Technology and education supporting neurodivergent children

EDUSPEKTRUM

From fragmented data to measurable progress

We develop educational environments and digital tools that support therapists, educators and families working with children and young people on the autism spectrum.

Multimodal progress analysis · child #A-0417 Live
47 sessions 12 weeks 4 data sources
Input4 / 4
  • SI tracking18 sessions
  • AAC sessions14 sessions
  • Education15 sessions
  • Specialist notes31 entries
AI analysisrun R-0417
  • Align timelines
  • Normalize measurements
  • Cross-source correlation
  • Rank recurring patterns
Outputfor review
3 recurring patterns detected
  • Movement stability+14%
  • Task completion trend↑
  • Communication changedetected
Measurements and correlations for specialist review — not clinical conclusions. Illustrative · synthetic data

01 — The problem

A child may work with multiple specialists across sensory integration, communication, education and other therapeutic activities. Each specialist sees only part of the picture.

EDUSPEKTRUM is developing an AI-powered system that connects these fragmented observations and measurements over time.

By analyzing information across multiple sources, AI can identify patterns, correlations and changes that may be difficult to notice when each therapy is evaluated separately.

02 — Two parallel sources

AI provides raw, measurable data independent of the therapist’s subjective assessment.

Measurement
→ Quantitative data

Duration, repetitions, movement and task performance — recorded without a clinical judgment attached.

Therapist
→ Observation & interpretation

Professional experience, context and clinical reasoning.

Specialist decision
Both, side by side

Interpretation and professional decisions remain with qualified specialists.

03 — Pilots & use cases

Where we are applying it

These are pilot implementations in development, not completed studies with published results.

SI tracking · session 12 / 12 Live
Child moving between soft blocks during a sensory integration session, with pose-tracking skeleton overlaid
Pose tracking · 15 joints
Session metricsvs S01
Balance stability 0.82+0.11
Range of motion 118°+9°
Repetitions 24+7
Task completion 86%+19 pp
Session 01 → Session 12
S01S04S08S12
Balance stabilityTask completion
Illustrative visualization of the pilot system. Not real patient data. 12 sessions
Pilot · in development

AI-assisted Sensory Integration tracking

Traditional SI documentation relies heavily on therapist observations and periodic assessments.

We are introducing additional AI-assisted tracking during sensory integration sessions. The system can collect and compare measurable session data over time — such as movement patterns, activity duration, repetitions, task completion and changes in performance.

Instead of replacing the therapist’s assessment, the system provides an additional layer of raw, measurable data that is independent of subjective interpretation.

A therapist can then see
  1. How performance changes across sessions
  2. Whether specific activities correlate with measurable improvement
  3. Trends that are difficult to notice during individual sessions
  4. Differences between observed progress and quantitative measurements
  5. Long-term changes across multiple forms of support
Use case · in development

Connecting multiple therapies

Children often participate in several forms of therapeutic and educational support simultaneously. Important information can remain distributed across notes, documents, assessments and different specialists.

Our AI system is designed to analyze these sources together and search for relationships between them. For example, changes observed during sensory integration sessions can be compared with communication, educational or functional data from other activities.

Session notes Assessments Documentation Measurements

The objective is not to let AI make clinical decisions, but to give specialists a broader and more measurable view of the child’s development.

Cross-therapy analysis · child #A-0417 4 sources
Sensory Integration
AAC
Education
Therapy notes
● AI analysis · cross-source correlation
Shared trend4 of 4 sources
Co-occurring change · W06–W09 · r = 0.62 For specialist review · synthetic data

04 — AI as a second layer of observation

A human specialist naturally focuses on the activity taking place in front of them.

AI can simultaneously analyze a much larger history of measurements and observations.

This allows us to investigate signals that could otherwise be overlooked because they are:

01
Distributed across different therapies
02
Separated in time
03
Subtle individually
04
Visible only when multiple data sources are compared

05 — Inside the system

What specialists actually see

Interface previews from the pilot, shown with synthetic data. No real child is represented.

Depth & pose

3D movement analysis

A depth sensor turns each session into a point cloud. Joint trajectories, symmetry and range of motion are measured frame by frame — the same way in every session, for every child.

  • Tracked joints15
  • Sampling30 fps
  • Outputparameters per session
3D movement analysis · depth frame 01842
Hand R trajectoryFoot L trajectory
Movement parameters
Centre of mass sway2.4 cm
Step length L / R0.41 / 0.38 m
Gait symmetry0.93
Knee flexion, max112°
Reach, dominant64 cm
Cadence98 steps/min
30 fps · 15 joints · 2 840 points Illustrative · synthetic data

Longitudinal view

One child, many sources, one timeline

Weeks of sessions from different specialists are brought onto a single timeline. The system highlights recurring patterns and shows the data behind them — the specialist decides what they mean.

