Medical image segmentation by hybridizing multi-agent system and reinforcement learning agent

Mahsa , Chitsaz (2009) Medical image segmentation by hybridizing multi-agent system and reinforcement learning agent. Masters thesis, University of Malaya.

[img] PDF
Dissertation_Mahsa Chitsaz.pdf

Download (4MB)

Abstract

Image segmentation is still a debatable problem although there have been many research work done in the last few decades. First of all, every solution for image segmentation is problem-based. Secondly, medical image segmentation methods generally have restrictions because medical images have very similar gray level and texture among the interested objects. Therefore, this dissertation presents a framework to extract simultaneity several objects of interest from head Computed Tomography images. The proposed method contains two phases; training and testing. A Reinforcement-Learning method is proposed for the training phase, and a new Multi-Agent system is proposed for the testing phase. In the training phase, a few images are used as a trained image whereas the RL agent will find the appropriate value of each object or region in the input image. The outcome of this training phase is transferred to the next phase, testing phase. In this phase, the images are segmented by some priori knowledge and the properties of local agent. Proposed reinforcement learning model attains significant result in segmentation accuracy; the accuracy is more than 95% for each region in the image and the mean computation time of all datasets is less than 13 seconds. Moreover, the number of training data set for PRLM can be one or a small number of images. Also, PRLM has the ability to segment simultaneously an image into some distinct regions. Proposed multi-agent model attains considerable result in segmentation accuracy; the accuracy is more than 90% for each region in the image and the mean computation time of all datasets is less than 7 seconds. Furthermore, PMAM is capable to segment simultaneously an image into some distinct regions.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Medical image segmentation; Hybridizing; Multi-agent system; Learning agent; Computed tomography images
Subjects: Z Bibliography. Library Science. Information Resources > Z665 Library Science. Information Science
Depositing User: MS NOOR ZAKIRA ZULRIMI
Date Deposited: 16 Jul 2013 08:20
Last Modified: 16 Jul 2013 08:20
URI: http://repository.um.edu.my/id/eprint/453

Actions (login required)

View Item View Item