Truace
Subscribe

Machine learning classification of ADHD using a virtual email sorting task

Source article: Virtual reality (bio-)markers differentiate ADHD from depression and controls

The heterogeneity of attention-deficit/hyperactivity disorder (ADHD) and its symptom overlap with other mental disorders limit the clinical utility of isolated diagnostic markers. Virtual reality (VR)-based paradigms offer potential to aid individual diagnostic decisions. This study investigated the diagnostic value of classification models based on multimodal data derived from a virtual email sorting task (VEST). Data were pooled from two studies and used to distinguish unmedicated participants with ADHD (n = 4…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1999· cite & verify
Virtual reality (bio-)markers differentiate ADHD from depression and controls
230315-F-ZB805-0222 - Shaw AFB stays postured at Red Flag-Nellis 23-2 by Staff Sgt. Madeline Herzog. Public domain
Trace impact readingContested
P 68The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 69The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

Researchers pooled data from two studies and had participants complete a virtual email sorting task while task performance, head movement and eye tracking were recorded. Participants also rated their task-related ADHD symptoms. The groups were unmedicated ADHD, medicated ADHD, major depressive disorder and controls.

Support vector machines were trained with nested cross-validation to compare objective, subjective and combined inputs. The results point to VR with machine learning as a possible aid to diagnostic decisions, but the samples were small and one comparison failed to beat chance, so clinical readiness is not shown.

Main points

  1. The sample comprised 43 unmedicated ADHD, 23 medicated ADHD, 16 major depressive disorder and 21 control participants.
  2. Self-reported inattentiveness alone gave the best ADHD versus control result, at 0.85 balanced accuracy.
  3. A 13-feature model mixing subjective and objective measures separated ADHD from depression at 0.75.
  4. Objective-only features gave 0.64 for medicated versus unmedicated ADHD, not significantly above chance.

The gain

Support vector machine models trained on VR task data distinguished ADHD from controls (0.85 balanced accuracy) and from major depression (0.75), suggesting a route to more precise diagnosis.

The problem

A model using only objective VR measures did not separate medicated from unmedicated ADHD participants significantly above chance (0.64 balanced accuracy), limiting its ability to capture medication effects.

The rundown

The study starts from the problem that ADHD is heterogeneous and overlaps in symptoms with other mental disorders, which limits the usefulness of any single diagnostic marker. VR paradigms are proposed as a way to support individual diagnostic decisions.

During the virtual email sorting task, the team recorded task performance, head actigraphy and eye tracking, and also assessed task-related ADHD symptoms by self-report. Models were then compared across objective, subjective and combined parameter sets.

The authors conclude that self-report is useful for detecting ADHD, while objective measures show promise for distinguishing disorders. Their medication-effect result was not significant, so that use remains unconfirmed.

What this doesn’t fix

The medicated versus unmedicated ADHD model did not perform significantly above chance, and the group sizes are small, so the findings are preliminary.

Sources

  1. Peer-reviewedEuropean Archives of Psychiatry and Clinical Neuroscience2026-10-14

The debate