High-performance neural interfaces and functional neurosurgeries have historically faced a fundamental trade-off: to gain high-fidelity access to the brain's electrical architecture, clinicians had to accept substantial surgical risks. Traditional methods—such as open craniotomies to place dense subdural electrode grids or direct parenchymal penetration for intracortical microelectrode arrays—carry significant risks of infection, intracranial hemorrhage, and permanent tissue damage. As noted by Chengyu Li, these complexities and medical hazards create a steep barrier to widespread clinical adoption, limiting life-changing assistive technologies and diagnostic tools to a fraction of the patients who need them.
To overcome this barrier, a major paradigm shift is underway. Researchers are developing techniques that minimize or entirely eliminate surgical trauma to the brain parenchyma. By leveraging alternative anatomical pathways, sparse sampling geometries, and advanced computational decoding, these modern approaches match the efficacy of traditional open-brain procedures while drastically improving patient safety.
From 2D Surface Grids to 3D Percutaneous Networks
For decades, identifying the source of drug-resistant seizures required a large craniotomy to lay a flat, two-dimensional grid of electrodes directly onto the brain's surface. While effective at capturing cortical activity, this approach is highly invasive and carries a surgical morbidity rate that can reach up to 17%.
As explained by Hui Ming Khoo et al., Stereo-electroencephalography (SEEG) bypasses the need for a craniotomy by utilizing thin, one-dimensional depth electrodes inserted through tiny percutaneous punctures. This shift from a 2D surface grid to a sparse 3D network of depth electrodes drops the surgical morbidity rate to under 1% and reduces hemorrhage risks fourfold, as documented by Hui Ming Khoo et al.
Traditional Subdural Grid Stereoelectroencephalography (SEEG)
(Large Craniotomy) (Percutaneous Holes)
___________ _ _ _ _
/ _______ \ / | | | | | | \
| |[o o o]| | | | | | | | | |
| |[o o o]| | | | | | | | | |
| |[o o o]| | | | | | | | | |
\___________/ \___________/
High Morbidity (~17%) Low Morbidity (<1%)
2D Surface Mapping 3D Deep Network Mapping
The intuition behind why SEEG works so effectively relies on the brain's volumetric connectivity. Seizures are rarely strictly surface phenomena; they propagate through deep networks. By probing deep structures directly, SEEG provides a more accurate three-dimensional delineation of the epileptogenic zone.
Furthermore, these diagnostic tracts can double as therapeutic conduits. Once the problematic node in the neural network is identified, clinicians can pass radiofrequency currents through the exact same electrodes to perform radiofrequency thermocoagulation (RF-TC), destroying the seizure-causing tissue without a scalpel. To maximize the ablation volume through these narrow tracts, Xiaotong Fan et al. introduced a three-dimensional cross-bonding technique that links multiple SEEG electrodes together, creating larger, more effective therapeutic lesions that cured or significantly reduced seizures in 95.2% of patients. This high-density, targeted approach has also been successfully adapted by Xinhua Wang et al. to safely treat complex pediatric epilepsy cases, such as those presenting with tuberous sclerosis complex.
Crucially, the utility of sparse depth electrodes extends beyond epilepsy mapping into brain-computer interfaces (BCIs). While conventional speech neuroprostheses have relied on dense, highly invasive cortical grids, Maarten C. Ottenhoff et al. demonstrated that audible speech can be synthesized directly from sparse SEEG depth recordings. By utilizing an encoder-decoder deep learning framework with attention mechanisms, the authors compensated for the lower spatial density of the electrodes, achieving high-reconstruction correlations and proving that deep brain structures contain sufficient signal architecture for complex neural decoding.
Endovascular BCIs: Navigating the Vasculature
If depth electrodes minimize skull trauma, endovascular interfaces eliminate brain tissue trauma entirely. Rather than drilling through bone and penetrating the brain parenchyma to record motor intent, this approach treats the brain's blood vessels as natural, non-destructive conduits.
