Action sequencing and error production in stroke patients with Apraxia: Behavioral modeling using Bayesian logic networks

Individuals with Apraxia often suffer from cognitive impairments during the execution of activities of daily living (ADL). In this study, we used a statistical relational learning approach (Tenorth, 2011) to model the behavior of apraxic patients and neurologically healthy individuals (n = 14 in each group) during ADL performance. Video analysis indicated that apraxic patients committed more errors than control participants, typically committing omission, addition, and substitution errors. The results of the Bayesian Logic Network (BLN) approach indicate that the relevance of the nodes (i.e., actions) differed between the control participants and apraxia patients. Furthermore, there were more nodes in the patient group, which is likely a result of addition and substitution errors, or by alternative ways of solving the task using a different set of tools. Overall, the results of the present study highlight the variability inherent in ADL performance, which need to be considered when developing action and error prediction models.