CONFIDENTIAL
01 / 14
IVF laboratory quality framework

Measure variability.
Preserve biology.
Improve calibration.

Calibryo is a reproducible framework for measuring inter-observer variability in embryology assessment, designed to support laboratory quality systems without imposing external scoring standards.

CONFIDENTIAL
02 / 14
Positioning

Calibryo is not a scoring system.

It is a measurement framework designed to quantify inter-observer variability in embryo assessment while preserving biological interpretability.

Calibryo does not define correctness. It measures agreement structure within a cohort and makes variability visible for quality-management discussion.

Core principle
The laboratory retains clinical authority. Calibryo provides an auditable measurement layer: where reviewers agree, where they diverge, and how these patterns evolve over repeated campaigns.
CONFIDENTIAL
03 / 14
Clinical context

Embryo assessment remains partly subjective.

Embryo morphology and time-lapse event annotation depend on trained human assessment. Inter- and intra-observer variability have been described in the embryology literature, particularly for morphology grading and developmental timing.

In routine laboratory practice, this variability is often discussed qualitatively but rarely measured systematically with a repeatable, exportable and auditable workflow.

“The purpose is not to replace expert judgement, but to make disagreement measurable.”
Calibryo methodological position — internal quality-management use.
CONFIDENTIAL
04 / 14
Reference model

No external gold standard is assumed.

Reference values are computed using a leave-one-out group reference for each reviewer × item × video.

For time-based events, the reference is derived from the group distribution. For ordinal or categorical items, comparison is performed according to predefined item-specific logic.

Why leave-one-out?
The reviewer being evaluated is excluded from their own reference calculation. This avoids circularity and keeps the comparison internal to the cohort without pretending to define an external “truth”.
CONFIDENTIAL
05 / 14
Deviation classification

Transparent categories, not black-box scores.

Aligned
Exact match or within predefined tolerance.
Agreement class
Acceptable
Moderate deviation according to item type.
Agreement class
Major
Deviation requiring discussion or calibration review.
Agreement class
Denominator discipline
Missing, biologically non-applicable and excluded rows are handled explicitly. They are not silently mixed into agreement denominators.
CONFIDENTIAL
06 / 14
Workflow

From campaign setup to auditable report.

1
Configure
Create campaign, define reviewers, videos and item set.
2
Invite
Generate secure personal reviewer URLs.
3
Review
Independent video scoring with autosave and submission.
4
Export
Generate normalized, analysis-ready CSV files.
5
Report
Dashboard, individual reports and trajectory output.
CONFIDENTIAL
07 / 14
Data integrity

Every output is traceable to raw CSV.

01
Raw export
Direct normalized extraction of reviewer submissions.
02
Analysis-ready long table
One row per reviewer × video × item, including comparability and exclusion flags.
03
Summaries and reports
Operator, item, video, dashboard and trajectory outputs generated from the same auditable base.
Design constraint
No hidden transformations. No opaque weighting. No marketing score. Results can be audited independently from exported files.
CONFIDENTIAL
08 / 14
Dashboard

Actionable disagreement, not aggregate judgement.

Avoided by design
Single global score without biological interpretation.
External gold standard claim without validation.
Thresholds presented as normative clinical truth.
Black-box ranking of embryologists.
Calibryo output
Top calibration issues expressed in item, video and biological time units.
Reviewer-specific deviation patterns for training discussion.
Scatter plots against internal group reference.
Methodology shown directly inside the report.
CONFIDENTIAL
09 / 14
Reviewer Trajectory

Calibration can be tracked over time.

The Reviewer Trajectory module adds a cross-campaign layer for monitoring reviewer behaviour over repeated calibration campaigns.

Deviation
Evolution of reviewer distance from group reference.
Longitudinal metric
Bias
Directionality of numeric or time-based deviation.
Longitudinal metric
Drift
Sustained worsening pattern when sufficient real campaigns exist.
Longitudinal metric
Important limitation
The longitudinal layer is implemented structurally. Valid clinical trend interpretation requires multiple real campaigns.
CONFIDENTIAL
10 / 14
Current limitations

Limitations are explicit.

The current deployed dataset includes one real / demo-ready campaign state. Longitudinal behaviour cannot yet be clinically validated from a single campaign.

Simulated trajectory points may be used for interface evaluation and product demonstration, but they are explicitly separated from real data and must not be interpreted as clinical results.

Validation path
Additional repeated campaigns are required to validate trend stability, drift detection, training effects and cross-laboratory generalizability.
CONFIDENTIAL
11 / 14
Technical profile

Designed for low-friction deployment.

Local
Portable Node.js deployment for restricted lab environments.
Deployment mode
Cloud
Railway-hosted demo deployment with persistent storage.
Deployment mode
File
Campaigns remain self-contained folders, without database migration.
Data model
Architecture principle
Template app, runtime campaign folders, export engine, reporting engine and trajectory module remain separated. This preserves stability and enables additive development.
CONFIDENTIAL
12 / 14
Strategic fit

A quality layer for TLM ecosystems.

Calibryo complements time-lapse workflows by focusing on the reliability of human annotation. It does not compete with embryo selection tools; it supports reproducibility of the assessment process.

TLM
For time-lapse platforms
Adds structured calibration and QA value around existing video workflows.
QA
For laboratories
Supports internal training, harmonization and evidence-based calibration sessions.
AI
For future data infrastructure
Improves the traceability and consistency of human-labelled embryo assessment data.
CONFIDENTIAL
13 / 14
Product status

Advanced prototype, ready for external validation.

Operational today
• Reviewer video app
• Secure reviewer URLs
• Autosave and final submission
• Campaign exports
• Dashboard and individual reports
• Reviewer Trajectory module
• EN / FR / DE interpretation layer
Next validation steps
• Repeated real campaigns
• Multi-TLM test files
• Cross-lab usability testing
• Formal validation of drift and trend outputs
• Documentation package for external scientific review
CONFIDENTIAL
14 / 14
Selected references

Bibliography for scientific context.

Alpha Scientists in Reproductive Medicine and ESHRE Special Interest Group of Embryology.
The Istanbul consensus workshop on embryo assessment: proceedings of an expert meeting. Human Reproduction, 2011.
Paternot G. et al.
Intra- and inter-observer analysis in the morphological assessment of early-stage embryos. Reproductive Biology and Endocrinology, 2009.
Baxter Bendus A.E. et al.
Interobserver and intraobserver variation in day 5 blastocyst grading. Fertility and Sterility.
Kirkegaard K. et al.
Time-lapse monitoring as a tool for clinical embryo assessment. Human Reproduction Update.
Note.
This deck deliberately avoids unsupported numerical claims. Quantitative claims should be added only when derived from CALIBRYO exports or explicitly supported by cited peer-reviewed literature.
CONFIDENTIAL
Appendix

Available for scientific review and live demonstration.

Full workflow demonstration: Admin setup, secure reviewer app, reporting pipeline, dashboard interpretation and Reviewer Trajectory module.

Demo deployment
calibryo-production.up.railway.app
Project
CALIBRYO
Document
Investor Desk V5 · Expert-grade