A Novel Multi-Stage Approach to the Detection of Visuo-Spatial Neglect Based on the Analysis of Figure-Copying Tasks
R. M. Guest, M. C. Fairhurst · 2002 · Proceedings of the Fifth International ACM Conference on Assistive Technologies (Assets '02) · doi:10.1145/638249.638278
Summary
This paper presents a computer-based technique for improving the detection of visuo-spatial neglect (VSN), a dysfunction typically caused by stroke that causes patients to fail to respond to stimuli on the opposite side of the visual field from the brain lesion. Early and accurate diagnosis of VSN is critical because inadequate detection leads to performance deficits during rehabilitation and prolonged recovery times. The conventional assessment — the Rivermead Behavioural Inattention Test (BIT) — relies on subjective evaluation by therapists who score figure-copying drawings based on whether expected components are present, yielding scores from 0 to 4 across six drawings. This subjective approach suffers from poor inter-rater reliability and limited sensitivity. The authors propose a two-stage improvement: first, objective rule-based static analysis that algorithmically determines pass/fail based on whether drawn components meet defined geometric criteria (e.g., for a cross shape, whether five sub-boxes are present and form a cross pattern); and second, novel dynamic analysis that captures timing and kinematic features of the drawing process itself using a graphics digitisation tablet placed under the paper. The key insight is that the process of drawing — not just the final product — contains diagnostic information about VSN.
Key findings
Data was collected from 30 VSN subjects (identified via BIT scores), 58 stroke control subjects (stroke patients without VSN), and 13 healthy age-matched controls across four hospital centres in East Kent. Using the rule-based static assessment alone, 53.3% of true VSN subjects failed the cross-copying test, compared to only 21.1% of the stroke control group — demonstrating that conventional component-based assessment misses nearly half of VSN cases. The critical finding is that among drawings that passed the static assessment (i.e., looked "normal"), dynamic features revealed significant differences between VSN and control groups. VSN subjects showed significantly more pen lifts (11.21 vs 5.84 for stroke controls, p=0.003), longer total movement time (16.30s vs 7.12s, p=0.002), longer total drawing time (14.61s vs 12.17s), and longer overall execution time (30.91s vs 19.35s, p=0.001). These dynamic features indicate that VSN subjects use a disjointed, component-based drawing strategy — drawing each element separately with extended planning pauses between strokes — even when their final drawings appear normal. This means drawings conventionally classified as "normal" by static analysis can be identified as abnormal through dynamic process analysis, substantially improving diagnostic sensitivity.
Relevance
This paper demonstrates the value of analysing the process of task performance rather than just the outcome — a principle with broad implications for accessibility assessment and assistive technology. The finding that nearly half of VSN cases are missed by conventional assessment has direct clinical consequences: undetected neglect leads to poorer rehabilitation outcomes and longer recovery. For accessibility practitioners, the work highlights that visible output may not reveal underlying cognitive or perceptual difficulties — a user may produce acceptable-looking work while struggling significantly with the process. This principle applies beyond VSN to other conditions where task completion masks processing difficulties, such as cognitive fatigue, attentional deficits, or executive function impairments. The computer-based approach also addresses a persistent challenge in clinical accessibility assessment: replacing subjective human judgement with objective, reproducible metrics that can be standardised across assessors and clinical settings.
Tags: stroke recovery · neurological conditions · clinical tools · assessment · automated testing · rehabilitation · visuo-spatial neglect · occupational therapy · computer vision