From Lab to Wearable
The Neuropic Pod represents seven years of clinical research into mast cell activation syndrome (MCAS), neuroinflammation, and the immune system's role in chronic illness. Every biomarker we track, every algorithm decision, every threshold—informed by peer-reviewed studies, clinical collaborations, and direct patient data from over 2,000 MCAS patients across 15 medical centers.
Traditional wearables measure surface-level metrics: step count, heart rate, sleep. They're designed for the general population and utterly insufficient for patients with complex immune dysregulation. The Pod goes deeper. It captures the subtle signals that matter to MCAS patients—the early warning signs that appear minutes or hours before a full flare: skin conductance shifts linked to mast cell degranulation, temperature fluctuations revealing inflammatory cascades, heart rate variability indicating systemic stress and immune load before any noticeable symptoms appear.
The journey from bench to wearable required solving a fundamental problem: how do you detect mast cell activation non-invasively? Clinically, MCAS is diagnosed through tryptase levels, histamine metabolites, and cell surface markers—all requiring blood draws and laboratory analysis. But patients need real-time insight into their own immune activation, moment by moment, day and night. That's what drove our research.
We began with a simple question: if mast cells release their granules during activation, what measurable changes occur at the skin surface? Years of research in our lab, combined with clinical partnerships at leading immunology centers, revealed that the autonomic nervous system—which controls skin conductance, temperature, and heart rate—responds acutely to mast cell degranulation. These responses precede clinical symptoms by minutes to hours. That observation became the foundation of the Pod.
The Pod's hardware represents precision engineering at a scale never before attempted in a consumer wearable. We developed custom biosensors with femtofarad-level sensitivity, thermal sensors accurate to 0.05°C, and optical heart rate sensors that work through different skin tones. But hardware alone isn't enough. The real innovation lies in the software—the algorithms that transform raw signals into actionable medical intelligence.
The Five Biomarkers
Our multi-year research program identified five precision biomarkers critical to understanding mast cell activation in real time. These biomarkers were selected through a rigorous process: we analyzed data from over 50,000 hours of continuous monitoring in clinical settings, cross-referenced findings with published MCAS literature, and validated each marker's predictive power against clinical flare documentation from our patient cohort.
Skin Conductance (EDA)
Skin conductance, also called electrodermal activity, measures the electrical properties of your skin—which change dramatically when sweat glands activate. In MCAS patients, mast cell degranulation triggers parasympathetic and sympathetic nervous system responses, causing measurable changes in skin conductance within 30-90 seconds of activation. Our research found that skin conductance spikes during flares are 3-5x larger in MCAS patients compared to healthy controls, and the pattern of activation is distinct enough to differentiate mast cell activation from anxiety or normal stress responses. The Pod's EDA sensors achieve microsiemens-level resolution, allowing us to detect the subtle tone shifts that precede overt flares.
Dermal Temperature
Local skin temperature at the application site reveals inflammatory cascades before systemic temperature changes. When mast cells degranulate, they release mediators including histamine, tryptase, and heparin—all of which trigger inflammatory cascades. These cascades increase local blood flow and metabolic activity, raising skin temperature by 0.2-0.8°C within minutes. Critically, this local temperature rise occurs before systemic symptoms like flushing or fever. We've validated this across multiple patient cohorts: temperature inflection points preceded patient-reported flare onset by an average of 47 minutes, giving patients and their care teams precious warning time.
Heart Rate Variability (HRV)
HRV—the variation in time between heartbeats—is a direct measure of autonomic nervous system tone. In healthy individuals, your heart rate varies slightly beat-to-beat, reflecting a balanced parasympathetic and sympathetic nervous system. In MCAS patients during flares, this balance destabilizes dramatically. HRV drops sharply, reflecting acute sympathetic activation. But more importantly, the pattern of HRV changes is distinct. We've developed proprietary algorithms that analyze HRV across multiple time windows—from beat-to-beat variation to hourly patterns—to distinguish mast cell-driven HRV changes from anxiety, exercise, or other stressors. This multi-scale analysis gives us 87% sensitivity and 91% specificity for detecting MCAS flares in our validation cohort.
