乳腺MRI分割与基于光线投影的三维重建

Breast tumor MRI segmentation and 3D reconstruction based on ray casting method

目的乳腺MRI(magnetic resonance imaging)序列成像连续、位置固定,但二维图像无法提供便于观察的肿块的深度与外形信息,为此提出乳腺及其肿块分割与三维重建方法。方法首先利用最大类间方差法与区域生长法提取乳腺外轮廓,然后选用基于梯度和能量信息的水平集方法准确分割乳腺肿块,并对分割结果采用基于包围盒的光线投影算法进行体绘制。结果对患者的多组MRI序列进行分割实验并对几种分割算法进行客观评价,三维重建实验得到乳腺与肿块的立体图像以供医生观察。结论分区处理的水平集分割与基于光线投影的三维重建为医生提供更加形象的立体视觉效果,不仅有助于对肿块病变程度的诊断,而且能帮助外科医生在保乳手术中准确判断开刀位置。

Objective MRI sequences in breast are continuous with fixed positions, yet 2D images cannot conveniently provide the information of the depth and the whole shape of the tumor to be observed. In this case, the paper proposes a method of 3D reconstruction on the segmentation of breast and tumor. Methods Firstly, the contour of the breast is extracted by the use of the combination method of Otsu and region growing based on region processing. Then, it makes use of the level set based on energy and gradient to segment the tumor exactly from breast. Finally, it chooses volume rendering to deal with the segmentation results and adopts the ray casting method based on the bounding box to realize the reconstruction. Results This paper conducts the experiments on multiple groups of patients＇ MRI, and evaluates objectively on several segmentation algorithms. Experiments of 3D reconstruction get the stereoscopic image of the breast and tumor to be observed for doctors. Conclusions Segmentation based on level set of region processing and 3D reconstruction based on ray casting method provide a more vivid 3D visual effect for doctors, which can not only contribute to the diagnosis of the severity of the tumor, but help surgeons to decide the surgical site accurately in breast-conserving surgery.