Start with the question
Define the knowledge gap, intended population, protocol, measurement plane, task, and decision before choosing a tool.
Research & Validation
Methods should be inspectable before results are trusted.
Biosystems One develops engineering software with explicit intended uses, deterministic verification, documented limitations, and a path toward independent, protocol-specific validation.
Defined scopeKnown ground truth where possibleProvenance retainedLimitations published
01 / Research philosophy
Build knowledge before building claims.
Scientific credibility does not come from polished outputs alone. It comes from a traceable relationship between a question, a method, the data, an analysis, and the limits of the conclusion.
Our software is designed to help qualified users inspect that relationship—not conceal it behind a single score or automated recommendation.
Keep provenance visible
Distinguish manually placed, tracked, constrained, estimated, interpolated, and user-entered observations.
Separate verification from validation
Correct software behavior on a known fixture is necessary evidence, but it does not establish fitness for every real protocol.
Publish the boundary
Limitations, exclusions, uncertainty, and unresolved validation work belong beside the result—not in hidden fine print.
02 / Validation methodology
A staged evidence model.
Each stage answers a different question. Passing one stage does not substitute for the next.
- 01Requirements and intended use
What should the method do?
Define the users, input conditions, outputs, tolerances, exclusions, and decision context before evaluating performance.
- 02Deterministic verification
Does the implementation behave as specified?
Use known inputs and ground truth to test calculations, thresholds, error handling, provenance, and repeatability.
- 03Reference comparison
How does it compare with an independent method?
Compare against independently digitized recordings, established reference implementations, or qualified manual calculations.
- 04Protocol-specific validation
Is it suitable for this population and protocol?
Evaluate representative tasks, raters, capture conditions, missing data, and failure cases using a pre-defined analysis plan.
- 05External review and reporting
Can others examine the evidence?
Report datasets, methods, version identifiers, uncertainty, deviations, and limitations in a form suitable for independent review.
RMS error95th-percentile errorLost-track rateFalse-lock rateInter-rater variability
03 / Intended use
Use depends on product and protocol.
These statements define the current product boundary; they do not replace an institution’s own method review.
Open beta
MotionLab
Intended: Two-dimensional, single-camera analysis of visible landmarks in gait, cycling, occupational movement, teaching, and related technical recordings.
Not intended: Validated three-dimensional motion capture, observation of hidden landmarks, or unreviewed clinical or safety decisions.
Launch preparation
ErgoThrive
Intended: Structured occupational-ergonomics evidence, method inputs, calculation review, revisions, and reporting by qualified professionals.
Not intended: Automatic determination that a job is safe, representative, compliant, or suitable for a worker or population.
Invite-only closed beta
HF Studio
Intended: Controlled workflow and evidence management for qualified medical-device human-factors teams and supervised education.
Not intended: Clinical decisions, automatic safety or regulatory conclusions, certification, QMS replacement, or guaranteed regulatory acceptance.
Catalogue foundation
B1 Ergonomics
Intended: Evidence-aware discovery and review of workplace-equipment records using visible source and evaluation states.
Not intended: Purchasing, supplier fulfilment, automatic product recommendations, compatibility decisions, or claims that demonstration fixtures are evaluated products.
Validation stage
MotionLab Performance
Intended: Private on-device review of supported training movements and cautious longitudinal comparison.
Not intended: Diagnosis, injury prediction, medical decision-making, hidden-landmark measurement, or unreviewed coaching conclusions.
04 / Known limitations
A result is only as strong as its acquisition and assumptions.
These are active engineering boundaries, not hypothetical disclaimers.
Two-dimensional observation
Perspective, camera alignment, lens distortion, scale calibration, variable frame rate, and out-of-plane movement can affect measurements.
Appearance can change
Large displacement, rotation, scale change, deformation, blur, repetitive backgrounds, and occlusion can require manual correction.
Landmarks remain estimates
Pose confidence is not anatomical ground truth. Hidden points are not observed, and AI estimates require protocol-specific comparison with manual digitization.
Observation quality governs scoring
Assessment engines cannot determine whether a sampled posture, load, frequency, duration, or task is representative.
Fixtures are not populations
Passing deterministic tests establishes regression behavior for defined inputs; it does not establish general accuracy or external validity.
No automatic acceptance
No platform output by itself proves safety, compliance, certification, clinical suitability, or submission readiness.
05 / Sample datasets
Small, inspectable fixtures.
These files are synthetic examples for software verification and teaching. They contain no human-subject, clinical, workplace, or customer data.
CSV
Public sample
MotionLab synthetic trajectory
Ten illustrative schema rows from a synthetic cyclic marker path, including ground truth, estimate, error, frame time, and provenance. This is not the underlying 36-frame regression fixture.
