Deep Learning for Medical Imaging module (BE41003)

You will learn fundamental and advanced concepts and methodologies of machine learning for imaging, and relate this to real-world problems in medical imaging and computer vision.

You will experience different approaches to machine learning, including supervised and unsupervised techniques, with an emphasis on optimisation and deep learning methods such as 'convolutional neural networks' and 'generative adversarial networks' (types of artificial neural network.

You will gain the skills to work with X-ray, CT, MRI, and ultrasound image reconstruction and how they are applied in medicine and analysis such as classification, segmentation, and registration applied to datasets in healthcare.

During practical sessions, you will be presented with medical applications whilst working in a team.

What you will learn

In this module, you will:

• be introduced to linear regression and use iterative numerical solvers
• understand convex optimisation and regularisation and use the Stochastic Gradient descent method
   
• be introduced to Convolutional Neural Networks (CNN) and Generative Adversarial Network (GAN)
• learn about deep recurrent neural networks, equivariant learning and generative machine learning
• understand the importance of Support Vector Machines (SVM)
   
• be introduced to CT, MRI reconstruction and unrolling deep iterative methods for image reconstruction

By the end of the module, you will be able to:

• understand machine learning basics and deep feed-forward networks
• use SVM for image segmentation and analysis
• give insight into future trends, limitations, and challenges

Assignments / assessment

• Group assessment (50%)
  • interim presentation (10%): an outline of the topic. The presentation should provide an overview of the design and the used methods
  • coding assignment (20%): implement and reproduce the deep learning techniques associated with the selected topic. Each student will contribute to the development of a shared library of coding functions and scripts in Python related to the project
  • final presentation and report (30%): summarise the project content and all gathered results
• Individual Class Tests (50%)
  • discuss machine learning methods for solving image computing problems and explaining competing approaches

Teaching methods / timetable

• in-person seminars
• workshops
• independent study