What are you going to do?
This PhD research initiative aims to develop advanced statistical and machine learning methods to facilitate the early diagnosis of Alzheimer’s disease, a condition that disrupts neural network functionality. Graph-based machine learning techniques are essential for this purpose due to their ability to incorporate network structures. Graph neural networks (GNNs), a subset of deep learning that leverages graph structures, have shown promising results. However, they fall short in quantifying model uncertainty, an essential factor for diagnosing Alzheimer’s disease. Bayesian methods provide mathematically grounded frameworks to address model uncertainty, but often with significant computational demands. The main objective of the research line is to develop a GNN that incorporates Bayesian graphical methods for Alzheimer’s detection. The entire project is divided into two PhD subprojects. The first subproject, currently in progress by an existing PhD student, aims to develop GNNs that are both computationally efficient and grounded in Bayesian principles.
This vacancy is for the second PhD subproject, which intends to apply the Bayesian framework alongside GNNs to analyze real-world data related to Alzheimer’s cases. The PhD student will use the Bayesian graphical method to identify the brain structure of Alzheimer’s patients and patients with a healthy brain. Subsequently, the PhD student will implement a GNN to categorize brain imagery into either typical brain function or Alzheimer’s affected states, using various imaging modalities like MRI, fMRI, and PET, along with non-imaging data such as demographic and genetic information. Our partnerships with hospitals provide us with access to pertinent data for this research.
Tasks and responsibilities
The PhD student will work in close collaboration with the supervisory team, alongside the current PhD student on this project, and additional academic staff. The responsibilities will encompass:
- Developing and applying advanced statistical and machine learning techniques, in particular, Bayesian statistical methods and graph neural networks;
- developing open-access software tools (such as R and Python packages or C++ libraries) for applying the newly developed algorithms/models and techniques to real-world datasets;
- working in close collaboration with the hospital to understand the data and work on data collection and cleaning;
- writing up findings for publication in prestigious machine learning and statistical journals;
- presenting research findings at leading conferences;
- attending classes and seminars (including those offered at other universities) to further develop thinking and research skills;
- conducting teaching (to a limited degree), including undergraduate tutorials and the supervision of BSc dissertation projects.
