Background Paediatric brain tumours are the second most common paediatric malignancy yet remain the leading cause of childhood cancer related deaths. This is despite advances in both surgical techniques and adjuvant oncological therapies. Survival prediction for this heterogenous group of tumours is challenging, due to the large number of predictive variables and relatively small numbers of cases. Consequently, it is difficult to provide patients, their parents and carers with accurate, individual prognosis and survival predictions. There is therefore a currently unmet need to harness full patient data and generate outcome predictions on an individual, personalised basis. Furthermore, the addition of advanced metric MRI sequences to routine paediatric brain tumour imaging generates a wealth of data about the biological behaviour of tumours that is not currently integrated into such outcome prediction. Aims To trial machine learning methodology in the survival prediction for paediatric brain tumours, utilising acknowledged prognostic makers plus advanced metric MRI sequences. Main results We have demonstrated that machine learning algorithms are capable of handling the complex, feature rich high-dimensional data set that accompanies paediatric brain tumour patientâs imaging. In addition, we have shown the importance of combining acknowledged prognostic markers in prediction modelling, and that that the addition of molecular and advanced MRI data creates the strongest prediction models. This preliminary work validates the need for further evaluation of our CoxNet L1 penalised model with larger, complete and unseen data sets to assess predictive capability. We have also identified a novel radiological phenotype of two rare paediatric brain tumours, with potential impact on both surgical management strategies and survival. Conclusions We describe an exemplar work relating to the collation and integration of whole patient data, with the aim of generating personalised survival outcome prediction in paediatric brain tumours.
Date of Award | 6 Jan 2025 |
---|
Original language | English |
---|
Awarding Institution | - The University of Manchester
|
---|
Supervisor | Andrew Brass (Supervisor), Ian Kamaly-Asl (Supervisor), Stavros Stivaros (Supervisor) & John-Paul Kilday (Supervisor) |
---|
- paediatric brain tumour
- paediatric neuro oncology
- paediatric neuro radiology
- artificial intelligence
- machine learning
Diagnosis, risk stratification and outcome prediction in paediatric central nervous system tumours; current techniques and the potential role for the use of advanced multiparametric resonance imaging techniques through the application of machine learning artificial intelligence systems.
Pringle, C. (Author). 6 Jan 2025
Student thesis: Phd