Thomas J. Oxley et al. pioneered this concept with the "Stentrode," an electrode array integrated onto a self-expanding medical stent. Entering through the jugular vein via standard neurointerventional surgery, the device is guided through the vascular tree until it rests inside the superior sagittal sinus—a major blood vessel running directly adjacent to the motor cortex.
Jugular Vein Entry ----> Vascular Highway ----> Superior Sagittal Sinus
(Adjacent to Motor Cortex)
|
[ Stentrode Array ]
|
Records through vessel wall
How can an electrode inside a blood vessel capture high-fidelity neural control signals without touching a single neuron? The answer lies in the transmission of local field potentials through the thin vascular wall. As shown by Kriti Kacker et al., the Stentrode reliably detects high-frequency electrical activity, specifically within the gamma and high-gamma bands (30 to 150 Hz), which correlate strongly with motor intent. Because the device is safely endothelialized within the blood vessel, it avoids the chronic immune and inflammatory responses that typically degrade direct brain tissue implants over time, allowing patients with severe paralysis to safely retain long-term control over digital interfaces.
Bypassing Surgical Implantation Entirely
The ultimate reduction in surgical invasiveness is the complete elimination of incisions. For therapeutic interventions, this is achieved through MR-guided Focused Ultrasound (MRgFUS). As detailed by Haruhiko Kishima et al., this technique coordinates an array of 1,024 acoustic transducers to send ultrasound waves through the intact skull. While any single wave passes through the tissue harmlessly, their convergence at a single, mathematically precise focal point deep within the brain generates enough thermal energy to ablate hyperactive circuits. Coupled with real-time MRI thermometry, surgeons can treat movement disorders like essential tremor with zero physical penetration.
On the interface side, researchers are bypassing intracranial surgery by extracting high-throughput control signals from peripheral or behavioral channels, using smart software to fill the information gap.
- Context-Aware Eye Tracking: Traditional eye-tracking interfaces often struggle with the "Midas Touch" problem, where every glance is mistakenly interpreted as a command. Aldo A. Faisal and William W. Abbott demonstrated that eye movements possess an incredibly high natural information throughput that can rival invasive BMIs at a fraction of the cost. To turn this raw throughput into meaningful control, Aldo A. Faisal et al. paired wearable eye-tracking with "Action Grammars"—predictive algorithms that understand the contextual steps of human tasks. This allows users to control a robotic arm for complex reaching and grasping tasks with only 10 minutes of training, achieving high-tier BCI performance non-invasively.
- Surface EMG Decomposition: For motor restoration, instead of placing arrays inside the motor cortex, Jordyn E. Ting et al. utilized a non-invasive wearable sleeve array wrapped around the forearm of a person with tetraplegia. By applying advanced signal decomposition to weak surface electromyography (EMG) signals, the authors isolated the firing patterns of individual motor units driven by residual spinal pathways. This algorithmic enhancement allowed them to decode distinct finger movements with over 75% accuracy, proving that high-resolution control signals can be harvested from the periphery without surgical risk.
Balancing Safety and Capability
The clinical community remains highly analytical of these advancements. As a worldwide survey by Simon C. Williams et al. highlights, neurosurgical teams carefully weigh the ethical, safety, and practical implications of neural interfaces, showing a strong preference for rehabilitative applications over purely augmentative ones.
By shifting the focus away from massive open surgeries toward percutaneous depth networks, endovascular pathways, and intelligent non-invasive decoding, this diverse suite of technologies addresses the core safety concerns of clinicians. They demonstrate that we do not need to disrupt the brain's physical boundaries to understand or heal its internal circuits.
Key Foundational Work
Progress in the brain–computer interface: an interview with Bin He (2019) — contributed 2 concept elements to this direction.
Ultra-low cost eyetracking as an high-information throughput alternative to BMIs (2011) — contributed 1 concept elements to this direction.
Optimized SEEG-guided radiofrequency thermocoagulation for mesial temporal lobe epilepsy with hippocampal sclerosis (2019) — contributed 1 concept elements to this direction.