Abstract:Objective To explore the improved algorithm of active contour model and apply it to the
automatic segmentation of spinal CT image. Methods First, fuzzy corner algorithm based on image gray
was adopted to mark the corner feature set of the subject goal. Second, the concave hull algorithm based
on α shape was used to outline the initial contour which is close to the real subject goal. Finally, this
initial contour was taken as the initial evolutionary conditions of active contour model to achieve image
automatic segmentation. Results Different segmentation algorithms were chosen to conduct simulation
experiment. Qualitative analysis showed that the proposed method retained a clear and complete image
edge and details. Quantitative evaluation results indicated that the improved algorithm can get a maximal Dice
similarity coefficient and a minimal Hausdorff distance measure, and can precisely segment subject goal
even in a noisy environment. Conclusion The fuzzy corner algorithm and concave hull algorithm can
effectively avoid the blindness of the initial contour selection, thus the active contour model evolves more
rapidly so that can acquire a more accurate subject goal contour. The proposed algorithm is a feasible for
spinal CT image segmentation and is more stable and universally applicable than other algorithm even in
a noisy environment, thus has greater clinical application value in terms of goal analysis.