Prediction of individual patient survival and GBM tumour characteristics
Glioblastoma multiforme (GBM) is the most common and aggressive malignant brain tumour with average survival of just over one year. Biopsy of tumours by surgery is the gold standard for classification of histology and molecular characteristics but is only a small sample and may therefore not be accurate all the time. Routinely acquired MR imaging of the brain allows diagnosis and planning for surgery & radiotherapy. A new way of analysing tumour volume with certain MRI characteristics, termed Texture Analysis (TA), can allow prediction of patient survival, tumour grade and molecular status, but is not routinely done.
The aims of this PhD studentship are a) to develop new computer coding for TA, incorporating the best existing methods and b) combine it with ‘machine learning’ (semi-automated computer analysis of MRI scans learnt from previous patients’ imaging) to more accurately predict patient survival and tumour characteristics. Furthermore, GBM tumour growth models will be developed and fitted to NHS patients’ MRI scan data, to c) determine whether this added information (collected over time to monitor tumour growth) enhances the accuracy of our predictions – i.e. aiming for clinically useful information unique to each individual patient.