Pinto Adriano, Pereira Sergio, Correia Higino, Oliveira J, Rasteiro Deolinda M L D, Silva Carlos A
Annu Int Conf IEEE Eng Med Biol Soc. 2015 Aug;2015:3037-40. doi: 10.1109/EMBC.2015.7319032.
Gliomas are among the most common and aggressive brain tumours. Segmentation of these tumours is important for surgery and treatment planning, but also for follow-up evaluations. However, it is a difficult task, given that its size and locations are variable, and the delineation of all tumour tissue is not trivial, even with all the different modalities of the Magnetic Resonance Imaging (MRI). We propose a discriminative and fully automatic method for the segmentation of gliomas, using appearance- and context-based features to feed an Extremely Randomized Forest (Extra-Trees). Some of these features are computed over a non-linear transformation of the image. The proposed method was evaluated using the publicly available Challenge database from BraTS 2013, having obtained a Dice score of 0.83, 0.78 and 0.73 for the complete tumour, and the core and the enhanced regions, respectively. Our results are competitive, when compared against other results reported using the same database.
神经胶质瘤是最常见且侵袭性最强的脑肿瘤之一。这些肿瘤的分割对于手术和治疗规划很重要，对随访评估也很重要。然而，这是一项艰巨的任务，因为其大小和位置各不相同，而且即使借助磁共振成像（MRI）的所有不同模态，勾勒出所有肿瘤组织也并非易事。我们提出了一种用于神经胶质瘤分割的判别式全自动方法，利用基于外观和上下文的特征来输入到极端随机森林（Extra-Trees）中。其中一些特征是在图像的非线性变换上计算得出的。使用来自2013年BraTS公开可用的挑战数据库对所提出的方法进行了评估，对于完整肿瘤、核心区域和强化区域，分别获得了0.83、0.78和0.73的Dice分数。与使用同一数据库报告的其他结果相比，我们的结果具有竞争力。
