User guided biomedical image segmentation and usable interfaces

Download files
Access & Terms of Use
open access
Copyright: Hosseini, Mohammadreza
Altmetric
Abstract
The ability to measure objects boundary and dimensions in medical and biomedical images and to assure a safe interaction with these image in the experimental laboratory using computer vision technology are the principal goals of this thesis. The accurate estimation of the dimensions of bacterium in the biomedical images has a significant impact on observing the reactions to a new environment. Accurate measurement of organs boundary in medical images has a great impact on designing reliable surgical plan. The safe interaction with usable interfaces will decrease any chance of bacterial contamination. In a high-resolution microscopic image, where the growth rate of bacteria may be less than 1/10 of a pixel width, more accurate estimation of object dimensions is required. Segmentation methods that rely on image intensity for segmentation tend to overestimate the bacterium dimensions. Instead, removing the blurring effect will result in more accurate dimension estimation. In this thesis, a novel approach to bacterium segmentation and measurement based on a new bacterium model is presented. The model can be used to estimate the bacterium dimensions and orientation with high accuracy. For medical images of human organs, on the other hand, automatic segmentation methods may not produce expected boundaries due to non-rigidity of organs. Therefore, interactive segmentation approaches based on expert feedback are becoming popular. These are often based a single feature of the feedback. In this thesis, the use of multiple features of user feedback in defining the energy function in energy-based segmentation methods is evaluated. Experiments on lung MDCT (multi-row detector computed tomography) images show that the proposed method performs better than a comparative automatic and single feature interactive segmentation methods. Interfaces that rely on human gestures to control a machine is a promising approach for increasing biological laboratory safety. An important issue is the usability of these new interfaces, which has not been thoroughly studied. Usability study of touchless and touch-based interfaces for interaction with biomedical images is conducted in this thesis. The knowledge gained based on user experience is then used to design a new interface more aligned with user needs.
Persistent link to this record
Link to Publisher Version
Link to Open Access Version
Additional Link
Author(s)
Hosseini, Mohammadreza
Supervisor(s)
Sowmya, Arcot
Bednarz, Tomasz
Creator(s)
Editor(s)
Translator(s)
Curator(s)
Designer(s)
Arranger(s)
Composer(s)
Recordist(s)
Conference Proceedings Editor(s)
Other Contributor(s)
Corporate/Industry Contributor(s)
Publication Year
2016
Resource Type
Thesis
Degree Type
PhD Doctorate
UNSW Faculty
Files
download public version.pdf 5.42 MB Adobe Portable Document Format
Related dataset(s)