Superior classification of visuospatial neglect in the subacute phase through free-viewing and considering overcompensation.

Abstract

Visuospatial neglect is a common disorder of spatial attention after stroke. Traditional paper-and-pencil tasks are widely used for diagnosis but show limited sensitivity in the subacute phase, partly due to compensatory and (over)compensatory strategies. State-of-the-art measures, including gaze bias assessed with eye-tracking during free viewing and refined line bisection scoring, may offer greater diagnostic accuracy but have rarely been tested in clinical samples. We compared traditional measures (Centre of Cancellation, directional bisection error) and state-of-the-art measures (endpoint weightings bias, mean horizontal gaze position during free viewing) in stroke patients admitted to rehabilitation (n = 34) and age-matched healthy controls (n = 32), examining which measures best discriminated neglect from non-neglect in the subacute phase. We also tested whether direction-independent (absolute) deviations better capture neglect than directional deviations, given that overcompensation can mask deficits in standard scoring. Direction-independent scores showed superior discrimination. Furthermore, a logistic regression model based on state-of-the-art measures outperformed one based on traditional measures. The strongest neglect detection was achieved by combining directional bisection error with mean horizontal gaze position. These findings support the complementary value of eye-tracking-based assessment alongside traditional neglect tasks, and highlight the importance of accounting for (over)compensatory behaviour.