A maximal CPET produces tens of thousands of data points; the standard report collapses them to a single number. Aevox surfaces what gets discarded — substrate oxidation, threshold confidence, glycogen-aware fueling, training zones anchored to physiology — in a report your clinicians can defend.
Most reports stop at “consume 30–60 g/h.” Aevox derives a per-athlete glycogen pool from VO₂max, body comp, and substrate oxidation — then tracks depletion across the event distance and intensity the athlete is actually training for.
Same model in the live tool. Glycogen pool scales with VO₂max, mass, age, and sex; depletion follows a published intensity-and-duration curve. Adjust intensity or duration to re-score the prescription in real time.
VE/VO₂ and VE/VCO₂ scatter, with VT1 and VT2 placed by an ensemble of detectors and surfaced with a confidence score. Where detectors disagree, the band widens. Where they agree, you sign off and move on.
Four detectors — V-slope, VE/VO₂ nadir, VE/VCO₂ nadir, excess CO₂ — vote on each threshold. Confidence = % agreement, with a draggable override and full audit trail.
Each card is interactive in the live report and exportable as PDF, JSON, or signed share link.
Where does fat oxidation peak? Where does the crossover happen? At what HR does carb burn dominate? The substrate curve answers all three — and tells the athlete how to fuel it.
Six zones placed off VT1, VT2, fatmax, and peak — not 220-minus-age. Translate physiology into the prescription a coach hands an athlete.
Running economy is the oxygen cost to cover one kilometer. The biggest predictor of race performance after VO₂max — and the only metric that tells you whether your training is making you faster for free.
Aevox started as the analysis pipeline for a graduate research program at the University of Colorado Colorado Springs — not as a SaaS retrofit of a startup idea.
Substrate oxidation and threshold scoring were calibrated in graduate work against Douglas-bag gas collection — the gold-standard reference method. Formal multi-site validation is a planned next step with our pilot partners.
Live in pilot at the Hybl Performance Center via CommonSpirit Health, alongside EXOS in Colorado Springs. Real athletes, real reports, real clinician sign-off.
I built Aevox during my graduate work to score the analyses I kept wishing the cart software did itself. Every formula in this report has a citation behind it — happy to walk you or your medical director through the methods. jack@aevox.health
Threshold confidence scoring, audit trails, methods you can stand behind. The pipeline doesn't replace your clinical judgment — it puts it on a foundation you can sign your name to.
Manual scoring is the bottleneck on a clinical-grade test. Aevox compresses report turnaround from days to minutes — and the report your clinic ships looks like it was built by a research lab.
No. Your cart still runs the test and exports the breath-by-breath data. Aevox reads that export and produces the interpretation report — substrate, thresholds, fueling, zones, economy. You keep your existing protocols and reference equations.
COSMED (K5, Quark, Omnia), ParvoMedics TrueOne 2400, Vyaire Vmax & OxyConPro, MGC CCM Express, Cortex Metalyzer 3B, Geratherm Ergostik, plus generic breath-by-breath CSV. New carts added on request — typically a 2–3 day turnaround.
Once the CSV is uploaded, the full pipeline — threshold detection, substrate oxidation, fatmax search, zone derivation, fueling estimation, and economy quantification — runs in parallel and a clinician-ready report is back in minutes, not days.
We treat every test as PHI. Files are processed in-region, encrypted in transit and at rest, and never used to train shared models. A signed BAA is available for clinical partners on request.
Yes. Every threshold, every formula, every classification has a citation and an audit trail in the report. Detectors and confidence scores are visible — overrides are first-class and recorded with the clinician's name.
Two weeks. Week 1: connect your cart export, configure your reference equations and protocols, cross-check against three of your historical tests. Week 2: live shadow-scoring on new tests with a clinician sign-off step. Then you're in production.
Reference equations are configurable per-protocol. We support pediatric percentile norms (Cooper Institute, Wasserman), and the threshold detector ensemble works the same way at any size — confidence scoring tells you when a test is unusual.
Question we missed? jack@aevox.health.