Early diagnosis of mild cognitive impairment: A case study in approaches to inductive-logic programming

Recent rapid advances in data collection routines in clinical science have led to a trend of storing patient data in a heterogeneous database. The lack of existing computing tools to enable operability to use machine learning on these heterogeneous data sources is a barrier to the healthcare sciences. Healthcare data is usually complex and highly context-dependent, and it requires modern computational tools to handle the complexity of such data. This study sought to utilize the data collected from virtual reality (VR)-based software and a leap-motion device used for learning in mild cognitive impairment (MCI) cases to enable early detection of MCI by analyzing the classification rules for errors (action slips) based on finger-action transitions when performing instrumental activities of daily living (IADL). Finger motion was recorded as a time-series database. An induction technique known as Inductive-Logic Programming (ILP), which uses logical and clausal language to represent the training data, was then used to discover a concise classification rule using logical programming. We were able to generate rules on how action transitions of the finger in the experiments were related to the pattern of micro-errors that indicate the difference of error regarding the length of the no-motion state of the finger.