Real-Time and Self-Adaptive Stream Data Analysis

Abstract

In recent years, with the advances in hardware technology, abundant medical surveillance data streams can be easily collected using various kinds of medical devices and sensors. Accurate and timely detection of abnormalities from these physiological data streams is in high demand for the benefit of the patients. However, the state-of-the-art data analysis techniques face the following challenges: First, the raw data streams are in sheer volume, which is attributed to both the number and the length of the data streams. Massive data streams pose a challenge to storing, transmitting, and analysing them. Second, multiple physiological streams are often heterogeneous in nature. These data streams collected from different devices have different value ranges and meanings. Domain knowledge is required for fully understanding them. Third, multiple physiological data streams are not independent. As a matter of fact, they often exhibit high correlations. Abnormalities can be evidenced not only in individual stream but also in the correlation among multiple data streams.