Cognitive prognosis of acquired brain injury patients using machine learning techniques

Abstract

The cognitive prognosis of acquired brain injury (ABI) patients is a valuable tool for an improved and personalized treatment. In this paper, we explore the task of automatic cognitive prognosis of ABI patients via machine learning techniques. Based on a set of pre-treatment assessments, distinct classifiers are trained to predict whether the patient will improve in one or any of three cognitive areas: attention, memory, and executive functioning. Results show that variables such as the age at the moment of the injury, the patient's etiology, or the neuropsychological evaluation scores obtained before the treatment are relevant for prognosis and easily yield statistically significant accuracies. Additionally, the prognostic relevance of these and other variables is studied by means of standard feature selection methodologies. The outputs of the present paper add to the discussion on current cognitive rehabilitation practices and push towards the exploitation of existing technologies for improving medical evaluations and treatments.