Background:
In-vivo models play an essential role in pre-clinical studies for research into infectious diseases, and lead directly to the development of vaccines and therapeutic agents as well as playing a vital role in the study of the migrational dynamics and differentiation processes of stem cells. Biophotonics-based imaging systems, especially Bioluminescence Imaging are highly sensitive and non-invasive techniques based on the detection of visible and near-infrared light produced by luciferase-catalysed reactions (bioluminescence) or by excitation of exogenous (fluorescent) molecules. These methods allow for the non-invasive detection and visualisation in 3D of functional activity within intact living models [1,2].
Deep learning has been actively investigated for various applications such as image classification, computer vision, and regression tasks, and has shown state-of-the-art performance. In diffuse optical tomography (DOT), and Bioluminescence Tomography (BLT) the accurate estimation of the bulk optical properties and light emitting bioluminescence sources within a medium is important as it directly affects the overall image quality, and derived cellular activity [3]. In this work, we aim to exploit deep learning to develop a set of a Deep Learning based image reconstruction algorithms to not only estimate the optical properties of a medium, but also the size and location of the light emitting sources within.
Aims:
We propose to develop, evaluate and validate a set of Deep Learning (DL) based image reconstruction algorithms that will provide quantitatively accurate information about not only luminescence activity in-vivo but also the often assumed and unknown optical attenuation properties of the tissue being imaged. The aim would be then to optimize and learn non-linear spatial transformations between measured and modelled data through the use of non-linear computational models of light propagation in tissue [4], providing a 3D map of cellular activity and location. This will be a significant advancement from most existing deep learning based image recovery methods that it will learn non-linear spatial transformations from training data with known corresponding spatial transformations.
Important milestones in this project are:
- Development and validation of DL based tomographic algorithms for BLT based on assumed optical attenuation of tissue.
- Development and validation of DL based tomographic algorithms for DOT and BLT based on multi-spectral data
- Testing and evaluation using phantom and in-vivo models to better help and guide radiotherapy treatment of cancer.
The student will gain experience in:
- Computing and data analysis (integrating multiple tomographic and 3D datasets, Data reduction)
- Biomedical sciences
- Physical sciences (Understanding and learning to utilize equipment designed to measure biological variables through differential absorption/emission of light spectra using tomographic near infrared spectroscopy.
Key Skills
Applicants should have a very good BSc (Honours) (First or Upper Second class) degree or a Master degree (with Distinction or Merit) in either Computing Science, Physics or related discipline.
Essential Knowledge and Experience:
- Programming experience (preferably some in Matlab and Python and some basic image processing techniques)
- Strong communication skills (including written/spoken English)
- Some experience in numerical modeling, particularly Finite Element models
- Solid skills in maths/statistics
- An ability to think independently and critically analyse different sources
Desirable requirements
- Knowledge of deep learning techniques/packages (e.g. Keras, TensorFlow,)
- Experience in the development, application and deployment of signal processing techniques
Applicants should have good personal and communication skills, strong professionalism and integrity, and be capable of working on their own initiative.
Continue with Facebook
