Segmentation Of Image Sequences Of Neuron Stem Cells Based On Level-set Algorithm Combined With Local Gray Threshold

Image segmentation plays an extremely important role in medical image processing. Medical images typically own problems of noise and edge blur because of uneven tissues and organs wriggling. Thus the main research direction of image segmentation is to avoid impact of noise and image sharpness, and get accurate contour of target area.Neural stem cell (NSC) owns the abilities of division and self-regeneration. We can cure the nervous disease, such as stroke, spinal cord injury, senile dementia etc by artificial injecting NSC. Observing the procession of NSC division and movement can increase diagnosis accuracy and cure rate. As the precondition, NSC image segmentation becomes more and more important.In time lapse of neuron stem cells image sequences imaged by optical microscope, for solving the problems of low contrast ratio between objects and background and clustered cells'interference, we present a novel algorithm, which is based on a level-set without re-initialization. Adding curvature term to accelerate convergence, iteration terminating condition is changed to measure norm energy for decreasing complexity. Local gray threshold is combined with the result of curve evolution for clustered cells' separation at last. The presented algorithm is applied in two sequence images of 120 frames. The results show that the algorithm can not only solve the problem of focal shift but also separate adherent and clustered cells successfully as well as keep cells'shape and location. After compared with watershed and traditional level-set algorithms, this algorithm can improve the success rate to 30%-40%, which has made a good preparation for cells recognition and cells tracking in further research work.