Steerable3D: An ImageJ plugin for neurovascular enhancement in 3-D segmentation

Abstract

Abstract Purpose Image processing plays a fundamental role in the study of central nervous system, for example in the analysis of the vascular network in neurodegenerative diseases. Synchrotron X-ray Phase-contrast micro-Tomography (SXPCT) is a very attractive method to study weakly absorbing samples and features, such as the vascular network in the spinal cord (SC). However, the identification and segmentation of vascular structures in SXPCT images is seriously hampered by the presence of image noise and strong contrast inhomogeneities, due to the sensitivity of the technique to small electronic density variations. In order to help with these tasks, we implemented a user-friendly ImageJ plugin based on a 3D Gaussian steerable filter, tuned up for the enhancement of tubular structures in SXPCT images. Methods The developed 3D Gaussian steerable filter plugin for ImageJ is based on the steerability properties of Gaussian derivatives. We applied it to SXPCT images of ex-vivo mouse SCs acquired at different experimental conditions. Results The filter response shows a strong amplification of the source image contrast-to-background ratio (CBR), independently of structures orientation. We found that after the filter application, the CBR ratio increases by a factor ranging from ~6 to ~60. In addition, we also observed an increase of 35% of the contrast to noise ratio in the case of injured mouse SC. Conclusion The developed tool can generally facilitate the detection/segmentation of capillaries, veins and arteries that were not clearly observable in non-filtered SXPCT images. Its systematic application could allow obtaining quantitative information from pre-clinical and clinical images.