Congratulations to Jonathan Goh for successfully passing his MPhil to PhD transfer
Monday 17 November 2008
Jonathan successfully transferred on Friday 14th November with his work titled "Semantic Analysis and its Scalability in Medical Images". Jonathan is supervised by Dr Lilian Tang and Dr Tony Browne, and was examined by Dr Jonathan Clark and Dr Bogdan Vrusias.
Abstract:
This research is interested in investigating how to develop a framework based on multiple classifiers and contextual reasoning that is robust and reliable over large-scale data. Using retinal images as the application domain, the use of multiple classifiers will be examined to determine whether they can improve the overall performance in the detection of various symptoms in retina images exhibited in Diabetic Retinopathy and ultimately asses its scalability over a wide variety of images. A number of preliminary experiments have shown that multiple classifiers achieves better accuracy compared with using the best performing classifiers. In order to differentiate between visually similar symptoms, a model is built incorporating Bayesian network for contextual reasoning to improve analysis accuracy. In this work, the use of Bayesian network for detecting micro aneurysms is experimented and the results have shown that by using this technique, false positives is decreased by at least 50%. The future wok for my PhD dissertation will focus on the development of multiple classifier combination strategies and contextual reasoning models in order to address the scalability of semantic analysis in a dynamic changing large image medical archive.