Differenzierung verschiedener Lungenmorphologien in einem experimentellen Maus-Asthmamodell mittels automatischer Bildanalyse

In dieser Arbeit wurde der Versuch einer Unterscheidung von Lungenmorphologien in Asthmafragestellungen mittels automatischer Bildanalysesoftware vorgenommen. Die zu untersuchenden Parameter waren das Volumen, die Masse und die Dichte der Lunge. Diese Parameter wurden an zuvor in einer andere...

Full description

Saved in:
Bibliographic Details
Main Author: Inan, Selcuk
Contributors: Renz, Harald (Prof. Dr.) (Thesis advisor)
Format: Dissertation
Language:German
Published: Philipps-Universität Marburg 2015
Subjects:
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents: This thesis was an attempt to differentiate lung morphology in an asthma mouse model using automatic image-analysis software. The parameters to be examined were volume, mass and the density of the lung. These parameters were evaluated from data, obtained in a previously performed study of an animal experiment. In that study, twelve mice were divided equally into two groups. The first group consisted of mice in the control group that received only a sham treatment. The second group was the asthma group, in which asthma was induced using a well-established acute mouse model. After an initial imaging of both groups, in all mice an artificial asthma attack was induced, which was triggered by methacholine and captured by imaging. The aim was to determine if it is possible to characterize, on the basis of the mentioned three parameters, a change in the lung morphology of the different experimental groups. The first step was the visual evaluation of the obtained images in terms of their visual differences. Therefore, the images of all groups were compared on a workstation with each other. The main idea behind the software development was that the software can only find differences, if differences are detectable in the images. Afterwards, the statistical analysis of the parameters volume, mass, and density were evaluated. The result of statistical analysis showed, that, based on the parameter volume, only four of the six comparison groups could be distinguished from each other by means of sufficiently small p-values. The combination of control group before methacholine treatment vs. asthma group before methacholine treatment and control group after methacholine treatment vs. asthma group after methacholine treatment could not be discriminated from each other based on the p-values. This circumstance can be explained by the fact that the volume of each phenotype is an individual value, which may be very different for each mouse. For the differentiation of the various groups on the basis of the parameter “mass”, no statistically significant differences could be found. This parameter also depends strongly on the phenotype of each mouse, and may highly vary individually, so that it was not possible to define a reasonable parameter based on the “mass”. By the definition of the third parameter, the density, individual differences were normalized. Hence, five of the six pairs could be distinguished from each other by the means of p-value. Only the pair of control group after methacholine treatment vs. asthma group after methacholine treatment was not discriminable. This was due to the effect of meth-acholine, which let the images appear visually identical after its application. The experiments of the present study were performed in 2012 and were completed before the pub- lication of Changani et al. (2013). They were performed completely independently from the publication in the literature. Conclusion: a classification can be successfully performed based on the density parameter. Therefore, an automatic, numerically based data-analysis is possible without subjective influences. The results in terms of classification between different experimental groups in this thesis, are identical to the classification method based of the fractal dimension analysis published by Obert and Coworkers (2015) 5 . Thus, a classification can be made computationally easier using the density evaluation and is therefore preferra- ble. Outlook: an application and evaluation of the software to human data would be the next interesting step, to assist the radiologist at work.