  • Sources combinedSI · AAC · education · notes
  • Patternsflagged for review
  • Identityanonymized
Longitudinal progress · child #A-0417 · anonymized W01 → W12
12weeks
47sessions
4data sources
Movement stability+14%
W01W12
Task completion+19 pp
W01W12
AAC initiations / session3.1 → 5.4
W01W12
Patterns detected3

  1. Task completion is higher in weeks with two or more SI sessions

    r = 0.58 · 9 of 12 weeksFor review

  2. AAC initiations rose from W07, alongside a new classroom routine

    +1.6 per session · 3 sources agreeFor review

  3. Balance variability is lower in morning sessions

    31 morning vs 16 afternoon sessionsFor review
Statistical observations for specialist review. Not a diagnosis. Illustrative · synthetic data

06 — Real-world implementation

Built in real educational environments

Our solutions are being introduced and tested in real educational and therapeutic environments. This allows us to develop AI tools together with specialists, observe how they perform in everyday practice, and continuously improve them based on real workflows and measurable data.

Specialized primary school

ASY — Grodzisk Mazowiecki

EDUSPEKTRUM develops and introduces its solutions in ASY, a specialized school in Grodzisk Mazowiecki supporting neurodivergent students.

The environment allows us to connect educational, therapeutic and longitudinal data and develop tools that help specialists understand progress across different areas of support.

Focus
AI-assisted analysis Progress tracking Specialist documentation Cross-therapy data
Therapeutic and educational preschool environments

Równowaga Preschool Network

Our solutions are also being introduced across the Równowaga preschool network, providing an environment for developing tools for younger children and multidisciplinary specialist teams.

This includes exploring AI-assisted analysis of therapeutic activities, communication and AAC, sensory integration tracking, and longitudinal development data.

Focus
Early development AAC Sensory integration Multidisciplinary data Progress over time

One system — different stages of development

Working across school and preschool environments allows us to study how data, communication and therapeutic support evolve at different stages of a child’s development.

  1. Równowaga Preschool Network
    • Early development
    • AAC
    • SI
    • Therapeutic data
  2. EDUSPEKTRUM AI layer
    • Data integration
    • Measurement
    • Longitudinal analysis
    • Patterns
  3. ASY School
    • Education
    • Therapy
    • Progress tracking
    • Specialist insights

What we do

Four areas of work

01

Education & therapeutic support

We create specialized educational environments designed around the individual needs of neurodivergent children.

02

AI-assisted tools for specialists

We develop tools that help therapists and educators analyze educational and therapeutic information, organize documentation and monitor progress.

03

Communication & AAC

We explore AI-supported augmentative and alternative communication tools designed to improve communication and independence.

04

Research & development

We investigate how artificial intelligence can help specialists identify patterns across educational and therapeutic data and personalize support.

AI supports qualified professionals — it does not replace their professional judgment.

FAQ

Does AI replace the therapist?

No. AI provides additional analysis and measurable data. Interpretation and professional decisions remain with qualified specialists.

What does the system analyze?

Depending on the activity, the system can work with structured session data, measurements, educational information, specialist documentation and longitudinal progress data.

Why combine data from different therapies?

A change visible in one area may be related to activities taking place elsewhere. Analyzing multiple sources together can reveal patterns that are difficult to identify from isolated records.

What do you mean by measurable data?

We distinguish measurements from professional interpretation. The system can record measurable parameters such as duration, repetitions, movement or task performance without assigning a subjective clinical judgment to them.

How is AI used in sensory integration sessions?

We are developing AI-assisted tracking that can capture selected measurable parameters during activities and compare them across sessions. This gives therapists an additional longitudinal dataset alongside their professional observations.

Can AI identify progress that a therapist might miss?

It can identify statistical patterns, changes and correlations across larger datasets that may not be apparent from individual sessions. These findings are presented to specialists for interpretation rather than treated as diagnoses.

Which AI models do you use?

We don’t rely on a single AI provider. Different parts of the system use different tools: computer-vision models for pose estimation and movement tracking, statistical and time-series methods for longitudinal analysis, and large language models — such as Claude, alongside other commercial and open-source models — for working with documentation and specialist notes. Each model is chosen for a specific task and checked against specialist annotations. Where possible, sensitive data such as video is processed on infrastructure we control.