Capture
A parent or nurse photographs the Kenya MOH Road-to-Health growth chart with a supported iOS or Android phone. Capture is designed to continue without an internet connection.
How it works
From a paper mark to a verified digital record.
NeoScan’s proposed workflow combines on-device document correction, server-side computer vision, human verification, and WHO growth analysis.
The complete workflow
Ten connected steps. One traceable record.
Technical processing stays behind a simple user experience: take a clear photo, check the values, and understand the trend.
Document detection
Google ML Kit is specified to find the chart boundaries, remove surrounding clutter, and auto-crop the page without asking the user to draw corners manually.
Image preprocessing
The mobile and backend pipeline normalizes orientation, corrects perspective up to the specified 30-degree capture angle, and reduces shadows or glare before inference.
AI chart analysis
OpenCV separates blue or black pen marks from printed chart lines. A custom YOLOv11 model is specified to identify axes, percentile curves, plot points, and handwritten text noise.
Measurement extraction
Detected point coordinates are translated into age-in-days and weight or height values using the fixed geometry and stepped age blocks of the MOH chart.
x: 98 days → y: 6.0 kg
Localized to the active age block before value conversion.
User verification
The app presents each extracted value before saving. A user can correct a misread value; the final record is marked as manually verified and the override is logged.
14 weeks6.0 kgVerified
6 months6.4 kgReview
WHO growth analysis
The verified value is matched to the child’s sex and exact age in days. The WHO LMS formula is used to calculate a Z-score and place the value in a reference percentile band.
Z = ((Y / M)L − 1) / (L × S)
An extended WHO formula is specified for values beyond ±3 SD.
Trend visualization
The app draws the child’s historical points beside WHO 3rd, 15th, 50th, 85th, and 97th percentile curves. Each point can reveal its date, value, and calculated position.
Anomaly flagging
Defined rules can flag a drop across two or more major percentile bands or block a mathematically impossible growth rate. These flags support review; they do not diagnose illness.
Growth trend needs review
A significant change across percentile bands was detected. Please ask a qualified healthcare professional to review it.
Secure sharing
A parent can give explicit, revocable consent to a registered clinic. The design also specifies a password-protected PDF report and timestamped audit records for sharing actions.
Offline sequence
Capture now. Process when connected.
Queued records retain their original scan timestamps so later anomaly checks can run in the correct chronological order.
Offline
Your scan has been saved securely. It is designed to sync automatically when internet returns.
OfflineEncrypted local storageConnection restoredSync
Explore by audience