González-Villà Sandra, Valverde Sergi, Cabezas Mariano, Pareto Deborah, Vilanova Joan C, Ramió-Torrentà Lluís, Rovira Àlex, Oliver Arnau, Lladó Xavier
Institute of Computer Vision and Robotics, University of Girona, Ed. P-IV, Campus Montilivi, 17073 Girona, Spain.
Institute of Computer Vision and Robotics, University of Girona, Ed. P-IV, Campus Montilivi, 17073 Girona, Spain.
Neuroimage Clin. 2017 May 8;15:228-238. doi: 10.1016/j.nicl.2017.05.003. eCollection 2017.
In recent years, many automatic brain structure segmentation methods have been proposed. However, these methods are commonly tested with non-lesioned brains and the effect of lesions on their performance has not been evaluated. Here, we analyze the effect of multiple sclerosis (MS) lesions on three well-known automatic brain structure segmentation methods, namely, FreeSurfer, FIRST and multi-atlas fused by majority voting, which use learning-based, deformable and atlas-based strategies, respectively. To perform a quantitative analysis, 100 synthetic images of MS patients with a total of 2174 lesions are simulated on two public databases with available brain structure ground truth information (IBSR18 and MICCAI'12). The Dice similarity coefficient (DSC) differences and the volume differences between the healthy and the simulated images are calculated for the subcortical structures and the brainstem. We observe that the three strategies are affected when lesions are present. However, the effects of the lesions do not follow the same pattern; the lesions either make the segmentation method underperform or surprisingly augment the segmentation accuracy. The obtained results show that FreeSurfer is the method most affected by the presence of lesions, with DSC differences (generated - healthy) ranging from - 0.11 ± 0.54 to 9.65 ± 9.87, whereas FIRST tends to be the most robust method when lesions are present (- 2.40 ± 5.54 to 0.44 ± 0.94). Lesion location is not important for global strategies such as FreeSurfer or majority voting, where structure segmentation is affected wherever the lesions exist. On the other hand, FIRST is more affected when the lesions are overlaid or close to the structure of analysis. The most affected structure by the presence of lesions is the nucleus accumbens (from - 1.12 ± 2.53 to 1.32 ± 4.00 for the left hemisphere and from - 2.40 ± 5.54 to 9.65 ± 9.87 for the right hemisphere), whereas the structures that show less variation include the thalamus (from 0.03 ± 0.35 to 0.74 ± 0.89 and from - 0.48 ± 1.08 to - 0.04 ± 0.22) and the brainstem (from - 0.20 ± 0.38 to 1.03 ± 1.31). The three segmentation approaches are affected by the presence of MS lesions, which demonstrates that there exists a problem in the automatic segmentation methods of the deep gray matter (DGM) structures that has to be taken into account when using them as a tool to measure the disease progression.
近年来，人们提出了许多自动脑结构分割方法。然而，这些方法通常是在无病变的大脑上进行测试的，病变对其性能的影响尚未得到评估。在此，我们分析了多发性硬化症（MS）病变对三种著名的自动脑结构分割方法的影响，即FreeSurfer、FIRST以及通过多数投票融合的多图谱方法，它们分别采用基于学习、可变形和基于图谱的策略。为了进行定量分析，在两个具有可用脑结构真实信息的公共数据库（IBSR18和MICCAI'12）上模拟了100例患有总共2174个病变的MS患者的合成图像。计算了皮质下结构和脑干在健康图像与模拟图像之间的骰子相似系数（DSC）差异和体积差异。我们观察到，当存在病变时，这三种策略都会受到影响。然而，病变的影响并不遵循相同的模式；病变要么使分割方法表现不佳，要么出人意料地提高分割精度。所得结果表明，FreeSurfer是受病变影响最大的方法，DSC差异（生成图像 - 健康图像）范围为 -0.11 ± 0.54至9.65 ± 9.87，而FIRST在存在病变时往往是最稳健的方法（-2.40 ± 5.54至0.44 ± 0.94）。病变位置对于FreeSurfer或多数投票等全局策略并不重要，在这些策略中，无论病变存在于何处，结构分割都会受到影响。另一方面，当病变覆盖或靠近分析结构时，FIRST受到的影响更大。受病变存在影响最大的结构是伏隔核（左半球为 -1.12 ± 2.53至1.32 ± 4.00，右半球为 -2.40 ± 5.54至9.65 ± 9.87），而变化较小的结构包括丘脑（0.03 ± 0.35至0.74 ± 0.89以及 -0.48 ± 1.08至 -0.04 ± 0.22）和脑干（-0.20 ± 0.38至1.03 ± 1.31）。这三种分割方法都受到MS病变的影响，这表明在将深部灰质（DGM）结构的自动分割方法用作测量疾病进展的工具时，必须考虑到存在的问题。
