FIND Surgical Sciences, Inc., doing business as FIND Neuro, has received US Food and Drug Administration 510(k) clearance for CN-Suite, a neurological decision-support platform designed to analyze intracranial electroencephalography and help clinicians identify and prioritize brain regions that may be driving seizures during presurgical evaluation for drug-resistant epilepsy. FDA records show that CN-Suite, filed under K260563, was found substantially equivalent on August 16, 2026 and classified as software for source localization using electroencephalography or magnetoencephalography. The company says the cleared claims cover focal and multifocal epilepsy in patients aged three years and older, giving the Cambridge, Massachusetts-based startup a regulatory pathway into specialist epilepsy surgery centers rather than conventional outpatient seizure monitoring.
The problem CN-Suite is trying to solve is not simply detecting when a seizure occurs. Patients undergoing stereoelectroencephalography already have electrodes implanted deep within the brain, giving specialists detailed recordings from multiple anatomical locations during actual seizures. The harder interpretation is determining which areas are true drivers of the epileptic network and which are responding after abnormal electrical activity has already propagated from somewhere else. Removing too little tissue can leave the critical network intact and allow seizures to persist, while removing unnecessarily broad regions can increase neurological risk.
Why is finding the epileptogenic zone harder than identifying the first abnormal EEG signal?
A seizure is a network event. Electrical activity can spread rapidly through connected brain structures, which means the site showing obvious abnormal activity is not always the region responsible for destabilizing the network. Experienced epileptologists combine seizure semiology, scalp EEG, imaging, intracranial recordings and other clinical information when developing a surgical hypothesis, but interpretation can remain challenging in multifocal disease or when propagation occurs quickly.
CN-Suite attempts to quantify directional information flow between intracranial EEG contacts. The research version uses delay-adjusted wavelet-based transfer entropy to estimate whether activity at one contact helps predict activity at another and in which direction, then feeds those relationships into a locked machine-learning classifier that assigns each contact a criticality score. A high score is intended to identify a region acting more like a seizure driver, whereas a low score is intended to represent tissue that behaves more like a responder or noncritical network node.
This changes the nature of the software from a seizure detector into a network-analysis tool. CN-Suite is not independently deciding which part of the brain should be removed; it is producing quantitative information intended to sit alongside physician interpretation during one of the highest-stakes decisions in epilepsy care.
What did the multicenter validation show before FDA clearance?
A company-funded multicenter study analyzed 60 patients aged two years and older with focal or multifocal drug-resistant epilepsy who underwent stereoelectroencephalography followed by surgery at four US Level 4 epilepsy centers. The algorithm had been trained on an independent group of 37 patients and locked before investigators analyzed the validation cohort, reducing the risk that model parameters were retrospectively optimized around the patients being used to judge performance.
In the original validation analysis, criticality values were significantly more concentrated in surgically treated tissue among patients with favorable Engel I or II outcomes, producing a Cohen’s d of 0.74 with a 95% confidence interval of 0.39 to 1.06 and a P value of 0.003. High-criticality contacts formed compact spatial clusters, with a nearest-neighbor distance of roughly 9 millimeters compared with approximately 17 millimeters expected by chance. Sensitivity reached 80% when procedures involved 10 or fewer treated contacts, while contact-level specificity was reported at 84%.
A later version of the analysis expanded the modeling approach and reported an effect size of 1.12 for predicted surgical success between favorable and less favorable outcome groups, again finding that residual high-criticality tissue outside the treatment zone was associated with poorer outcomes. These findings remain available as a medRxiv preprint rather than a peer-reviewed clinical publication, and the study was funded by FIND Neuro, with several authors employed by or financially connected to the company. Those limitations do not negate the result but are important when judging the strength of independent validation.
Why could the software be particularly relevant to minimally invasive epilepsy procedures?
The spatial clustering result could have practical implications for laser interstitial thermal therapy, where clinicians attempt to ablate relatively focused targets through a minimally invasive probe rather than performing a broad open resection. If high-criticality contacts genuinely identify compact seizure-driving networks, they could help physicians prioritize which regions should be included in the treatment volume.
The algorithm performed best in smaller focal procedures, while sensitivity declined as the size of resection increased. That pattern makes biological sense because a software tool identifying a relatively concentrated driver network may be more informative when the treatment objective is focused ablation than when a surgeon removes a broad anatomical region containing many different EEG contacts.
The study also found an interesting pattern among surgical failures: high-criticality tissue often remained outside the treated boundary. Retrospectively, that creates a plausible explanation for persistent seizures and could identify candidates for reevaluation or repeat intervention. Prospectively, the more important possibility is using the information before surgery to determine whether a proposed treatment plan adequately covers the predicted network.
What does FDA clearance establish and what does it not establish?
FDA classified CN-Suite as a Class II neurological software product and cleared it through the traditional 510(k) pathway after finding it substantially equivalent to an existing legally marketed device. The decision permits FIND Neuro to market the software according to its cleared indications, but it should not be described as FDA proving that use of CN-Suite increases seizure-free surgery rates.
The validation data are retrospective: physicians made the original surgical decisions without CN-Suite, and the software was subsequently tested against treatment locations and outcomes. That design is appropriate for demonstrating whether the algorithm’s predictions align with successful treatment patterns, but the strongest clinical evidence would come from a prospective study examining whether physicians using CN-Suite actually make different decisions and whether those decisions improve seizure freedom without increasing neurological complications.
That distinction is particularly important because decision-support software can perform impressively against historical datasets without necessarily changing clinical outcomes. Surgeons may already recognize many of the regions the software identifies, or they may be unable to treat an algorithmically important area because it overlaps language, motor or other essential cortex.
Why does epilepsy surgery need more quantitative decision support?
Drug-resistant epilepsy is typically defined after failure of two appropriately chosen and tolerated antiseizure medication regimens, yet many eligible patients spend years cycling through additional drugs despite relatively low chances that another medication will produce durable seizure freedom. Surgery can be highly effective when a sufficiently localized epileptogenic network can be removed or ablated safely, but the diagnostic pathway is intensive and requires multidisciplinary expertise.
Intracranial EEG generates large quantities of dynamic information, creating exactly the type of environment where quantitative analysis could complement human pattern recognition. The potential value is not replacing epileptologists but helping them prioritize signals within a network too complex for any one visual EEG feature to capture.
FIND Neuro has therefore reached an important commercial threshold: CN-Suite is no longer merely a research algorithm awaiting FDA review. The next threshold is more difficult. Epilepsy centers need evidence that adding its criticality map to existing presurgical workflows changes treatment planning in a way that produces more seizure-free patients, especially when every additional millimeter of brain targeted for ablation carries consequences that an algorithm cannot judge by network mathematics alone.
