A case-based reasoning system for radiotherapy treatment planning for brain cancer.
PhD thesis, University of Nottingham.
In this thesis, a novel case-based reasoning (CBR) approach to radiotherapy treatment planning for brain cancer patients is presented. In radiotherapy, tumour cells are destroyed using ionizing radiation. For each patient, a treatment plan is generated that describes how the radiation should be applied in order to deliver a tumouricidal radiation dose while avoiding irradiation of healthy tissue and organs at risk in the vicinity of the tumour. The traditional, manual trial and error approach is a time-consuming process that depends on the experience and intuitive knowledge of medical physicists. CBR is an artificial intelligence methodology, which attempts to solve new problems based on the solutions of previously solved similar problems. In this research work, CBR is used to generate the parameters of a treatment plan by capturing the subjective and intuitive knowledge of expert medical physicists stored intrinsically in the treatment plans of similar patients treated in the past.
This work focusses on the retrieval stage of the CBR system, in which given a new patient case, the most similar case in the archived case base is retrieved along with its treatment plan. A number of research issues that arise from using CBR for radiotherapy treatment planning for brain cancer are addressed. Different approaches to similarity calculation between cases are investigated and compared, in particular, the weighted nearest neighbour similarity measure and a novel non-linear, fuzzy similarity measure designed for our CBR system. A local case attribute weighting scheme has been developed that uses rules to assign attribute weights based on the values of the attributes in the new case and is compared to global attribute weighting, where the attribute weights remain constant for all target cases. A multi-phase case retrieval approach is introduced in which each phase considers one part of the solution. In addition, a framework developed for the imputation of missing values in the case base is described.
The research was carried out in collaboration with medical physicists at the Nottingham University Hospitals NHS Trust, City Hospital Campus, UK. The performance of the developed methodologies was tested using brain cancer patient cases obtained from the City Hospital. The results obtained show that the success rate of the retrieval mechanism provides a good starting point for adaptation, the next phase in development for the CBR system. The developed automated CBR system will assist medical physicists in quickly generating treatment plans and can also serve as a teaching and training aid for junior, inexperienced medical physicists. In addition, the developed methods are generic in nature and can be adapted to be used in other CBR or intelligent decision support systems for other complex, real world, problem domains that highly depend on subjective and intuitive knowledge.
Thesis (University of Nottingham only)
||radiotherapy, cased-based reasoning, CBR, intelligent decision support systems
||Q Science > Q Science (General)
||UK Campuses > Faculty of Science > School of Computer Science
Hatton, Mrs Kirsty
||13 Jul 2015 07:57
||15 Sep 2016 16:08
Actions (Archive Staff Only)