No abstract

With the development of the Internet of Medical Things (IoMT), indoor wireless sensor networks (WSNs) have been used to monitor Alzheimer's disease patients daily and help guide their behaviors. Alzheimer's disease may seriously impact patients’ memory, and thoughts of “what should I do” can unexpectedly form in their mind. This cognitive impairment can affect patients’ independence and well-being. As a basic infrastructure for future healthcare systems, a WSN can collect patient behaviors, such as their positions and states, to support safety and health analyses. Therefore, this paper proposes a probabilistic model checking-based method to predict patient behaviors and detect abnormal behaviors related to mild cognitive impairment to help patients rebuild their confidence and perception. First, the layout of the home environment is abstracted as a formal grid, and a user activity model (UAM) is proposed in the form of a discrete-time Markov chain (DTMC) to describe patients’ activity based on data collected by sensors. Second, because Alzheimer's patients with mild cognitive impairment (A-MCI) often forget their next daily activities, we classify and describe their daily behaviors as verification requirements in the form of probabilistic computation tree logic (PCTL). Then, the UAM is input into a probabilistic model checking tool and compared against the verification property to calculate the associated probability values and assess temporal behaviors; the activity with the largest probability is selected for behavior guidance. Third, we demonstrate the process of detecting abnormalities, including activities with abnormal temporal behaviors and activities with normal temporal behaviors but unexpected probabilities that may be repeated more than twice. The key states are extracted from the UAM to specify the verification properties for abnormality detection. Finally, a case study is presented to demonstrate the usability and feasibility of our proposed method.