Individual projects
Deep learning for medical image segmentation and reconstruction
Background: Over the last decade, machine learning techniques have demonstrated the potential to learn complex models and information from images. Deep learning in particular has been shown to be a powerful tool in computer vision applications such as image classification or face recognition. However, the application of deep learning in medical imaging is challenging due to the large size of the images (millions of voxels in 3D or 4D) and the small size of the training databases.
Challenge: The aim of this project is two-fold:
- To implement deep learning techniques to build generative models from medical images.
- To use the generative models for image segmentation and reconstruction.
Skills: The project will require a good understanding of machine learning techniques as well as some knowledge of computer vision. The project also requires very good programming skills.
Visualization of massive-scale medical image datasets
Background: UK Biobank is the world’s largest population health study, collecting health and lifestyle records from over 500,000 subjects in the UK. For a subset of 100,000 subjects UK Biobank is also collecting medical images in form of Magnetic Resonance Images (MRI) of the brain, heart and whole-body. State-of-the-art machine learning techniques for dimensionality reduction such as manifold learning provide powerful approaches to uncover the relationship between images (i.e. their similarities and dissimilarities). However, the intuitive visualization of these relationships is still challenging.
Challenge: The aim of this project is three-fold:
- To implement state-of-the-art dimensionality reduction techniques for massive-scale medical image datasets such as UK Biobank.
- To explore visualization techniques for massive-scale medical image datasets such as UK Biobank to allow the user to interactively query the image databases
- To implement a web-based user interface for visualization using the facilities of the visualization studio of the Data Science Institute at Imperial.
Skills: The project will require a good understanding of machine learning techniques as well as some knowledge of computer graphics and/or computer vision. The project also requires very good programming skills.
Segmentation of cardiac MR and CT images using machine learning
Background: Improvements in the quality, spatial and temporal resolution of cardiac imaging in combination with increased availability of these advanced imaging techniques to diagnose patients and plan treatments provides clinicians with a wealth of data describing patient physiology and pathology. However, the large size of these four dimensional data sets can make extracting useful information from these images challenging and time consuming. Developments in shape analysis, computational modelling and motion tracking can all support the analysis and interpretation of these images. All of these techniques require anatomical labelling of the image to provide a common reference frame for comparing and interpreting results.
Challenge: Rapid and robust anatomical labelling (or segmentation) of the four chambers of the heart remains a major challenge for cardiac image processing due to the significant variations in cardiac anatomy under pathological conditions, cardiac motion compromising image quality and the presence of image artefacts from medical devices in close proximity to the heart.
Methods: This project will aim to develop an automatic image segmentation algorithm for labelling the four chambers of the heart from MRI and cardiac CT images. In particular, the project will explore model-based segmentation algorithms based on multi-atlas segmentation. These techniques provide an initial approximation of the segmentation of the heart. This segmentation will then be refined using machine learning techniques such as Random Forests.
Application: Retrospectively gated CT data sets image the heart with an isotropic 0.4-0.5mm resolution and 80-100ms temporal resolution. Automated segmentation of the four chambers of the heart combined with motion tracking will allow the endocardial wall motion for each of the four chambers to be tracked. The relative function of each region can then be quantified in terms of contractile synchrony, ejection fraction and contraction timing, automatically reducing down the complex imaging data set to known quantitative descriptions of cardiac function that clinicians can use to inform patient selection and treatment.