Fully automated side branch detection in intravascular optical coherence tomography pullback runs

Abstract

Side branches in the atherosclerotic lesion region are important as they highly influence the treatment strategy selection and optimization. Moreover, they are reliable landmarks for image registration. By providing high resolution delineation of coronary morphology, intravascular optical coherence tomography (IVOCT) has been increasingly used for side branch analysis. This paper presents a fully automated method to detect side branches in IVOCT images, which relies on precise segmentation of the imaging catheter, the protective sheath, the guide wire and the lumen. 25 invivo data sets were used for validation. The intraclass correlation coefficient between the algorithmic results and manual delineations for the imaging catheter, the protective sheath and the lumen contour positions was 0.997, 0.949 and 0.974, respectively. All the guide wires were detected correctly and the Dice's coefficient of the shadow regions behind the guide wire was 0.97. 94.0% of 82 side branches were detected with 5.0% false positives and the Dice's coefficient of the side branch size was 0.85. In conclusion, the presented method has been demonstrated to be accurate and robust for side branch analysis.

Automatic segmentation of coronary morphology using transmittance-based lumen intensity-enhanced intravascular optical coherence tomography images and applying a localized level-set-based active contour method

“…Therefore, a direct segmentation of the lumen was now possible from display format images, without the need for sequential removal of the catheter and its protective sheath. 4 To obtain lumen intensity-enhanced IV-OCT images, first transmissivity maps were generated with n ¼ 6 value, from every polar domain IV-OCT image. Then a log-scale compression was applied to the transmissivity maps to reduce the dynamic range which was both inherently present and resulted from the exponentiation.…”

“…10 In such scenarios, lumen segmentation methods cannot guarantee an accurate lumen contour for the affected regions. Corrupted lumen contours can affect the lumen quantification methods such as geometrical analysis 3,4 of the contour to identify bifurcation points of the side branches. To overcome this, a scheme is presented wherein low-intensity lumen regions of the side branches were identified and subsequently the erroneous contour sections were masked out.…”

“…The degree of agreement between any two segmentations, i.e., O1 versus O2, O1 versus At, and O2 versus At (where At refers to the automatic method), was carried out using an intraclass coefficient (ICC) method. 4 For any two segmentations, the ICC method compared the radial distance (measured in pixels) of the lumen contour points from the image center. The radial distance error (i.e., differences in the measured radial distance) between any two approaches was also performed.…”

“…1 Automated assessment can reduce this error and produce more consistent image interpretation. Various techniques have been reported for lumen segmentation, [1][2][3][4][5][6][7][8] stent strut detection, 7 and plaque characterization. 1,9 These approaches are either semi or fully automated.…”

“…Thus, the wall-lumen interface provides a high degree of contrast and this is generally exploited by various techniques to extract the lumen boundary. Some of the methods previously employed include image thresholding, 1 A-line intensity-variation analysis, 2,3 intensity difference-dependent cost function and its minimization approach, 4,5 and labeling methods using Markov random field (MRF). 7 Alternatively, a combination approach with expectation maximization (EM) for labeling and graph-cut for lumen segmentation has been reported.…”