Disease identification is one of the most important aspects of a physician's duties. Radiologists play a very important role in disease recognition based on the increasing use of diagnostic images. Nowadays, medical scanning devices such as MRI and CT produce thousands of images per patient which makes the radiologist's job even more onerous; indeed radiologists look at approximately 50,000 images per day leading to fatigue and a higher probability for missing smaller lesions. Therefore it is critical to assist radiologists in their duties.In this PhD work, research is focused on developing segmentation-based mathematical algorithms and computer programs for automatically characterizing lung CT images. There are two kinds of segmentation methods; the first group contains the methods that find edges of all objects in the image and the second group contains the methods that focus on one object in the image. By assessing many segmentation methods and based on the concept of this project, the level set method, from the second group, has the capability to accurately find the boundary of an object in medical images. Although this method does not need any threshold for segmenting an object in an image, it does require the setting of seven parameters. Genetic algorithms were employed to optimize seven parameters of the level set method for use as a boundary detection method. A streamlined automatic mechanism, essential for successful and fast segmenting processes, provided the level set method with a good initial contour.This segmentation step was fundamental for further measurements such as bronchial lumen diameter and wall thickness measurement. The developed program automatically measures airway lumen diameter with exceptional repeatability. Also, by simulating manual methods used by radiologists for measuring luminal wall thickness, the automated fitting method consistently finds the wall thickness at the thinnest part, minimizing partial volume problems. The current standard for measuring luminal wall thickness is the full-width at half-maximum method. The technique formulated here is more accurate and reproducible and can be performed automatically. To find the lumen airway tapering, in order to recognize some prominent lung diseases, a method for tracing of an airway through various CT slices was developed. These measurements are of critical importance in the understanding of a number of lung diseases including asthma and COPD; this tracing step was fundamental for bronchial bifurcation angle measurement. This tracing method was extended to detect and follow bifurcation branches. Then, an estimation method for finding a fitted line through airway center points was developed. For the last portion of this PhD work an approach for measuring lung airway bifurcation angle from CT datasets, which is important in lung diseases such as asthma, was created. The current goal, characterizing lung CT images, was achieved as a working form of this software development. This mathematical approach has been shown to be accurate with phantom studies. The ultimate goal of this work was to develop software for recognizing key lung diseases and to compare serially (weeks or months apart) acquired images to assess any progress, regression, or stability in a disease. / Thesis / Doctor of Philosophy (PhD)
| Identifer | oai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/17317 |
| Date | 09 1900 |
| Creators | Heydarian Firouz Abadi, Mohammadreza |
| Contributors | Poehlman, Skip, Kamath, Markad, Computer Science |
| Source Sets | McMaster University |
| Language | English |
| Detected Language | English |
Page generated in 0.0017 seconds