About the Project
For a full project proposal and details on how to apply using our online recruitment portal please see icr.ac.uk/phds. Please note we only accept applications via the online application system apply.icr.ac.uk
The incidence and prevalence of brain metastases (BM) has increased over the last few years as advances in cancer treatments have led to better extracranial disease control and survival. Survival with BM has also improved significantly from a few months to 1 or more years with better understanding and treatment of BM. All patients with BM are considered at increased risk of seizures and are warned of this when diagnosed, and all are immediately banned from driving. Patients report significant detrimental consequences from this including increased social isolation, anxiety, loss of confidence and inability to work. However, we know that the majority will never have a seizure, and many will live the remainder of their lives in unnecessary fear. Furthermore, screening for BM is now undertaken more frequently in asymptomatic patients and smaller sites of disease detectable with newer MRI methods.
Stereotactic radiosurgery (SRS) is a highly-targeted form of radiotherapy used to treat BM. Treatment response is assessed with 3-monthly MRI scans but interpretation is challenging as SRS-related brain injury, ‘radionecrosis’, in responding patients can mimic disease progression and confound radiological assessment. Treatment of progressive disease is drastically different from treatment of radionecrosis and better response assessment is urgently needed to allow accurate MRI interpretation without needing to await serial scan results. The SAFER study (CCR:5266, CI:Nicola Rosenfelder), aims to determine which clinical- and MRI-features increase seizure-risk post-SRS using 3-monthly MRIs.
In this proposed project, the student will develop novel approaches using artificial intelligence and deep-learning to improve seizure-risk-stratification and response evaluation post-SRS for BM by combining clinically-relevant data sources including MR-imaging, patient demographics, and radiotherapy dose. These new algorithms will be tested and verified within the context of the SAFER study. Technical outcomes of this project will be highly applicable to cancer research utilising imaging data and radiotherapy.
- Develop new multi-parametric deep-learning (DL) algorithms that combine MR-imaging data with patient demographics and radiation dose to predict risk and response in cancer.
- Provide ‘explainability’ to the proposed algorithms and identify the importance of each of the data sources in predicting patient outcome.
- Test the accuracy of proposed algorithms in predicting the patient-specific risk of (1) Seizure events due to presence of brain metastases (2) Tumour response from patient survival (3) Radionecrosis versus progressive disease
- Compare and contrast such deep-learning approaches with traditional statistical approaches including radiomics.
- Collate developed tools into robust research software that can be used for analysis of clinical trial data by radiologists.
Keywords/ Subject Areas:
- Medical Physics
- Artificial Intelligence
- Brain Metastasis
- Machine Learning
- Magnetic Resonance Imaging
- Computer Vision
Based on your current searches we recommend the following search filters.
Based on your current search criteria we thought you might be interested in these.