An Independent Active Contours Segmentation framework for bone micro-CT images

Micro-CT is an imaging technique for small tissues and objects that is gaining increased popularity especially as a pre-clinical application. Nevertheless, there is no well-established micro-CT segmentation method, while typical procedures lack sophistication and frequently require a degree of manual intervention, leading to errors and subjective results. To address these issues, a novel segmentation framework, called Independent Active Contours Segmentation (IACS), is proposed in this paper. The proposed IACS is based on two autonomous modules, namely automatic ROI extraction and IAC Evolution, which segments the ROI image using multiple Active Contours that evolve simultaneously and independently of one another. The proposed method is applied on a Phantom dataset and on real datasets. It is tested against several established segmentation methods that include Adaptive Thresholding, Otsu Thresholding, Region Growing, Chan-Vese (CV) AC, Geodesic AC and Automatic Local Ratio-CV AC, both qualitatively and quantitatively. The results prove its superior performance in terms of object identification capability, accuracy and robustness, under normal circumstances and under four types of artificially introduced noise. These enhancements can lead to more reliable analysis, better diagnostic procedures and treatment evaluation of several bone-related pathologies, and to the facilitation and further advancement of bone research.