脑深部刺激是一种有效地治疗帕金森病的外科疗法，其疗效与电极能否植入到丘脑底核功能区内密切相关。通常，外科医生通过手工方法确定植入区域和靶点位置，能否采用计算机辅助方法一直是这个领域密切关注的问题，本文研究了丘脑底核的自动定位算法，通过术前靶点自动分割和定位以及术后电极位置评估等系列工作对靶点定位问题展开系统化研究，为临床实践提供准确、可靠、有效的定位方法。首先基于患者术前核磁图像实现了丘脑底核的半自动分割，为靶点自动定位提供了技术支持。在与传统方法比较基础上，本文采用窄带限制的水平集方法完成了丘脑底核分割，并讨论了核磁图像失真、循环终止条件及不同初始轮廓曲线对分割结果的影响，验证了该方法的有效性和鲁棒性。其次，基于核磁图像中核团间位置关系提出了丘脑底核的自动定位方法，目的是帮助外科医生术前有效地定位靶点，制定最佳的手术方案。在前章方法的基础上，结合最大似然估计实现对核团轮廓的自动描绘。在自动描绘的基础上，结合核团的形态特点，采用隐式样条曲线方法实现了核团三维重建，有助于定位的可视化。最终结合丘脑底核与红核间位置关系实现了核团的自动定位。通过对比专家定位与自动定位靶点到标准靶点距离，发现自动定位方法与专家系统定位方法没有显著差异，并且具有较小的预测误差，表明了该算法的有效性。最后，手术定位的术后评估是对定位精度的最终判断标准，通常是采用术后核磁成像，本文提出了利用术后核磁图像中电极伪影估计电极位置的新方法。手术定位的术后评估是对定位精度的最终判断标准，通常采用术后核磁成像进行电极位置的评估。本文从植入角入手，建立了电极周围磁场分布的模型，根据MRI空间编码过程计算基于植入角度的电极伪影，利用术后核磁图像中电极伪影实现了电极位置的估计，有助于外科医生利用术后靶点位置研究疗效最佳的刺激靶点，指导医生选择最佳的植入角度和程控，获得最好的植入效果。通过电极伪影实验和在体实验的核磁图像的分析，数据表明该方法可以测量电极在丘脑底核中的位置，并具有较小测量误差。
Deep brain stimulation （DBS） of the subthalamic nucleus （STN） is an effective treatment for advanced Parkinson’s Diseases （PD）.Thetherapeutic benefitsof DBS forreducing motor symptoms are closely related to the accurate placement of the electrodes in the motor sub-region of the STN.Neurosurgeons commonlyuse manual methods to locate the STN for electrode insertion during surgery, which carries the risk of human error.Therefore, automatically locating the STN for DBS therapy provides potential benefit in functional neurosurgery.This dissertation providesa computational process applied to 3TMagnetic ResonanceImaging （MRI） in preoperative segmentation and postoperative assessment of PD patients for DBS that automatically locates the STN. The methods of STN segmentation and reconstruction were designed in order to automaticallylocate the STN based on preoperative MRI.An algorithm for 3.0T MRI image segmentation using the narrow limited level set method was designed to reconstructthe STN. The impact on segmentation was analysed according to image distortion, loop iteration break and arbitrary initial contour.Implicitpolynomial surfaces were used for the reconstruction of the STN segmentation.Together, the analyses indicated that our method was effective.The automatic location method based on relative position of the nucleus in the MRI image was presented to help surgeons effectively locate thepreoperativetarget and prepare a best plan for operation.An automated description of the STN boundary was createdusing a maximum likelihood estimate and level set methods according to the segmentation method. Based on this automatic description, the location method was proposed. Comparison of the Euclidean distances and dice overlap coefficient showed no significant differences with the segmentation-based method, with the present method having smaller prediction errors and being more robust than expert systems.Due to artefacts from the electrodes in postoperative MRI images, determining the electrode position remains problematic.In our study, the method for estimation of the electrode position throughpitching angle and deflection angle was presented in order to calculate theposition between the electrode and the STN.This was aimed to correlatethetherapeutic effect and electrode position,foraidingneurosurgeons in theirappropriatepost-surgical analyses. Following theassessmentof artefactsin MRI datafromphantom and 10 PD patients, comparisonsbetween Euclidean distances and data on the correlation coefficients showed that this method is suitable for measuring the electrode position with a prediction error.
