Pen kinematics for clinical research

Every pen stroke is a measurement.

A measurement platform for research teams in psychiatry and neurology. Run standardized drawing tasks on ordinary tablets, in the browser, and get comparable motor measures across sites and visits.

How a study runs Become a design partner
Runs in the browserNo installs on study devices
Certified per deviceSampling and calibration checked
Open metric definitionsVersioned, citable, reproducible
Built for multi-site studiesPseudonymous codes, CSV export
Duration
0.0s
Path
0mm
Mean speed
0mm/s
Variation
0%
Pen lifts
0
Time in air
0.0s
Sample rate
0Hz
Capture quality
–
DEMO · synthetic trace
The problem

Motor side effects are studied with tools that don’t travel.

Psychiatric medication can cause tremor, slowing and shrinking handwriting. Studies that measure these effects still rely on clinician rating scales or desktop software tied to one tablet in one lab.

Rated by eye

Coarse clinical scales depend on who scores them and miss small changes.

Locked to one lab

Desktop tools need installation and specific hardware at every site.

Devices differ

Sampling rate and screen size change between tablets, so raw numbers don’t compare.

Metrics are opaque

When the computation is hidden, results are hard to reproduce or publish.

How it works

From protocol to dataset in four steps.

The research team keeps ownership of the study, its ethics approval and its participants. Ductus is the measurement instrument.

  1. Set up the study

    Choose the task protocol, sites and number of sessions. Participants get pseudonymous codes.

    Research team

  2. Certify each tablet

    A two-minute check measures sampling rate and timing jitter, then calibrates millimetres with a ruler.

    Site staff

  3. Run sessions

    The tablet locks to the protocol. Every stylus sample is recorded with a quality score for the session.

    Participant

  4. Analyse and export

    Metrics are computed with a fixed, versioned definition and exported as CSV for statistical analysis.

    Research team
What we measure

Three motor tasks. Every measure has a unit.

A short protocol, about five minutes in total, chosen for sensitivity to tremor, slowing and loss of rhythm.

Archimedes spiral

≈ 90 s · both hands

Tracing a printed spiral from the centre outwards.

Tremor frequency
Hz
Tremor amplitude
mm
Path regularity
mm

Lines between targets

≈ 60 s

Straight strokes between pairs of dots.

Stroke velocity
mm/s
Endpoint accuracy
mm
Oscillation at stop
Hz

Alternating tapping

2 × 10 s

Tapping two targets in turn as fast as possible.

Tapping rate
taps/s
Rhythm variability
%
Decline over time
%
Capture qualityRecorded for every session
Hz
Effective sampling rate
ms
Timing jitter between samples
0–100
Session quality score with invalid-session flags
Reproducible by design

Same participant, two tablets, matching numbers.

Results are only useful if another site can get them too. That is why device checks, open definitions and quality flags are built into every session.

Device certification

Each tablet passes a sampling and timing check and a ruler calibration before it can join a study.

Open, versioned metrics

Every measure is defined in an open-source library. Exports record the version used, so results can be cited.

Quality flags

A model trained on labelled sessions flags unusable recordings, such as an unfinished task or the hand resting on the screen.

metrics / tremor_frequencyv1.0
# Dominant tremor frequency in the spiral task
input:      spiral trace, resampled to 200 Hz
signal:     radial deviation from ideal path (mm)
filter:     band-pass 3–15 Hz, 4th-order Butterworth
spectrum:   Welch, 2 s windows, 50% overlap
output:     peak frequency (Hz) + peak power
valid_if:   capture_quality ≥ 70
            and trace covers ≥ 3 turns

Illustrative definition. Published definitions live in the open-source repository.

Flagship protocol In design

Motor side effects of psychiatric medication, designed together with clinical researchers. Sessions are repeated across the study so each participant is compared with their own baseline.

For research use

  • Not a medical device and not for diagnosis
  • No treatment recommendations or alerts
  • Used within studies approved by the research team
Technology

Built on Microsoft Azure.

Each study is isolated, data stays in an EU region and is encrypted at rest, and participants are identified only by code.

Static Web AppsCapture app and researcher portal
FunctionsSession intake and study logic
Container AppsSignal processing service
Blob StorageRaw stylus traces, per study
Cosmos DBStudies, sessions, metrics
Entra External IDResearcher accounts and roles
Machine LearningQuality model, versioned runs
Key VaultKeys and secrets
Application InsightsMonitoring and alerts
Log AnalyticsAudit trail and incidents

Tablet → Functions → Blob + Cosmos DB → Container Apps (metrics) → Machine Learning (quality) → Researcher portal → CSV export

Get in touch

Running a study on motor effects? Let’s talk.

We are looking for research groups to shape the flagship protocol with us as design partners.

Contact

Write to the team about a study, a pilot or the protocol.

contact@ductus-research.com

Source code

The metric library and the capture app are developed in the open.

github.com/~example~