Causal representation learning for robust healthcare predictions

As part of the Canon Medical and RAEng funded project on Healthcare AI [1], we are looking to expand the team with a highly motivated PhD student working on disentangled representations, computer vision, and medical image analysis to start as soon as possible but no later than September 2022.

We frequently hear that AI will aid the work of experts. To succeed in this an AI must make accurate and confident predictions, generalizing to new and unseen data. Yet we still train AI models via supervision (i.e. by providing several examples) and require diverse data to do so. There must be a better AI that can understand concepts in the data even without explicit supervision, and behave well even on unseen data. The limits of supervision are particularly profound when we consider a more multimodal view to sensory input using both visual and text input, or consider asking an AI to make predictions about the future. Ultimately our goal is to develop decision support systems that can predict and assess the influence of healthcare treatments and interventions. To achieve this, this AI must be able to distinguish correlation from causation and to jointly learn the concepts and any causal relationships between them and the desired output task or tasks. 

Thus, topics of interest include:

  •  representation learning (to learn concepts);
  •  causality (to learn relationships between concepts);
  •  generative models (to constrain the representations within a distribution space); and
  • digital twinning of data generation and acquisition processes (to augment what we can describe physically and incorporate within the causal structure).

The candidate will:

  • join VIOS [3] a dynamic, international team of several academics, postdocs and students working in the area of healthcare and deep learning at the University of Edinburgh –a powerhouse in AI, engineering and medicine;
  • contribute to an exciting project where deep learning meets imaging and other healthcare data;
  • collaborate with the Edinburgh-based R&D centre [4] of Canon Medical –a conglomerate in healthcare;
  • interact with several clinical experts within the UK, Europe, the USA and Japan. 

Notes:

[1] Project page https://vios.science/projects/raeng

[2] Tutorial on disentangled representations, https://arxiv.org/abs/2108.12043 and https://vios.science/tutorials/dream2021

[3] VIOS website https://vios.science

[4] Canon Medical Research Europe website https://research.eu.medical.canon

Further Information: 

Candidates must apply via the UoE system “see the click here to apply”. The candidate can email Prof. Tsaftaris for inquiries. In the latter include a CV, mention this position in the subject, but note that this does not constitute a formal application.

The University of Edinburgh is committed to equality of opportunity for all its staff and students, and promotes a culture of inclusivity. Please see details here: https://www.ed.ac.uk/equality-diversity

Closing Date: 

Wednesday, February 16, 2022

Principal Supervisor: 

Assistant Supervisor: 

Eligibility: 

Minimum entry qualification - an Honours degree at 2:1 or above (or International equivalent) in a relevant science or engineering discipline, possiblysupported by an MSc Degree. Further information on English language requirements for EU/Overseas applicants.

A first class (or strong 2:1) honours degree or Distinction Masters level degree in Engineering, Computing, Mathematics, Physics, or relevant discipline is required. Candidates with an MSc equivalent training will be preferred. Demonstratable evidence of knowledge of machine learning (e.g. via coursework, projects, publications, work experience), and computational frameworks such as PyTorch, TensorFlow (e.g. via coursework or public repositories) are required. Evidence of prior publications of high caliber (e.g. in computer vision, image analysis or processing or machine learning) are desired but not essential criteria. The candidate should have a high level of analytical and investigative skills and a strong mathematical background. Ability to work within a team, collaborate and inspire others are essential criteria; thus, good communication and desire to own the project are sought-after abilities. 

Funding: 

Tuition fees + stipend are available for Home/EU and International students.

This position is fully funded for 42 months (3.5 years) and is open to all students with a preference for UK/EU nationals. International tuition fees can be covered for exceptional candidates. Funding source: Canon Medical and RAENG.

Further information and other funding options.

Informal Enquiries: