From an undocumented sensor board to a calibrated clinical trainer — in three weeks

Situation

• Life-science company in clinical training
• Custom pressure-sensor board, provenance lost
• Silicone training phantom with embedded masses
• No schematic, firmware source or protocol spec
• No calibration certificate
• Original developer no longer available
• Target environment: a consulting room with no network

Challenge

The pad emitted numbers, and nothing was known about them:

• The wire format, framing and sample rate were undocumented
• Channel order did not correspond to physical sensor position
• Readings were dimensionless counts, while the curriculum teaches force in Newtons
• No reference load cell was available to calibrate against
• The square sensor grid had to map onto an anatomical surface
• Results had to reach the study from a machine with no connectivity

What dataMaven delivered

Each layer had to be established by measurement before the one above it was possible:

• Characterised the serial protocol empirically — wire format, framing and rate
• Recovered the channel-to-position map by stimulating each sensor in a known order
• Built a per-sensor force model from repeated loading at known masses
• Fitted response curves with an explicit, measured saturation ceiling per sensor
• Mapped the square sensor grid onto the anatomical outline with a radial warp
• Built a guided exam that advances on the learner’s actions rather than button presses
• Scored coverage and applied pressure separately, against configurable ground truth
• Packaged the whole thing as an offline desktop application with durable local records

Technology

The build, end to end

Outcome

A calibrated trainer running beside the rig, and a written record of what the numbers mean:

• Force reported in Newtons, per sensor, with known error characteristics
• A single decoding module owning the wire format, so firmware revisions are a one-file change
• Live pressure and coverage feedback during the exam
• A debrief expressed as something the learner can go back and do differently
• Every exported record carrying the calibration it depends on
• Fully offline operation, with results accumulating into one collectable file
• Hardware and deployment documentation stating what is measured and what is provisional

Engagement snapshot

Deliverables

• Device Characterisation & Hardware Documentation
• Per-Sensor Force Calibration
• Guided Clinical Training Application
• Operator Tooling & Scoring Configuration
• Offline Desktop Packaging & CI Pipeline
• Deployment Guide

What changed — week by week