Detection of mild cognitive impairment based on mouse movement data of trail making test

Abstract

Mild cognitive impairment (MCI) has 10%–20% prevalence in the population above the age of 65, and a significant portion of these people will go on to develop dementia later in their lives. However, if MCI is detected early, preventative measures can be taken to delay the onset of severe symptoms. Current diagnostic methods for MCI are not suitable for regular wide-scale screening. Advances in machine learning algorithms in combination with digital movement data offer rich possibilities for automated MCI detection. This paper introduces a machine learning model that effectively predicts MCI based on only a few seconds of computer mouse movement. To our knowledge, studies directly comparable to ours have not been done before. On a dataset of 70 participants, we demonstrated 80% accuracy in distinguishing healthy controls from patients with MCI. This gives an opportunity to develop a cost-efficient and easy-to-use screening method that could aid the work of healthcare professionals.