Technology & Architecture

The 5-Layer Stack
Behind D-BAS™

A vertically integrated architecture — from custom sensor hardware to closed-loop AI feedback — designed for clinical-grade biosignal intelligence at scale.

Explore Architecture Scientific References →
Biosignal Channels

12+ Simultaneous Channels

The D-BAS™ hardware captures the following biosignal modalities in real time, with sub-millisecond synchronisation across all channels.

🧠
EEG
Electroencephalography — neural oscillations
❤️
ECG / HRV
Cardiac rhythm & autonomic tone
💧
EDA
Electrodermal — sympathetic arousal
🩸
PPG
Photoplethysmography — peripheral circulation
💪
EMG
Electromyography — neuromuscular signals
🌡️
TEMP
Multi-point thermal mapping
🫁
RESP
Respiratory rate & depth
🔬
SpO₂
Blood oxygen saturation
5-Layer Architecture

Every Layer Engineered
for the Next

Each architectural layer is optimised not in isolation, but as a component that maximises the fidelity and utility of its downstream layer.

01
Hardware

Signal Acquisition

The foundation of D-BAS™ is a proprietary multi-sensor array engineered to medical-grade tolerances. Sensor fusion at the hardware level ensures that all modalities share a common time base — a prerequisite for valid cross-channel correlation.

The hardware stack employs low-noise amplification circuits, active shielding, and adaptive impedance matching to ensure signal quality is maintained regardless of environmental conditions.

Active EMI shielding Sub-ms synchronisation
Channels12+
Sampling rate (EEG)1000 Hz
ECG resolution24-bit ADC
Time sync tolerance< 0.5 ms
ReferenceIEEE TBME ↗
02
Data

Multi-Channel Recording

Simultaneous capture of all biosignal streams into a unified data frame. A proprietary time-domain alignment algorithm ensures cross-channel phase coherence — essential for computing features such as neurocardiac coupling and sympathovagal balance.

Data compression, buffering and real-time streaming are handled by an embedded DSP layer before transmission to the processing pipeline.

Phase coherence On-device DSP Lossless compression
Data throughput~48 MB/session
Buffer depth30-min local
Compression ratio8:1 lossless
ProtocolBLE 5.2 / USB-C
03
AI / ML

AI Signal Processing

The processing layer employs a multi-modal transformer architecture trained on proprietary and publicly licensed biosignal datasets. It extracts biomarkers invisible to conventional threshold-based analysis — including ultra-low-frequency heart rate variability, cortical coherence indices, and cross-modal entropy signatures.

Models are continuously retrained on anonymised, consent-gated user data, ensuring consistent improvement over time.

Transformer architecture On-device inference Federated learning
Model typeMulti-modal Transformer
Inference latency< 50 ms
Biomarkers extracted40+
Accuracy (validation)
04
Modeling

Digital Twin Modeling

Every user generates a living digital biological model — a personalised Digital Twin — that is seeded from the first session and evolves continuously with each subsequent measurement. The twin encodes individual baseline profiles, deviation thresholds, and longitudinal trends across all tracked biomarkers.

Unlike population-average models, the D-BAS™ Digital Twin adapts to each individual's unique physiological signature, yielding a precision that is impossible with conventional reference range approaches.

Personalised baseline Longitudinal evolution Anomaly detection
Model update freq.Per session
Tracked parameters80+
Retention periodLifetime (user-owned)
StorageEncrypted cloud + local
05
Feedback

Closed-Loop Feedback

The final layer closes the loop: insights from the Digital Twin are translated into actionable, personalised interventions delivered in real time. These may include acute alerts, lifestyle recommendations, sleep optimisation protocols, or physician-ready biological state reports.

The closed-loop architecture mirrors principles from responsive neuromodulation systems, adapted for non-invasive consumer and clinical contexts.

Real-time alerts Physician reports Intervention engine
Alert latency< 2 s
Report formatsPDF / HL7 FHIR
Biological state report exportHL7 / SMART on FHIR
Intervention types12 categories
See the Products → Scientific References →