Graph Convolutional Neural Networks for Context-based Decision Making in Medical Systems for Disease Prevention

Data-driven medical systems for disease prevention rely on multimodal data streams for decision making. The representations of each data stream can be ordered in graphs. In this topic, students will explore graph convolutional neural networks for deciding steps within a medical system to prevent life-style diseases developing in its users.
Master

Future medical systems for preventing diseases will be based on multiple data streams customized by the user. These multimodal data streams can be data such as heart rate, glucose measurements, and nutritional values. The system must use these data streams to evaluate the user and give advice on how to avoid developing lifestyle diseases. An important distinction from these medical systems an d existing ones is that advice is based individual data and not just population data. Additionally, the granularity of the data streams help decide the accuracy of the evaluation. The objective of the thesis is to combine the data streams into graph representations and apply graph convolutional neural networks for generating advice on life choices for disease prevention. Examples of such advice can be a list of examples of meals with varying nutritional benefits that the user enjoys to make and eat.

Goal

Find new methods for graph-based context-aware decision making within a life-style disease prevention system.

Learning outcome

The student will gain deep insight into knowledge graphs, graph convolutional neural networks and medical disease prevention systems. The student will also gain experience working on a real world system.

Qualifications

  • Proficient programming skills with Python.
  • Basic knowledge of graph theory.
  • Basic knowledge of machine learning theory.

Supervisors

  • Pål Halvorsen
  • Rune Johan Borgli
  • Nitish Nag
  • Ramesh Jain

Collaboration partners

University of California, Irvine

References

ieeexplore.ieee.org/abstract/document/8690196
dl.acm.org/citation.cfm?id=3241913