Artificial Intelligence (AI) in population studies
We go through the various phases of observational studies based on primary and secondary data sources and present some state-of-the-art tools for collecting, washing, compressing, and analysing heterogeneous data.
This course is a part of
Observational population studies are key to gaining insights into safety and effectiveness of drugs and other treatments, as well as into the consequences of interventions. Traditionally, observational studies are carried out on well-curated data from registers. The digitalisation of the public and health care sector opens the opportunity to collect a huge amount of data of hitherto unseen depth, but also with the problems of irrelevant, missing, inaccurate, and erroneous data presented alongside the information carrying and clinically relevant evidence.
We go through the various phases of observational studies based on primary and secondary data sources and present some state-of-the-art tools for collecting, washing, compressing, and analysing heterogeneous data. Our focus is on examples from pharmacovigilance based on professional health care data but will also touch upon data from other sources such as IOT devices and the internet.
The course will introduce state-of-the-art technologies of AI based on deep learning. It will go through the various architectures and packages for data analysis.
Very good overview of all methods, very useful.
Course director on AI in population studies
Course details of AI in population studies
Key benefits - judge the capabilities of technologies
After the course, you will
- Have knowledge of the IT infrastructure of the healthcare sector in Demark
- Have knowledge of the Danish and international pharmacovigilance system
- Have knowledge of basic pharmacological concepts to better interpret and design observational studies
- Have a basic knowledge of the legal requirements for performing observational studies.
- Have insight into the workings of deep learning tools for health care data analysis
- Have experience with using tools for natural language processing, image processing, and representation learning
- Be able to judge the capabilities of technologies
- Have experience with training machine learning systems
- Have experience with drawing conclusions from observational studies
Course content - AI-based analysis of healthcare data
Throughout the course, we will use real and synthetic health data to illustrate all faces of observational studies and the use of AI tools to implement these.
Core elements
- Observational studies and statistical methods and power
- Data collection from the national and local data resources
- Legal aspects of the use of health data
- Data wash
- Representation learning of data features for use in statistical analysis. We relate the traditional statistical tools to their modern counterparts such as autoencoders and self-supervised learning methods
- Natural language processing. We give examples on how to analyse electronic health care records for finding diagnostic information.
- Image analysis. We give examples of how to transform visual information in medical scans into quantitative biomarkers.
- We give examples of real-world studies on the consequence of Covid-19 infections, on adverse effects of surgery and on side effects of drugs.
We introduce tools and platforms for representation learning based on transformer architectures. We introduce deep learning tools for your own simple programming. We introduce tools for controlled training and testing of deep learning tools. We will use simple scripting in Python.
Participant profile
The participant is expected to have a basic statistical understanding and have experience with scripting in Stata, R, Python or similar. The participant will have the need for AI-based analysis of healthcare data and can potentially bring their own application to the course as example for some of the exercises.
Location
University of Copenhagen
South Campus, Faculty of Law
Njalsgade 76
DK-2300 København S
Denmark
Contact
Copenhagen Summer University
csu@adm.ku.dk
+45 3533 3423
Time and Date
21-25 August 2023
09:00-16:30