Chen Mingyang, Elazab Ahmed, Jia Fucang, Wu Jianhuang, Li Guanglin, Li Xiaodong, Hu Qingmao
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
University of Chinese Academy of Sciences, Beijing, China.
Australas Phys Eng Sci Med. 2015 Dec;38(4):627-41. doi: 10.1007/s13246-015-0372-3. Epub 2015 Aug 29.
Cerebral glioma is one of the most aggressive space-occupying diseases, which will exhibit midline shift (MLS) due to mass effect. MLS has been used as an important feature for evaluating the pathological severity and patients' survival possibility. Automatic quantification of MLS is challenging due to deformation, complex shape and complex grayscale distribution. An automatic method is proposed and validated to estimate MLS in patients with gliomas diagnosed using magnetic resonance imaging (MRI). The deformed midline is approximated by combining mechanical model and local symmetry. An enhanced Voigt model which takes into account the size and spatial information of lesion is devised to predict the deformed midline. A composite local symmetry combining local intensity symmetry and local intensity gradient symmetry is proposed to refine the predicted midline within a local window whose size is determined according to the pinhole camera model. To enhance the MLS accuracy, the axial slice with maximum MSL from each volumetric data has been interpolated from a spatial resolution of 1 mm to 0.33 mm. The proposed method has been validated on 30 publicly available clinical head MRI scans presenting with MLS. It delineates the deformed midline with maximum MLS and yields a mean difference of 0.61 ± 0.27 mm, and average maximum difference of 1.89 ± 1.18 mm from the ground truth. Experiments show that the proposed method will yield better accuracy with the geometric center of pathology being the geometric center of tumor and the pathological region being the whole lesion. It has also been shown that the proposed composite local symmetry achieves significantly higher accuracy than the traditional local intensity symmetry and the local intensity gradient symmetry. To the best of our knowledge, for delineation of deformed midline, this is the first report on both quantification of gliomas and from MRI, which hopefully will provide valuable information for diagnosis and therapy. The study suggests that the size of the whole lesion and the location of tumor (instead of edema or the sum of edema and tumor) are more appropriate to determine the extent of deformation. Composite local symmetry is recommended to represent the local symmetry around the deformed midline. The proposed method could be potentially used to quantify the severity of patients with cerebral gliomas and other brain pathology, as well as to approximate midsagittal surface for brain quantification.
脑胶质瘤是最具侵袭性的占位性疾病之一，由于占位效应会出现中线移位（MLS）。MLS已被用作评估病理严重程度和患者生存可能性的重要特征。由于变形、形状复杂和灰度分布复杂，MLS的自动量化具有挑战性。本文提出并验证了一种自动方法，用于估计使用磁共振成像（MRI）诊断的胶质瘤患者的MLS。通过结合力学模型和局部对称性来近似变形中线。设计了一种考虑病变大小和空间信息的增强型Voigt模型来预测变形中线。提出了一种结合局部强度对称性和局部强度梯度对称性的复合局部对称性，以在根据针孔相机模型确定大小的局部窗口内细化预测中线。为了提高MLS的准确性，从每个容积数据中选取具有最大MSL的轴向切片，将其空间分辨率从1毫米插值到0.33毫米。所提出的方法在30例公开可用的呈现MLS的临床头部MRI扫描上得到了验证。它描绘了具有最大MLS的变形中线，与真实值的平均差异为0.61±0.27毫米，平均最大差异为1.89±1.18毫米。实验表明，当病理的几何中心为肿瘤的几何中心且病理区域为整个病变时，所提出的方法将产生更好的准确性。还表明，所提出的复合局部对称性比传统的局部强度对称性和局部强度梯度对称性具有显著更高的准确性。据我们所知，对于变形中线的描绘，这是第一篇关于胶质瘤量化以及来自MRI的报告，有望为诊断和治疗提供有价值的信息。该研究表明，整个病变的大小和肿瘤的位置（而不是水肿或水肿与肿瘤的总和）更适合确定变形程度。建议使用复合局部对称性来表示变形中线周围的局部对称性。所提出的方法可能潜在地用于量化脑胶质瘤和其他脑部病变患者的严重程度，以及近似脑矢状面进行脑量化。
