Detection that respects the patient.

Mission — pediatric first

Roughly fifty million people worldwide live with epilepsy. Twelve million have been misdiagnosed even by specialists. People with untreated epilepsy are three times more likely to die than the general population. About a third of patients do not respond adequately to medication.

We chose to start where the diagnostic gap is widest: pediatric epilepsy. Hospital EEG sessions are typically 30–60 minutes, in an unfamiliar environment, with a child who is not sleeping naturally. The data this produces is often inconclusive — and the next slot can be weeks away. Tracings sit in a queue; clinicians review them hours, sometimes days, after the event. Decisions that should be minutes are measured in shifts.

We are building the layer that closes this gap: a wireless, wearable EEG headset designed for comfortable long-term wear at home, and a clinical-grade detection engine that runs on the same scalp EEG montage hospitals already use. The mobile app and reference hardware are complete; the algorithm is validated on the CHB-MIT pediatric dataset; live clinical deployment is the next gate.

Founder

Dr. Serkan Yıldırım, PhD — Founder & CEO. Computer Sciences, with over twenty-five years of experience in software development, AI, and team leadership. His doctoral work focused on machine learning for biomedical time-series; the NSAS detection engine is a direct continuation of that research, built to clinical-grade software standards and currently entering its first hospital validation cycle.

The thesis underlying our classifier was validated on the CHB-MIT scalp EEG database, then re-validated cross-subject on held-out cohorts to ensure the model generalizes beyond the training population. The methodology and results are documented in the publications listed on our research page.

Co-founders

Erdal Çakıroğlu — Data Specialist. Software Engineering background (Melbourne Polytechnic University) with professional experience in data architecture, database administration, and systems engineering at industry-leading companies. At NSAS he owns the data architecture, embedded-systems integration, and the security envelope around clinical data — the layer that has to be right before a hospital lets us anywhere near a patient.

Göktuğ Yıldırım — Marketing, Sales & Communications. Drives go-to-market strategy, partner outreach, and the public face of NSAS. International education and a clinical-marketing instinct make him the bridge between the engineering team and the partners — hospitals, regulators, press — that have to understand what we are building.

Medical advisor

Prof. Dr. Meryem Aslı Tuncer, MD — Neurology, Hacettepe University Faculty of Medicine, Ankara. Over twenty years of clinical practice as a neurologist, with a Master's in Clinical Neuroimmunology and a professorship in Neurology since 2012. She advises NSAS on clinical methodology, neurology-specific product requirements, regulatory positioning, and the kind of evidence a neurologist actually wants to see before trusting an AI alert.

Her involvement is the reason we are confident describing what NSAS is and is not in clinical terms — and the reason we will not ship anything to a hospital that does not pass her review first.

Engineering partner

Novamio Technologies LLT (United States) is our contracted engineering partner. They handle a portion of the embedded-firmware and mobile-application engineering work under a formal cooperation protocol, alongside the in-house team in Finland. This arrangement gives us the senior engineering bench a pre-pilot company would otherwise have to fundraise to assemble.

The science, briefly

The platform ingests 22 channels of bipolar scalp EEG at standard clinical sampling rates. A short rolling window (~5 seconds) feeds a feature extractor — 92 spectral, statistical, and morphological features per channel — into a Random Forest classifier trained for binary seizure event detection.

We chose Random Forest deliberately. It is interpretable, fast on commodity hardware, and produces well-calibrated probabilities clinicians can reason about. Deep approaches were tested; they did not improve cross-subject sensitivity enough to justify the loss of explainability that matters in a regulated setting.

What we do not do — and where we are

• We do not diagnose epilepsy. We surface events. Diagnosis remains the clinician's responsibility.
• We do not replace EEG technicians or neurologists. We are building software that compresses the time between event and review.
• We have not yet processed real patient EEG in a live hospital setting. All performance numbers on this site are from the public CHB-MIT dataset under cross-subject validation. Ethics-committee applications are open at four hospitals (two in Turkey, one in Helsinki, one in the United States).
• The deployed architecture is designed to keep raw EEG on-premises; only event summaries and clinician-approved exports would leave the hospital network. This will be configured at each pilot site rather than assumed.

Where we are

The company is registered in Espoo, Finland as Neuro Scan AI Solutions Oy (Y-tunnus 3437722-8). Our clinical and research partners are spread across Europe and North America; the project-facing offices and contact addresses you will see on this site are in New York, Amsterdam, and Helsinki.

For partnership inquiries, see partnerships. For media, see press. For everything else: contact.