Segmentation of Blood Vessels Based on a Threshold that Combines Statistical and Scale Space Filters
This paper presents a strategy for segmenting blood vessels based on the threshold, which-combines statistic and scale space filter. By incorporating statistical information, the strategy is capable of reducing over-segmentation. We propose a three stage strategy which involves: (1) optimal selection of window size; (2) optimal selection of scale and (3) segmentation process. We compared our strategy to two commonly used thresholding techniques. Experimental results showed that our method is much more robust and accurate. Our strategy suggested a modification to Otsu's method. In this application the important information to be extracted from images is only the number of blood vessels present in the images. The proposed strategy was tested on manual segmentation, where segmentation errors less than 3% for false positives and 0% for false negatives are observed. The work presented in this paper is a part of a global image analysis process. Therefore, these images will be subject to a further morphometrical analysis in order to diagnose and predict automatically malign tumors.
Main Author: | |
---|---|
Format: | Digital revista |
Language: | English |
Published: |
Instituto Politécnico Nacional, Centro de Investigación en Computación
2005
|
Online Access: | http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S1405-55462005000400002 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | This paper presents a strategy for segmenting blood vessels based on the threshold, which-combines statistic and scale space filter. By incorporating statistical information, the strategy is capable of reducing over-segmentation. We propose a three stage strategy which involves: (1) optimal selection of window size; (2) optimal selection of scale and (3) segmentation process. We compared our strategy to two commonly used thresholding techniques. Experimental results showed that our method is much more robust and accurate. Our strategy suggested a modification to Otsu's method. In this application the important information to be extracted from images is only the number of blood vessels present in the images. The proposed strategy was tested on manual segmentation, where segmentation errors less than 3% for false positives and 0% for false negatives are observed. The work presented in this paper is a part of a global image analysis process. Therefore, these images will be subject to a further morphometrical analysis in order to diagnose and predict automatically malign tumors. |
---|