Movement & Posture
The Pod's inertial measurement unit (IMU) tracks three-axis acceleration and rotation, revealing how physical activity correlates with flare timing. This serves two critical functions. First, movement data helps us distinguish trigger-induced flares from baseline fluctuations—if a flare coincides with a positional change or activity shift, we can flag environmental or postural triggers. Second, during flares, patients often exhibit characteristic movement patterns: reduced activity, postural shifting, freezing. Recognizing these patterns helps us confirm flare detection with higher confidence. We've also discovered that certain movement-immune correlations are patient-specific; the Pod learns your individual patterns over time, increasing personalization of alerts.
Respiration Rate
Respiratory rate, derived from IMU data and validated against optical sensors, captures another autonomic response to mast cell activation. During flares, patients often experience tachypnea (rapid breathing) or dyspnea (shortness of breath), driven by both direct mediator effects and autonomic activation. Respiration rate changes typically lag other biomarkers by 1-3 minutes but provide crucial confirmation. In combination with the other four markers, respiration rate helps our algorithms distinguish true flares from false positives, improving specificity without sacrificing sensitivity.
AI Trained on Clinical Data
A wearable sensor is only as good as the intelligence that interprets its signals. We invested three years developing and validating our machine learning pipeline—far longer than typical consumer wearable development. Here's why: we refused to compromise on medical accuracy.
Our training dataset consists of 50,000+ hours of continuous biosensor data from MCAS patients, paired with ground-truth labels: precise timestamps when patients experienced flares (documented in real-time via our app and verified against clinical records), and periods of stability. We also incorporated data from related conditions—POTS, Long COVID, ME/CFS—where similar immune dysregulation plays a role, allowing our model to learn generalizable patterns while maintaining MCAS-specific sensitivity.
The model architecture itself is novel. We employ a multi-task learning framework: one branch predicts flare probability (is a flare happening now?), another predicts flare risk (will a flare occur in the next 1-6 hours?), and a third identifies which biomarker is driving the alert (skin conductance spiking? Temperature shifting? Multiple markers?). This architecture provides both immediate alerts and longer-horizon warnings, giving patients and providers time to prepare.
We validate our model using rigorous cross-validation: we train on 80% of our patient cohort and test on the remaining 20%, using proper temporal splits to prevent data leakage (training on future data and testing on past data). Across our validation set, we achieve 89% sensitivity and 94% specificity for flare detection—meaning the Pod catches nearly 9 in 10 flares while maintaining a false-positive rate low enough that patients don't suffer from alert fatigue.
But raw model performance isn't enough. We also studied false positives and false negatives in depth. What conditions trigger false positives? Intense exercise, anxiety, certain foods, caffeine, medication changes. We built secondary classifiers to contextualize alerts, reducing false positives by another 30% while maintaining sensitivity. False negatives? Typically occur during atypical flare presentations (patients with very mild symptoms) or when a patient forgets to wear the Pod. We actively flag gaps in data and prompt users to maintain consistent wear.
The model also learns and adapts. Each patient's immune system is unique. Over the first 30 days of Pod use, our personalization algorithm learns your individual baseline—your resting heart rate, your normal temperature patterns, your typical HRV range. Subsequent flare detection becomes increasingly personalized, improving specificity. A patient whose normal HRV is naturally low will have different thresholds than a patient whose HRV is naturally high. The Pod adjusts.
Continuous Monitoring, Real Time
The Pod runs 24/7, passively capturing data with zero manual input. No button presses. No apps to open. No diary entries to fill out. Just continuous intelligence flowing in real time to your phone, your provider, and your care team.
Operationally, this requires solving several hard problems. The Pod samples its five biomarkers at 100 Hz (100 times per second) for electrodermal activity, 50 Hz for temperature and heart rate, and 20 Hz for movement and respiration. That's approximately 500 MB of raw data per day—far too much to store on-device or transmit continuously. So we implemented on-device signal processing: the Pod's embedded processor compresses and analyzes the raw signals in real time, extracting relevant features (not just raw data) every 5 minutes. This reduces the data footprint by 99%, from 500 MB to 5 MB per day, while preserving all clinically relevant information.
Those 5-minute feature windows are where our machine learning model runs. Every 5 minutes, the Pod scores the probability of an active or imminent flare. When that probability exceeds your personalized threshold, the Pod triggers an alert—which appears on your phone, syncs to our cloud platform, and can be pushed to your care team's dashboard in real time (with your permission).
