Three-dimensional quantitative analysis of bone microvasculature in synchrotron micro-CT imaging
- Authors
- Publication Date
- Mar 26, 2021
- Source
- HAL-Descartes
- Keywords
- Language
- English
- License
- Unknown
- External links
Abstract
Breast cancer is the most frequently diagnosed cancer in women worldwide. Breast cancer bone metastases cannot only bring about bone destructions but also facilitate the formation of undesirable vascularization. Angiogenesis is the formation of new blood vessels and can be a rate-limiting step in the growth of metastatic bone tumors. It is necessary to develop efficient methods to image and analyze three-dimensional (3D) bone and vessels. Synchrotron radiation microcomputed tomography (SR-μCT) has a significant advantage of yielding high spatial resolution images with a high signal-to-noise ratio. SR-µCT has been applied to the quantitative analysis of bone up to the micrometer scale. In addition, SR-µCT coupled with a contrast agent permitted to visualize simultaneously the 3D bone microstructures and vascular networks in preclinical studies. Segmentation is an important step in image analysis. Previously, the bone and vessels of rats, simultaneously imaged using SR-µCT with a contrast agent, have been segmented automatically using 3D region growing method. To make more models of pathologies available, it would be desirable to image mice. However, this transition is not straightforward. As opposed to in rat bone, vessels may appear to be in contact with the bone surface in mice. This precludes the correct segmentations of bone and vessels using 3D region growing. In this thesis, we proposed an algorithm based on marker-controlled watershed in conjunction with the monogenic signal phase asymmetry. To evaluate the accuracy and robustness of the proposed method, a series of synthetic volumes were generated to mimic the real vessel, bone and background structures. Different contrasts between various compartments, different noise levels, as well as the segmentation of thin structure were considered. The simulation study indicates that the algorithm is performant in other multi-class segmentation problems. We reported the application of the proposed method to the real datasets. 3D SR-μCT was used to image the bone vasculatures on tibia of mice, with contrast agent. The segmentation quality was evaluated using the Dice coefficient and the Matthews correlation coefficient (MCC) by comparing to the manual segmentations on representative small volumes. The results indicate that segmentation quality with the proposed method is much improved compared to hysteresis thresholding or gradient based watershed segmentations. To characterize bone and vasculatures, several quantitative parameters were extracted. Finally, statistical analysis was performed to study the influence of anti-angiogenesis drugs on bone and vessels, in the context of breast cancer bone metastases.
