贝叶斯优化的RSF模型脑肿瘤图像分割新方法

摘要

由于核磁共振成像(MRI,magnetic resonance imaging)模糊、灰度不均,使得脑肿瘤图像分割精确度不高,给出了一种贝叶斯优化的自适应RSF模型.传统RSF模型的水平集分割性能受初始化和控制参数影响较大,需要大量人工干预,限制了其在实际中的应用.利用贝叶斯估计的自适应性,自动提取初始轮廓,并用于RSF模型细分割脑肿瘤图像,得到了一种脑肿瘤MRI图像分割新方法.结果表明,实验采用 Jaccard 系数和分割时间评估分割方法的精度和效率,与RSF-mean shift方法相比,其分割精度提高20% 以上,分割效率提高32% 以上.%Blurred images and uneven gray level of magnetic resonance imaging(MRI)make the brain tumor image segmentation imprecise.In order to solve the problem,a new brain tumor image segmentation method is proposed,which is an adaptive RSF(Region-Scalable Fitting) model optimized by Bayesian estimation.The traditional RSF model's level set segmentation performance depends on proper initialization and optimal configuration of control parameters, requiring a lot of manual intervention,which limits its practical application.This method uses the adaptability of Bayesian estimation,automatically extracting the initial contour,and is also used in RSF model to subdivide brain tumor images,which get a new image segmentation method for MRI brain tumor.The Jaccard index and segmentation time are used to evaluate the precision and efficiency of the new method.The results show that compared with the RSF-mean shift method,the segmentation precision increases by more than 20% and the segmenta-tion efficiency increases by more than 32% by the new method.