This month the DHIN is pleased to feature Dr. Shifeng Liu. Shifeng is a Research Associate in the new Biomedical Informatics and Digital Health theme.
Please tell us a little bit about yourself. Hi, I am Shifeng. I love cooking, photography and I am a foodie. I completed my PhD last year and am working as a Research Associate in the Biomedical Informatics and Digital Health Theme. Throughout my PhD I mainly focused on Natural Language Processing (NLP) in computer science. Over the past two years, I became interested in information extraction in the biomedical domain as my first step into digital health. Currently, I am applying the latest NLP techniques in processing text in biomedicine, aiming to extract useful information for understanding and further usage. I enjoy having discussions with academics around, but not limited to, machine learning techniques and NLP
How do you define digital health?
It is hard for me to set a definition of digital health. From my perspective, digital health is a booming and dynamic multidisciplinary area involving computer science, medical science, health science, etc. It could be a toolbox that applies machine learning and communication techniques for telemedicine; speeds up the processing of medical images; extracts information from medical reports and electric medical records. It can be used to analyse reports and thus support decision making for doctors and clinicians. Also, it can help people understand their body and lives via mobile health devices, and also understand medical records with simplified text description!
What do you think will enable digital health projects and innovations to succeed?
Short answer: Data integration from heterogeneous sources. Health related data is represented in multiple formats, including unstructured text, structured table data, images, speeches, etc. And it can originate from diverse resources, such as personal mobile devices, social media, medical records, clinical trial reports, etc. Combining such data could connect multiple resources, discover the undetected relations among them, and thus enable a better and more thorough understanding of health. This also contributes to a sophisticated analysis of multiple streams of data and the development of customer-centric services.
What do you think are the biggest challenges facing digital health at the moment?
The short answer is data understanding and interpretation. In the current stage, I think we are still using the data in a shallow way even though we have plenty of machine learning tools. We have to admit, even for data from a single source, it is still not easy to understand and represent the data that fully fits the diverse requirements from doctors, clinicians, patients, service providers, etc. Due to insufficient data understanding and interpretation, there are still gaps between the collected data and real-world data. And this hinders the expansion and extension of digital health applications.
Do you have any interesting resources or helpful networks people should know about?
As I focus more on how to apply machine learning techniques into digital health, I would recommend these following resources:
- [Book] Machine Learning: A Probabilistic Perspective by Kevin P. Murphy
- [Book] Deep Learning By Ian Goodfellow
- [Book] Speech and Language Processing by Dan Jurafsky and James H. Martin Link: https://web.stanford.edu/~jurafsky/slp3/
- [scholar] ACL Anthology Link: https://www.aclweb.org/anthology/index.html
A very big thank you to Shifeng for taking the time to be our April member feature!