Automatic segmentation method for bone and blood vessel in murine hindlimb.

PURPOSE

The goal of this paper is to address three problems existing in vessel extraction of murine hindlimb. First, the bone can hardly be separated from blood vessels because the intensity of contrast enhanced blood vessels is similar to that of bones. Second, as an automatic blood vessel segmentation method, the vesselness method is sensitive to sharp boundaries, resulting in false positive effect in nonvascular regions. Finally, thin blood vessels are always broken after segmentation because of the low signal-to-noise ratio.

METHODS

The proposed automatic segmentation method for bone and blood vessel in this paper includes three important modules. (1) To eliminate the interference of bones on the segmentation of blood vessels, the authors employ split Bregman method to segment bones in the first place. (2) The authors propose an edge extension strategy to cope with the false positive effect of the vesselness method on the sharp boundaries of hindlimb after the removal of bones. Then, the authors segment the blood vessels using the vesselness method combined with multiscale bi-Gaussian filtering. (3) The authors reconnect the broken blood vessels after segmentation based on centerline and morphological dilation.

RESULTS

The bones' segmentation from the murine hindlimbs was conducted using the split Bregman, manual, and thresholding methods, respectively. Compared with the thresholding method, the split Bregman method could finely segment the bones from blood vessels, and the results were comparable to that of manual segmentation. After removing bones, the vesselness method combined with the bi-Gaussian filtering with and without edge extension was performed. The vesselness results with the edge extension strategy could effectively eliminate the false positive effect on sharp boundaries in nonvascular regions. Some of the blood vessels segmented by thresholding from the vesselness results were disconnected. Thus, the authors employed the vascular connection method based on centerline and morphological dilation to connect the broken blood vessels. Compared with the vascular connection utilizing the spatial-variant and -invariant morphological closing methods, the proposed vascular connection method reconnected the broken blood vessels and meanwhile maintained the nonbroken ones unchanged.

CONCLUSIONS

Our proposed method is suitable for the segmentation of bones and blood vessels in murine hindlimbs. For the segmentation of bones, the split Bregman method improves the distinguishability between bones and blood vessels, since both the intensity information and the geometrical size are exploited. For the segmentation of blood vessels, vesselness method with the edge extension strategy eliminates the false positive effect on the nonvascular sharp boundaries. After segmentation, the proposed vascular connection method based on centerline and morphological dilation can reconnect the broken blood vessels without affecting the nonbroken ones.