Support value based stent-graft marker detection

• Published in: Pattern recognition Vol. 46; no. 3; pp. 962 - 975
• Main Authors: Zheng, Sheng, Yang, Changcai, Kaptein, Bart L., Hendriks, Emile A., Koning, Olivier H.J., Lei, Bangjun
• Format: Journal Article
• Language: English
• Published: Kidlington Elsevier Ltd 01-03-2013; Elsevier
• Subjects: Applied sciences; Artificial intelligence; Biological and medical sciences; Computer science; control theory; systems; Correlation analysis; Detectors; Directional support value transform (SVT); Diseases of the cardiovascular system; Exact sciences and technology; Fluoroscopic roentgenographic stereophotogrammetric analysis (FRSA); Investigative techniques, diagnostic techniques (general aspects); Markers; Medical sciences; Multiscale correlations; Noise; Pattern recognition. Digital image processing. Computational geometry; Radiodiagnosis. Nmr imagery. Nmr spectrometry; Radiotherapy. Instrumental treatment. Physiotherapy. Reeducation. Rehabilitation, orthophony, crenotherapy. Diet therapy and various other treatments (general aspects); Spots; Stent-graft markers detection; Support value matrix; Surgical implants; Three dimensional; Support value matrix; Directional support value transform (SVT); Multiscale correlations; Fluoroscopic roentgenographic stereophotogrammetric analysis (FRSA); Stent-graft markers detection; Correlation; Stent; Variance; stereophotogrammetric analysis (FRSA); Vector support machine; Stereopsis; Photogrammetry; Stereophotogrammetry; Computer vision; X ray radiography; Measurement sensor; Characteristic curve; Marker; Experimental study; Poisson distribution; Mathematical morphology; Gray scale; Wavelet transformation; Fluoroscopic roentgenographic; Filter; Automatic measurement; Multiscale method
• ISSN: 0031-3203; 1873-5142
• DOI: 10.1016/j.patcog.2012.08.017

With the development of the fluoroscopic roentgenographic stereophotogrammetric analysis (FRSA), it is possible to make the three-dimensional (3D) dynamics of stent-graft. The stent-graft markers, however, are identified manually. In this paper we present a robust solution for automatic detection of stent-graft marker projections in FRSA X-ray images. Several directional support value (dSV) filters and the directional support value transform (dSVT) method are studied. Based on the dSV of the dSVT, a support value matrix is constructed, and the determinant of this matrix is then defined as the markerness measure. The corresponding multi-scale correlations of the rescaled markerness measures are computed for enhancing the multi-scale marker response peaks while suppressing the effects of stent-grafts and Poisson noise. The marker spots are subsequently located by finding the local maximum of the correlated markerness measures. The conditional variance Stabilizer (CVS) is further integrated into this framework for removing Poisson noises. Performance comparisons are carried out among the proposed dSVT, the CVS+dSVT, local threshold operation (LTO) and the frequently adopted spot detectors, including the morphological grayscale opening top-hat filter (MTH), wavelet multiscale products (WMP), and multiscale variance-stabilizing transform (MSVST) methods. The results from experiments on synthetic as well as real FRSA X-ray image data show that the proposed CVS+dSVT method performs better than other detectors, in terms of the free-response receiver operation characteristic (FROC) curves.
► We developed the directional support value analysis method (dSVT). ► Based on the dSVT, a support value matrix is constructed. ► Then the determinant of this matrix is defined as the markerness measure. ► The multi-scale measures are computed for enhancing the marker response peaks. ► The CVS is further integrated into this framework for removing Poisson noises.