DECCOG
Early Detection of Cognitive Decline

DECCOG (Plataforma per a la detecció primerenca de DEClivi COGnitiu) is a research and innovation project led by Starlab Barcelona in partnership with the BarcelonaBeta Resaerch Center of the Fundació Pasqual Maragall, funded by the Generalitat de Catalunya under the Nuclis d'R+D 2024 programme (Expedient ACE120/24/000179). Running from June 2024 to June 2027, the project is building a clinical-grade SaaS platform that brings AI-driven, EEG-based Alzheimer's screening directly into primary care — delivering results that are 70% cheaper than neuroimaging, 75% faster than standard neuropsychological tests, and accurate to 80–95%.

​

The Challenge: Catching Alzheimer's Before It's Too Late - Alzheimer's disease follows a long, silent prodromal course. The window for effective intervention — and ultimately for disease-modifying therapies to make a difference — lies years before clinical symptoms become apparent. Yet current diagnostic pathways are slow, expensive, and inaccessible at the point of first contact. DECCOG addresses this gap head-on, targeting the transition between Subjective Cognitive Decline (SCD) and Mild Cognitive Impairment (MCI), the critical juncture at which early intervention has the greatest impact.

​​

DECCOG combines Starlab's EEG-based neurocognitive assessment technology with state-of-the-art AI to deliver an automated, cloud-connected decision support tool validated for routine primary care use.

​

At its core is the COGBAT protocol — a streamlined 27-minute EEG battery comprising an N-Back working memory task, an Auditory Oddball paradigm, a Resting-State recording, and an ASSR measurement. The protocol was designed for clinical simplicity: no specialist operator required, minimal patient burden, and automated analysis via the proprietary STARFLOW cloud platform.

​

The AI engine is built on a Normative Models combined with Autoencoders — architectures well suited to capturing the complex, multivariate EEG signatures of early cognitive decline. Models are trained and validated against a harmonized multisite database of over 2,000 subjects (LEMON, ABM, GAADR, BetaAARC, ACE, and BBHI cohorts), providing the statistical depth needed for robust normative modeling.

​

By the close of the project (2027), DECCOG aims to deliver a validated, CE-mark-ready SaaS prototype at TRL 6, validated with 200 patients at BarcelonaBeta clinical facilities. The platform will be ready to be embedded within existing primary care workflows, requiring no specialist infrastructure and providing clinicians with a clear, interpretable risk report within minutes of assessment. DECCOG represents a major step toward making Alzheimer's early detection a routine part of preventive healthcare across Europe.