Early kinematic information and Machine Learning methods allow to detect visual reaching impairments in a patient with Parieto-Occipital lesion

Abstract

Patients with lesions of the parieto-occipital cortex typically misreach visual targets that they correctly perceive (Optic Ataxia, OA). Although OA has been described more than 30 years ago, distinguishing this condition from physiological behaviour using kinematic data is still missing. Here, combining kinematic analysis with machine learning methods, we compared the reaching performance of a patient with bilateral occipito-parietal damage with that of 10 healthy controls (mean age 21.52.0, 9 females and 1 male, right-handed). They performed visually-guided reaches towards targets located at different depths and directions in peripheral and foveal viewing conditions. Using the horizontal and sagittal deviation of the trajectories as input to a linear discriminant analysis (LDA) decoding algorithm, we computed accurate predictions of the patient’s deviations with respect to control’s deviations. For reaching to the left targets, the dataset reveals a considerable inter-individual variability in the predictability of trajectory deviations along movement execution in both peripheral and foveal reaching (chance level 50%). For reaching to the right space targets, instead, dataset and LDA decoding accuracy show low variability, that suggests high inter-individual consistency. In both peripheral and foveal reaching, there was a 100% decoding accuracy after half of the movement, indicating that the classifier was able to perfectly distinguish between patient and control participants (chance level 50%). For the targets located along depth direction and in both peripheral and foveal reaching, after the first 20% of movement, it was possible to discriminate the trajectory deviations of the patient from those of control participants, with the accuracy of classification significantly increasing from near to far targets (chance level 50%). This classification based on initial trajectory decoding was possible for both direction and depth component of the movement, suggesting the possibility to apply this method to characterize pathological motor behaviour in wider frameworks.