The SINEW project represents a large-scale study leveraging ambient intelligence devices and artificial intelligence techniques for non-intrusive sensing of human daily activities and cognitive health assessment in a home-based environment. 

It aims to determine the utility of continuous activity monitoring for early detection of mild cognitive impairment, a window of opportunity for timely intervention, and whether behavioral change patterns can be used to classify between normal cognition, mild cognitive impairment, and dementia. SINEW is arguably the first longitudinal study focusing on MCI detection through sensor data automatically captured from a sizable number of real homes. 

Whereby, data collection is conducted through an Internet of Things (IoT) architecture, which consists of a combination of wireless devices, an intelligent gateway, and a cloud server that serves as the backend of the home-based sensor network systems. In this study, the biomarkers are used to capture different aspects of daily living that include physical, activity, memory, and sleep. Specifically, a daily record of the digital biomarkers for a person captured in a day consists of a total of 20 biomarker features 

The SMU team provide technical expertise for this study by ensuring safe implementation and maintenance of the sensors in the homes of the participants, provide the sensor obtained data to the clinical team and apply artificial intelligence methods for predictive modelling. Our team obtained promising results showing that sensor-captured digital biomarkers are indeed indicative of the cognitive statue of the participants and can be exploited effectively to detect cognitive decline.

SINEW sensor network system is designed to collect the raw sensor readings data from the home of an individual participant and transmit the data to the cloud-based database via in-house gateway device. The raw sensor readings are then further processed to extract the digital biomarker features.

It consists of sensor network deployment, monitoring, and maintenance, as well as developing various heuristics in biomarker feature extraction and exploring various machine learning models to handle sensor data with a high missing rate. By developing a suite of customized feature extraction methods and predictive modeling techniques using machine learning models which are able to handle missing data, we have obtained promising results showing that sensor-captured digital biomarkers are indeed indicative of the cognitive statue of the participants and can be exploited effectively to detect cognitive decline.

References

• [2024/6] Ah-Hwee Tan, Weng-Yan Ying, Budhitama Subagdja, Anni Huang, Shanthoshigaa, Tony Chin-Ian Tay and Iris Rawtaer. Predicting Mild Cognitive Impairment through Ambient Sensing and Artificial Intelligence. 2nd IEEE International Conference on Artificial Intelligence (CAI 2024), Singapore, 25-27 June 2024. [PDF]

• [2024/4] Tony Chin Ian Tay, Ah-Hwee Tan, Rathi Mahendran, Tze Pin Ng, and Iris Rawtaer. In-Home Sensors for Monitoring Cognitive Decline: A Community-Based Study Protocol Paper. Alzheimer’s Association International Conference Neuroscience Next, April 22-25, 2024. (Poster)

• [2022/11] Seng-Khoon Teh,  Iris Rawtaer and Ah-Hwee Tan. Predictive Self-Organizing Neural Networks for In-Home Detection of Mild Cognitive Impairment. Expert Systems With Applications., Vol. 205, 2022, https://doi.org/10.1016/j.eswa.2022.117538.  [PDF]