An Unsupervised Data-driven Anomaly Detection Approach for Detection of Adverse Health Conditions in People Living with Dementia: Cohort Study (Preprint)

Abstract

Sensor-based remote health monitoring can be used for timely detection of health deterioration in people living with dementia with minimal impact on their day-to-day living. Anomaly detection approaches have been widely applied in various domains including remote health monitoring. However, current approaches are challenged by noisy and multivariate data and low generalizability.

We aim to develop an online and lightweight unsupervised learning-based approach to detect anomalies representing adverse health conditions using activity changes in people living with dementia. We demonstrate the effectiveness of our method over state-of-the-art methods on a real-world dataset of 9363 days collected from 15 participant households, by the UK Dementia Research Institute between August 2019 and July 2021. Our approach is applied to household movement data to detect urinary tract infections (UTI) and hospitalization.

We propose and evaluate a solution based on Contextual Matrix Profile (CMP), an exact and ultra-fast distance-based anomaly detection algorithm. Using daily aggregated household movement data collected via passive infrared (PIR) sensors, we generate CMPs for location-wise sensor counts, duration, and change in hourly movement patterns for each patient. We compute a single daily normalized anomaly score in two ways: by combining univariate CMPs, and by developing the multidimensional CMP. The performance of our anomaly detection method is evaluated relative to Angle-Based Outlier Detection (ABOD), Copula-Based Outlier Detection (COPOD) and Lightweight on-line detector of anomalies (LODA). We also use the multidimensional CMP to discover and present the important features associated with adverse health conditions in people living with dementia.

The multidimensional CMP yields 84.3% recall and offers the best balance of recall and relative precision compared to COPOD, LODA and ABOD when evaluated for urinary tract infections (UTI) and hospitalization. We validate early AM (midnight to 6 am) bathroom activity to be the most important cross-patient digital biomarker of anomalies indicative of UTI. We also demonstrate how CMP-based anomaly scoring can be used for a cross-patient view of anomaly patterns.

To the best of our knowledge, our work is the first real-world study to adapt the CMP to continuous anomaly detection in a healthcare scenario. The CMP inherits the speed, accuracy, exactness, and simplicity of the Matrix Profile, providing configurability, ability to denoise and detect patterns, and easy explainability to clinical practitioners. We address the need for anomaly scoring in multivariate time series healthcare data by developing the multidimensional CMP. With high sensitivity and a low alert rate, better overall performance than state-of-the-art methods, and the ability to discover digital biomarkers of anomalies, the CMP is a clinically meaningful unsupervised anomaly detection technique, extensible to multimodal data for dementia and other healthcare scenarios.