结合峰值检测的前列腺磁共振成像偏移场平滑拟合

Smooth fitting of bias field in prostate MRI with peak detection

目的研究前列腺磁共振图像中灰度不均匀现象(偏场)的校正方法。材料与方法从几组前列腺磁共振扫描数据中截取的横断面图像。将真实图像的分段常量特性与偏移场的平滑变化特性表达到图像模型中,构造一个能量函数,通过能量函数的最小化实现偏移场评估和组织分割。利用峰值检测技术自动获得能量函数的初始化参数,并用结合三角函数与多项式函数的一组基函数实现对偏移场的平滑拟合。结果定性的实验表明笔者的方法能对前列腺磁共振图像中的偏场现象进行有效的校正。另外通过与其他方法在变化系数、均方根、Jaccard相似度等指标下的定量对比发现,笔者的方法有更好的校正结果。结论结合峰值检测的偏移场校正方法能对前列腺磁共振图像中的灰度不均匀现象有效改善。

Objective： To study correction of the inhomogeneity of grayscale （Bias Field） in prostate MR image. Materials and Methods： Several transverse images derived from magnetic resonance scanning data of prostate. The piecewise constant property of the real image and the smooth change characteristic of the bias field are expressed in the image model. All energy function is constructed and the bias field estimation and tissue segmentation are realized by minimizing the energy function. The initial parameters of the energy function are obtained automatically by using the peak detection technique, and the smoothing fitting of the offset field is realized by using a set of basis functions combined with trigonometric thnctions and polynomial functions. Results： Some qualitative evaluations showed the significant improvement of prostate MR image with severe intensity inhomogeneity by using our method. The comparison with other methods in some quantitative evaluation indexes （Coefficient of variation, Root mean square and Jaccard similarity） is shown to demonstrate the better result of our method. Conclusion： Peak detection based bias correction method can perfect the intensity inhomogeneity in prostate MR image.