The alert itself is information-rich. It doesn't just say "flare detected." It tells you: which biomarkers are elevated (skin conductance spiking? Temperature rising? HRV dropping?), the confidence level of the detection (89% confident vs. 99% confident), your estimated flare severity (mild vs. severe), how this compares to your recent flare history, and suggested next steps (contact your provider? Take your prescribed medications? Track your symptoms?). This context is crucial for clinical decision-making.
Alert latency is sub-2-minute from signal acquisition to notification. You experience a flare. Within 90 seconds, the Pod has detected it, run our machine learning pipeline, and sent an alert to your phone and provider's dashboard. This latency is critical—in a 2-hour flare, 90 seconds of warning enables interventions that might prevent progression.
The Pod also integrates with your electronic health record (EHR) system, with your permission. Your provider sees Pod alerts in their clinical dashboard, alongside your vitals, medications, and appointment history. This transforms MCAS management from reactive (you wait for symptoms, then contact your provider) to proactive (the Pod alerts you and your provider, you intervene early).
Privacy by Design
Wearing a device that monitors your physiological signals 24/7 raises serious privacy questions. We designed the Pod with privacy as a first-class concern, not an afterthought.
All raw biosignal data processing happens on the Pod itself, before any data leaves your wrist. The Pod's embedded processor runs our machine learning models locally. You experience a potential flare, the Pod analyzes it, and determines a flare probability. Only the summarized result (flare/no-flare, confidence level, which biomarkers elevated) ever leaves the device. Your raw heart rate samples, your raw skin conductance waveforms, your raw temperature readings—none of that ever transmits to our servers or the cloud.
What data does leave the Pod? Metadata about your flares and your health patterns. For each flare alert: timestamp, severity estimate, biomarker contributions, your current activity level (from movement data). Over time, this metadata patterns reveal your triggers and cycles—information that's clinically useful and de-identified (no PII, no raw signals). This metadata is encrypted in transit and at rest.
Your raw data never leaves your Pod or phone (which stores a local copy of recent Pod data for your reference). If you choose to share your data with your care provider, you explicitly authorize a one-time or recurring sync to your provider's secure system. You can revoke access at any time. You can also request complete data deletion—all your Pod data, gone, immediately—though we encourage you to think carefully before doing so, since your data helps train the next generation of MCAS diagnostics.
We also built transparency features. You can see, at any time, exactly what data the Pod has collected about you. You can export your full dataset. You can audit the algorithms we run on your device (yes, really—we provide the model weights and source code). This transparency is essential for medical devices; you deserve to understand what's running on your body.
What's Next
The Neuropic Pod launches this year as a wearable for MCAS patients. But our vision extends far beyond a single condition.
Mast cell dysfunction is implicated in a growing set of chronic diseases. Long COVID patients show elevated mast cell activation markers. POTS (postural orthostatic tachycardia syndrome) involves mast cell-driven vascular dysregulation. ME/CFS (myalgic encephalomyelitis/chronic fatigue syndrome) shows immune activation patterns similar to MCAS. We're currently running prospective studies in each of these populations, validating whether the Pod's biomarkers transfer. Early data is promising.
Beyond MCAS and related conditions, we're exploring whether similar immune activation patterns appear in other diseases. Autoimmune conditions. Allergic disorders. Even neuroinflammatory diseases. The Pod's architecture—multi-biomarker sensing + on-device AI + clinical-grade validation—could become a platform for detecting immune activation across many conditions.
Technically, we're working on next-generation capabilities. A future version of the Pod will include skin impedance measurements, enabling us to assess barrier function (relevant in atopic dermatitis and other skin conditions). We're exploring non-invasive glucose sensing, relevant to metabolic dysfunction in ME/CFS. We're developing algorithms for detecting specific flare triggers by correlating Pod data with environmental sensors (pollen counts, air quality, temperature).
Our ultimate goal: a wearable that doesn't just alert patients to flares after they start, but predicts them hours in advance. We're not there yet—but our 6-hour predictive model (still in validation) is showing promise. Imagine waking up Monday morning and your Pod tells you: "Based on your patterns over the past 72 hours, you're 72% likely to experience a moderate flare between 2-6 PM. Consider reducing stressors this afternoon and have your medications ready." That's the future we're building toward.
The Pod represents a new model for how wearables should work: clinically rigorous, privacy-respecting, transparent, and genuinely integrated into patient care. It's not a gadget. It's a medical tool. And it's only the beginning.