Indexing Biosignal for Integrated Health Social Networks

Abstract

Rising medical costs and aging populations are major concerns for most countries, including developed countries. Some studies are now mining Health Social Networks (HSNs) as a way of dealing with these concerns. HSN provides a scalable, cost-effective, and fast method for collecting a large amount of user-generated data. However, patients usually have difficulty finding relevant information from social networks. This study aims to develop an Internet of Things (IoT) approach to find keywords to describe medical conditions using patients' biosignals. This study uses the Convolutional Neural Network (CNN) to encode ECG signals into word embedding vectors. Word embedding is a vector projection of words' sentimental features from a context. Similar keywords can be extracted given a vector. Therefore, keywords can be used to search for information from HSN. The average number of keywords correctly predicted is 2 to 3 out of 5. This approach improves the efficiency and effectiveness of information searching in HSNs using biosignal. This study is the first time that index biosignal in HSN.