Faculty of Computer Science and Information Technology >
PhD Theses : Computer Science >
Please use this identifier to cite or link to this item:
|Title: ||APPLICATIONS OF SOFT COMPUTING TECHNIQUES IN THE PREDICTION OF SURVIVAL IN CANCER|
|Authors: ||Baker, Oras F|
|Keywords: ||Soft Computing|
|Issue Date: ||2009 |
|Abstract: ||Over the last twenty years, Soft Computing has developed rapidly as a discipline and
method for the diagnosis and prognosis in Medical Informatics. Soft Computing
comprises principally of genetic algorithms, artificial neural networks, and fuzzy logics.
Soft computing techniques can be used to introduce non-linearity method for the
analysis of censored data. Mainly neural network models have been used as a soft
computing technique to analyse survival data, this realisation paved the way for this
research as one of the first to apply genetic algorithms and fuzzy logic systems to
nasopharyngeal carcinoma prognosis.
Medical prognosis is a prediction of the future course and outcome of a disease and an
indication of the likelihood of recovery from that disease. Prognosis is used as a tool to
assess the efficacy of treatment protocols, to monitor the progress of treatment
programmes and as an aid in choosing treatment types and methodologies.
The analysis is based on NPC cases seen in the UMMC; Data from 494 patients were
used to conduct this research for the purpose of finding the ideal soft computing
techniques for NPC survival rate. In addition a detailed discussion on genetic algorithms
and fuzzy logics was produced distilled from a wide variety of literature in the public
domain. A design of a genetic algorithm rule generator was discussed, and results based
on experiments with genetic algorithm rule generator were presented and analysed.
Furthermore a detailed discussion on fuzzy logics was produced. This discussion led to
the Exposition of new fuzzy logic system, which has been trained using an adaptive
neural fuzzy inference system and two different training methods were tested “The
neural networks back propagation method and the combination of least-squares and
back propagation gradient descent method (Hybrid method)”.
Nasopharyngeal carcinoma prognosis in this instance would involve a decision as to the
survival or non-survival of a patient within a time range of ten years. When a patient is
deemed to survive this is specified for at least ten years, when a patient is deemed not to
survive the time of expected death is provided to the nearest first year, up to ten years.
Our aim for this research is to develop new systems based on genetic algorithms and
fuzzy logic systems to determine efficiently NPC survival rate.
For this research two models are developed, namely:
1. A genetic algorithm that evolves algebraic rule-based classifiers for NPC
prognosis and a selection of subset of prognostic factors with high prognostic
2. A fuzzy logic system trained with ANFIS “adaptive neural fuzzy inference
system” with two different training strategies (The neural networks backpropagation
method and the combination of least squares and back-propagation
gradient descent method “hybrid method”) and both can successfully predict the
survival rate for a NPC patient.
The study done in this thesis was designed as a focused preliminary exploration into the
power of the genetic algorithm and fuzzy logic system to the prediction of cancer
survival, The results of the experiments were very positive, comparing the outcome of
the GA model with that of FL it shows the robustness of the GA model as prediction
The two principal designs indicate that the use of genetic algorithms and fuzzy logic in
NPC is definitely a fruitful endeavour. The results would suggest that genetic
algorithms as standalone classifier models are better (based on the system designed in
this research) for this sort of task than a fuzzy logic model.
In conclusion it is the main thesis of this research that new approaches to
nasopharyngeal carcinoma prognosis is presented, genetic algorithms and fuzzy logic
systems are indeed viable and useful approaches and both shows high accuracy. This is
where statistical regression and artificial neural networks have dominated the field of
cancer prognosis. From the results obtained in this thesis we have shown that it is
possible for researchers to undertake projects involving medical informatics using the
local scenario. We hope that this study could generate more efforts and researches, and
encourage more Malaysians to embark on researches involving the applications of
genetic algorithms and fuzzy logics in various types of medical applications.|
|Appears in Collections:||PhD Theses : Computer Science|