Mid-Parental Height
Uses parent heights to estimate a target-height range. It is a simple first-pass summary of family height genetics.
Model card
How does the AI work?
This page explains the calculation and AI layers inside Çocuk Gelişim in plain language. The platform provides decision support; it is not diagnosis, treatment or emergency triage.
Methods we use
Khamis-Roche
Combines age, sex, height, weight and parent heights. It is one adult-height method that does not require bone age.
Bayley-Pinneau
Uses bone age as an input. Hand-wrist X-ray interpretation belongs in clinical context and is not a standalone diagnosis.
Neyzi + Ensemble
Uses the Neyzi 2008 Turkish reference and agreement across methods to produce a confidence label.
Bone-Age AI
Provides an assistive signal from hand-wrist X-rays. It does not replace a radiology report or clinician judgment.
Limits and safety rules
Outputs are probabilistic and never guarantee final adult height.
Measurement error, missing parent heights, rapid weight change or chronic disease can distort predictions.
Precocious puberty, delayed puberty, falling growth velocity or eating-disorder concern requires clinician assessment.
Branch development context never produces suitability scores or ranking; read it with the child’s enjoyment, goals, load and safety.
Data and privacy
Personal profiles, child measurements and reports stay in authenticated areas. Federation and annual reports apply a k >= 10 anonymity gate before showing aggregate groups.
Clinical governance
The platform keeps scientific-board oversight and clinical warnings visible. Model outputs only make sense alongside clinician assessment, history, examination, labs and imaging.
How is a prediction produced, step by step?
- 1
Data entry
The family enters birth date, sex, height-weight measurements and optionally parent heights and Tanner stage. Measurement-technique hints are shown in the form.
- 2
Validation and warnings
Inputs are checked against age/validity ranges. Out-of-range ages or unusual values do not block prediction; they attach a visible warning to the result.
- 3
Multi-method computation
Every applicable method (MPH, Khamis-Roche, Neyzi percentile projection, Bayley-Pinneau when bone age exists) is computed separately with its own 95% confidence interval.
- 4
Ensemble combination
Method outputs are combined with prior weight × precision (narrower confidence intervals get more weight); agreement across methods drives the confidence label.
- 5
Report and sharing
The result is reported with its interval, warnings and per-method contributions. Families can forward the report to their clinician via in-platform messaging.
Method validity ranges
Out-of-range inputs do not block a prediction; the platform attaches a visible warning to the result instead. Flagged results should not be interpreted without clinical assessment.
| Method | Main inputs | Valid range | Known limitation |
|---|---|---|---|
| Mid-Parental Height | Mother + father height | Any age | Produces a wide range; ignores the child’s own growth data. |
| Khamis-Roche | Age, sex, height, weight, parent heights | 4.0–17.5 years | Error grows around the pubertal spurt (11–15 y); developed on US data. |
| Bayley-Pinneau | Bone age + height | Bone-age table range | Sensitive to the bone-age method used (Greulich-Pyle / TW). |
| Neyzi percentile projection | Height, age, sex | 0–18 years (Neyzi 2008, Türkiye) | Assumes the percentile channel stays stable; deviates when a child crosses channels. |
| Bone-Age AI | Hand-wrist X-ray | Pediatric hand radiographs | Assistive signal only; never replaces a radiology report and is sensitive to image quality. |
KVKK rights and data control
Data export: families can download all profile, measurement and report data in machine-readable form from the data-export page in profile settings.
Account deletion: permanent deletion is requested from the account-deletion flow in profile settings and removes child profiles and measurements.
Anonymity gate: federation and organisation reports enforce k ≥ 10 — no group smaller than 10 children is ever shown.
Access boundary: child data is visible only in authenticated accounts and to clinician/coach roles the family explicitly invites.
Health-form content is never sent to product analytics; analytics is limited to anonymous usage measurement.
Frequently asked questions
•Is the height prediction exact?
No. All methods are probabilistic and reported with a 95% confidence interval. The interval is more informative than a single number; no method guarantees final adult height.
•Does Bone-Age AI diagnose?
It does not. It produces an assistive bone-age signal from a hand-wrist X-ray and never replaces a radiology report or clinician assessment.
•Who can see my child’s data?
No one by default. Data stays in the authenticated account; only a clinician or coach the family invites can see it. Organisation reports apply a k ≥ 10 anonymity gate.
•Can I export or delete my data?
Yes. Data export and permanent account deletion are self-service in profile settings, and can also be requested through support.
•What should I do when a result carries a warning?
A warning means an input is outside the method’s validated range or a pattern needs attention. Review flagged results together with your pediatric clinician.
How to read a result
Interpret the interval, warnings and agreement between methods together instead of treating one number as certain. Contact a pediatric clinician when growth slows unexpectedly, puberty timing looks unusual or chronic disease symptoms are present.