays inserted into cortical veins, including the superior sagittal sinus and tributaries, are chronically stable and capture electrocorticographic signals without inducing parenchymal damage. Early human feasibility trials with the Stentrode™ system indicate accurate decoding of motor intention in patients with severe paralysis, allowing for functional digital communication and device control. Signal fidelity was greater than scalp EEG but smaller than subdural grids, with adaptive machine‐learning decoding improving decoding accuracy. Procedural safety was similar to venous stenting, with thromboembolic complications only and no long‐term parenchymal damage. Advances in miniaturized amplifier technology, wireless telemetry, and biocompatible flexible scaffolds enable long‐term clinical translation. Limitations are sparse electrode density caused by vascular anatomy, unknown long‐term endothelial response, and developing decoding algorithms. Conclusion Endovascular BCIs are a disruptive technology with potentially significant clinical impact. It is an innovative convergence between neural engineering and neurointervention. Although sparse, there is a growing body of early evidence that indicates safety, feasibility, and rehabilitation benefit. However, many challenges remain. Future trials will need to: (1) explore long‐term durability and neointimal integration, (2) enhance decoding accuracy, and (3) establish relative effectiveness when compared to standard BCI modalities. Furthermore, serious efforts must be made to address ethics and regulatory issues. Neurointerventionalists are best poised to standardize procedural protocols and guide translational research efforts.