Patterns in Cognitive Rehabilitation of Traumatic Brain Injury Patients: A Text Mining Approach

Traumatic Brain Injury (TBI) is a leading cause of disability worldwide, there is one TBI case every 15 seconds and in every 5 minutes someone becomes permanently disabled due to it. Brain injuries lack of surgical or pharmacological therapies, therefore Cognitive Rehabilitation (CR) is the generally adopted treatment. Computerized CR tasks are increasingly replacing traditional "paper and pencil" approaches. Nevertheless, CR plans are manually designed by clinicians from scratch based on their own experience. There is very little research on the amount and type of practice that occurs during computerized CR treatments and its relationship to patients' outcomes. While task repetition is not the only important feature, it is becoming clear that neuroplastic change and functional improvement occur after specific tasks are performed, but do not occur with others. In this work we focus on the preprocessing, patterns and knowledge extraction phases of a Knowledge Discovery in Databases (KDD) framework. We propose considering CR programs as sequences of sessions and pattern searching (association rules, classification models, clustering and shallow neural models) to support clinicians in the selection of specific interventions (e.g. tasks assignations). The proposed framework is applied to 40000 tasks executions from real clinical setting. Results show different execution patterns on patients with positive and negative responses to treatment, predictive models outperform previous recent research, therapists are provided with new insights and tools for tasks selection criteria and design of CR programs.