- Version
- 1.0
- Size
- 948 bytes
- Content
- Synthetic
- SHA-256
ba0f868c…d14ff830
JSON
Public sample
ErgoThrive RULA fixture
A deterministic input and expected result for the RULA-1993/1.2 calculation engine. The fixture expects a score of 3.
- Version
- 1.0
- Size
- 1,026 bytes
- Content
- Synthetic
- SHA-256
afa91b44…5b8b606
Dataset boundary
These samples are suitable for inspecting schemas and teaching verification concepts. They must not be described as clinical, epidemiological, workplace, population, or product-validation datasets.
06 / Example reports
Reports show scope and limitations beside results.
Open, print, or save these example engineering notes. They contain synthetic fixtures only.
Example report
MotionLab synthetic tracking validation
Method, recorded fixture result, limitation statement, and recommended external-validation measures.
Open printable report →Example report
ErgoThrive calculation verification
Engine identifier, illustrative inputs, expected result, and qualified-ergonomist review boundary.
Open printable report →07 / Publications
Referenced methods and published foundations.
Biosystems One has not listed a peer-reviewed company publication at this time. The products rely on established methods and literature, with implementation-specific qualifications documented separately.
RULA: a survey method for the investigation of work-related upper limb disorders
L. McAtamney and E. N. Corlett. Method reference for the ErgoThrive RULA engine.
PubMed / DOI →Rapid Entire Body Assessment (REBA)
S. Hignett and L. McAtamney. Method reference for the ErgoThrive REBA engine.
PubMed / DOI →Applications Manual for the Revised NIOSH Lifting Equation
T. R. Waters, V. Putz-Anderson, and A. Garg. Official method guidance used as a reference for lifting analysis.
CDC / NIOSH →Adjustments to Zatsiorsky–Seluyanov’s segment inertia parameters
P. de Leva. Anthropometric reference used in MotionLab’s explicitly labelled kinetics estimates.
DOI →Biomechanics and Motor Control of Human Movement
D. A. Winter, fourth edition. General biomechanical measurement and analysis reference.
Publisher record →A reference identifies a methodological source. It does not imply author, publisher, agency, or institutional endorsement of Biosystems One or its products.
08 / Technical notes
Implementation evidence, written for review.
Technical notes describe software behavior and engineering decisions; they are not substitutes for independent studies.
Tracking pipeline and deterministic validation
Pyramidal Lucas–Kanade estimation, normalized cross-correlation refinement, ambiguity checks, forward–backward rejection, occlusion handling, pose-assist provenance, and known limits.
Read example note →Registry-driven ergonomics engines
Typed method contracts, runtime input validation, deterministic tables, version identifiers, immutable result revisions, and method-specific qualifications.
Read example note →Human judgment and traceability architecture
Reviewer-owned decisions, exact-version relationships, standards metadata, controlled outputs, and explicit prohibition of automatic compliance or safety conclusions.
Review HF Studio boundary →Evidence-aware catalogue states
Typed records, visible source quality, neutral unevaluated states, and explicit separation between catalogue fixtures and governed claims.
Review B1 Ergonomics boundary →Private on-device performance analysis
Raw-video non-retention, local session history, experimental metric labels, and infrastructure for future reference comparison.
Review MotionLab Performance boundary →09 / Future publications
Questions we intend to answer with evidence.
These are publication priorities—not completed studies, accepted manuscripts, or guaranteed timelines.
- 01Protocol design
MotionLab protocol-specific accuracy
Compare calibrated 2D video measurements against independently digitized recordings across defined movement speeds, marker conditions, cameras, and raters.
- 02Domain review
ErgoThrive scoring concordance
Evaluate calculation-engine outputs, recommendation wording, and inter-rater workflows with qualified ergonomists using controlled fixtures and representative tasks.
- 03Usability research
HF Studio traceability workflow
Study whether qualified teams can build and review connected human-factors evidence with fewer missing relationships and clearer decision provenance.
- 04Reference comparison
MotionLab Performance measurement agreement
Compare supported on-device movement metrics with independently reviewed reference recordings across defined capture and activity conditions.
- 05Catalogue governance
B1 evidence-quality framework
Evaluate whether visible evidence and uncertainty states improve equipment-review decisions without implying endorsement or automatic suitability.
- 06Education
Reproducible teaching modules
Develop openly documented laboratory exercises that separate measurement, uncertainty, software verification, and protocol validation.
Research collaboration
Bring a defined question.
Share a non-confidential overview of the hypothesis, population, reference method, available data, institution, and expected research output. We will assess methodological fit and a responsible next step.
- Joint validation protocols
- Independent replication
- University teaching and capstone work
- Technical notes and peer-reviewed publication planning