New biomedical imaging and sensing technologies are generating data at an unprecedented pace. The increasing availability of electronic health records data has enabled data-driven inquiry into contemporary health care issues. Researchers in Duke BME are developing innovative data science, machine learning, and digital health modeling approaches to transform multi-scale biomedical data (e.g. multi-omics, imaging, wearable sensors, and electronic health records data) into actionable health insights.
Beginning in Spring 2019, Duke’s Department of Biomedical Engineering (BME) will offer two new hands-on courses in biomedical and health data sciences. Please see the linked course offerings document to learn more or continue reading below.
1. Data Science and Health (Spring 2019)
Instructors: Xiling Shen and Michael Gao (and Jessilyn Dunn)
Course websites (containing syllabus, projects, homework, resources):
Introduction: The goal of this course is for students to learn the tools needed to succeed in a modern data science environment. We will cover some topics at a high-level, focusing on building intuition, as well as dive into some important topics in detail. By the end of the course, students should be able to take a data science project from asking the right questions, through the data processing and modeling, and into a concise and well-delivered presentation that highlights their work. Practical skills that will be taught include:
Data Science Programming (Python or R)
Data Visualization (Seaborn, ggplot2)
Literate Programming (Jupyter Notebooks, RMarkdown)
Software Containerization (Docker)
Modeling (Scikit-learn, Various R packages)
Delivering effective presentations
In addition, the aim of this course is not to be exhaustive. It is impossible to cover all aspects of data science and healthcare in a single semester course. Rather, it aims to develop the data science mindset, as well as empower you to be able to seek the answers to questions that you may have on your own in an effective way.
2. Data Engineering With Tensorflow (Fall 2019)
Instructors: Xiling Shen and Ouwen Huang (and Jessilyn Dunn)
Course website (containing syllabus, course notes, resources): https://bme.ouwen.io/
Machine learning has appreciated a wide range of applications. Industry has developed many software tools and best practices to make data science reproducible, scalable to large datasets, and easy to productionize. This course will focus on applying various deep learning architectures using Tensorflow to solve problems in healthcare, and how to bring these models into production environments. Students will learn the pipeline of: (1) data acquisition, (2) data wrangling, (3) modeling, (4) performance assessment, and (5) deployment. The class is largely project based and will focus on software engineering, DevOps, databases, networking, and lightly touch on concepts in deep learning.
Duke BME is launching the new curriculum in biomedical and health data sciences in partnership with the Duke Institute for Health Innovation (DIHI). Students will learn to develop state-of-the-art machine learning technology using large, de-identified clinical datasets from Duke Health procured exclusively for this program. They will also have the opportunity to work alongside physicians on real clinical problems at Duke University Hospital.