Overview of My Research Projects:

Seizure Prediction in Epilepsy

October 01, 2023

Research Assistant, Supervisors: Prof. Raymond HF CHAN, Prof. Ho Man CHAN, Prof. Bee Luan KHOO, HK COCHE

This is part of Ms. HUANG Ke’s PhD program at City University of Hong Kong, where we are focused on using EEG and other multi-modal signals from open-access datasets to predict epileptic seizures using machine learning models. My contributions to this project include the following tasks:

1. Dataset Selection: I conducted extensive research to identify suitable open-access datasets and made selections based on EEG channels.
2. Feature Extraction: I utilized EEGLab and the Discrete Wavelet Transform (DWT) method to extract relevant features from the EEG data.
   
3. Deep Learning-Based Feature Extraction: I explored deep learning techniques for feature extraction from the multi-modal signals.

Cuffless Blood Pressure Estimation Using Multi-modal Bio-signals

August 01, 2023

Research Assistant, Supervisor: Prof. ZHANG Yuanting(张元亭), HK COCHE

This is part of Ms. XIANG Ting’s PhD program at City University, where we employ a variety of physiological signals, including PPG, ECG, IPG, and temperature, to estimate blood pressure using both machine learning and mathematical models. The data for all subjects was obtained from COCHE Lab, and my contributions to this project encompass the following tasks:

1. Data Preprocessing: I handled the preprocessing of the multi-modal data, ensuring its suitability for analysis.
2. Feature Extraction and Normalization: I extracted relevant features from the data and normalized them for consistency.
3. Machine learning models: I employed a combination of a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) model to extract features and estimate blood pressure.
4. Results from Multiple Input Signals: I fed various signal combinations and individual signals into the model, and the outcomes are as follows:

Non-invasive Blood Glucose Prediction

July 01, 2023

Research Assistant, Supervisor: Prof. Kannie WY CHAN, HK COCHE

This project is part of the HKSTP Ideation program, and our primary goal is to enhance the accuracy of non-invasive daily blood glucose estimation methods. My specific contributions to this project involve:

1. Clinical Protocol Development: I’ve created a comprehensive Clinical Protocol to guide our research and experiments.
2. Rapid Blood Glucose Monitoring: We’ve been utilizing the Contour Plus Elite system for swift blood glucose monitoring, making the process efficient and effective. You can see an example of a blood glucose test in the image below:

Cardiovascular Disease Classification

June 01, 2023

Research Assistant, Supervisor: Dr. CHEN Hanjie(陈翰杰), HK COCHE

In this project, we utilized the MIMIC-IV dataset to develop a machine learning model for cardiovascular disease (CVD) classification. My responsibilities encompassed:

• Identifying cardiovascular disease (CVD) subjects based on ICD codes.
• Exploring the construction of a knowledge graph to identify potential associations with the CVDs’ occupation.

Cuffless Blood Pressure Estimation using PPG and ECG

January 24, 2023

Research Assistant, Supervisor: Prof. ZHANG Yuanting(张元亭), HK COCHE

This research marked a significant milestone in my transition from Nursing to Biostatistics, as it introduced me to the world of signal processing and machine learning for disease estimation. I am deeply grateful to Prof. ZHANG for his invaluable guidance and patience throughout this project. My work during this period encompassed three main components:

1. Replicating Research on Blood Pressure Estimation: In this part, I focused on replicating a paper that used machine learning models to estimate blood pressure (BP) based on photoplethysmogram (PPG) and electrocardiogram (ECG) signals. My tasks included:
   • Data Preprocessing: I worked with a dataset of 600 patients from the MIMIC-II database.
   • Feature Extraction: I extracted 13 essential features from the PPG and ECG signals, including pulse transit time (PTT), various amplitudes (T-wave, Q-wave, R-peak, S-peak, Low-peak), QRS intervals, R to low-peak amplitudes, R-peak intervals, PPG pulse amplitude, PPG pulse width, PPG high to low interval, and PPG slope change standard deviation.
   • Machine Learning Models: I applied several machine learning models in BP estimation, including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Random Forest (RF), and Linear Models. Notably, the LSTM model outperformed the others.
2. Investigating the Relationship Between PPG, ECG, Reference BP, and Estimated BP:
   • I utilized LSTM and Random Forest (RF) models for estimating blood pressure waveforms.
   • I calculated the correlation coefficient (CC) for Reference BP & Estimated BP, PPG & Estimated BP, and ECG & Estimated BP.
3. Optimal Feature Set for Blood Pressure Estimation:
   • I Conducted a review of a published article: Beat-to-beat continuous blood pressure estimation with optimal feature set of PPG and ECG signals using deep recurrent neural networks.
   • I performed statistical analyses in this article.

Construction of a Knowledge Graph-Based Chronic Disease Risk Prediction Model and Health Management Pathway in the Context of Big Data

January 01, 2022

Undergraduate, Supervisor: Prof. SUN Hongyu(孙宏玉), Peking University, China

This research is a four-year project funded by the National Natural Science Foundation of China, with a focus on public health management. My responsibilities during the preparatory phase included:

• Utilizing Selenium and Requests to extract chronic disease risk factors from web content.
• Analyzing the collected data through natural language processing.
• Constructing a knowledge graph related to chronic disease risk factors and health management using the spaCy library.

Public Risk Perception, Behavior Patterns, and Emotional Guidance Under Major Public Health Emergency

July 01, 2021

Undergraduate, Supervisor: Prof. SUN Hongyu(孙宏玉), Peking University, China

This research project, supported by the Beijing Natural Science Foundation Committee, involved a qualitative study concentrating on the public’s preventive actions and responses to major public health emergencies. We conducted in-depth interviews across China and meticulously analyzed the gathered data to enhance our research. My contributions to this study encompassed two primary areas:

1. Public Reaction to COVID-19:
   • Conducted a comprehensive search for COVID-19-related literature regarding public risk perception and coping behaviors, utilizing databases such as PubMed. Conducted quality assessments and extracted relevant content.
   • Employed Stata to perform meta-analysis, subgroup analysis, and meta-regression, summarizing the findings and identifying sources of heterogeneity within the included literature.
   • Executed sensitivity analyses and Egger’s tests to ensure the robustness of our research outcomes.
   • Consolidated the above findings and authored a paper as the first author. This paper received citations from the World Health Organization (WHO).
2. Healthcare Providers’ Responses during Public Health Emergencies:
   • Developed an interview protocol for interactions with healthcare providers and conducted interviews in Beijing to collect research materials.
   • Applied Colaizzi’s 7-step analysis method and utilized NVIVO for key information extraction.
   • Formulated a time-based risk perception model.
   • Contributed to the paper’s development process.

Home Care Guide for the Elderly with Dementia

October 01, 2019

Undergraduate, Supervisor: Prof. WANG Zhiwen(王志稳), Peking University, China

This research project was conducted as part of the National College Student Innovation and Entrepreneurship Practice in China. It centered on the creation of a home care guide for elderly individuals with dementia. My responsibilities within this project included:

• Conducting a comprehensive review of dementia home care research on scholarly platforms, evaluating literature quality with input from three experts, and extracting essential content.
• Crafting the guidance document based on the findings from my review and valuable insights from clinical experts and professors.
• Utilizing AutoCAD software to produce 2D and 3D illustrations for the guide.