Vessel-detection in chest X-ray images

OData support
Dr. Horváth Gábor
Department of Measurement and Information Systems

In my thesis I worked with chest radiographs. On these images nodules may be seen. For the detection of these nodules automatic nodule detection algorithms were developed. These algorithms are able to detect as many or even more nodules than specialists. The problem of these algorithms that is they produce too many false positives. My main task was to decrease the number of these false positives without losing too many true positives.

For this task I searched for and examined an algorithm which can detect vessel opacities on a chest radiograph. I implemented the chosen the algorithm in Matlab. Then I tried to use the algorithm to improve the accuracy of a nodule detection algorithm but the results were inadequate. Thus I developed methods for improving the algorithm. The results of the new algorithm surpassed the original one's results. This means that it is possible to detect approximately the 26% of the false positives while losing the 10% of the true positives. In the end I proposed methods for further improvement.


Please sign in to download the files of this thesis.