Improving performance of automated coronary arterial tree center-line extraction, stent localization and tracking
Stent placement is a common procedure in Percutaneous Transluminal Coronary Angioplasty which helps many patients to avoid emergency heart bypass surgery or heart attack. In this procedure, a stent is implanted at the narrowing part of an artery to keep its lumen open allowing blood to flow normally...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2012
|
| Subjects: | |
| Online Access: | http://psasir.upm.edu.my/id/eprint/31411/1/FSKTM%202012%2028R.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Stent placement is a common procedure in Percutaneous Transluminal Coronary Angioplasty which helps many patients to avoid emergency heart bypass surgery or heart attack. In this procedure, a stent is implanted at the narrowing part of an artery to keep its lumen open allowing blood to flow normally through the artery. A potential risk of this treatment is the inaccurate placement of the stent (geographic miss), which could result in serious complications for the patient such as development of new stenotic lesions, increasing the likelihood of blood clot formation and the need for revascularization. Over the last decade, many algorithms have been developed to address this problem. However, most of them either fail to meet the demanding requirements of real-time assistant systems or to provide quantitative analysis for erification of stent placement task. In this research work, we report on new contributions to the major parts of a computer assisted stent positioning system. The first contribution is automatic detection of seed points which serve as a prerequisite step for centerline extraction algorithm. The solution consists of an algorithm for automatic collection of candidate seed points using efficient grid line searching mechanism and a validation method which uses local geometric and intensity based features as effective validation rules to discriminate between the actual seed point and false alarms. The experimental results show that combining the advantages of the geometric based validation and contrast based filtering as well as avoiding large quantization errors, lead to significant enhancement in the performance of the seed point detection algorithm in terms of balancing between the precision and recall. The second contribution is related to the robust and accurate extraction of centerlines for all vessel segments of the arterial tree in the angiogram images. This problem is addressed by proposing an accurate and robust centerline extraction method. Starting at each detected seed point, the centerline extraction method utilizes eigenvalues and eigenvectors of Hessian matrix for the pixels located on a semi-circular scanning profile for robust estimation of the next centerline point. The experimental validations show that the use of Hessian matrix results in significant improvement in the robustness of the tracing algorithm. The stent localization and tracking in angiogram image sequence is the topic of the third contribution. The proposed method combines special fast filtering, region of interest processing and graph based trajectory analysis approach to localize and track the radioopaque markers of the stent in fluoroscopic frame sequences. The most interesting finding was that the validation of the potential markers prior to building tracks of marker pairs, causes the landmark detection process to avoid dealing with a large number of outliers and misdetections. In total, the current study found that the proposed algorithms outperform their well-established existing counterparts indicating their suitability to be adopted in practical computer assisted stent positioning systems. |
|---|