Abstract
Importance of computed tomography (CT) images lies in imaging speed, image contrast & resolution and clinical cost. It has found wide use in detection and diagnosis of brain diseases. Unfortunately reported works on CT segmentation is not very significant. In this paper the gaps in automatic brain segmentation are identified and addressed with possible solutions which results into a robust, automatic computer aided diagnosis (CAD) system. Adaptive thresholding, automatic seed point finding, knowledge driven region growing and multilevel masking are key potentials of this work. Two types of masks are created. One mask is used to segment brain matter and another one restricts inclusion of non-intracranial area of nasal image slices. This second mask is a global reference mask for all slices in a dataset whereas the brain intensity mask is implemented on adjacent slices and automatically updated for next slice. Mask propagates independently in two different directions keeping reference slice at centre. Successive propagation and adaptive modification of brain mask have demonstrated very high potential in brain segmentation. Presented result shows highest sensitivity i.e. 1 and more than 96% accuracy in all cases. These segmented images can be used for any brain disease diagnosis or further image analysis as no information within brain is altered or removed.
Abstract (translated)
计算机断层扫描(CT)图像的重要性在于成像速度,图像对比度和图像对比度。分辨率和临床费用。它已被广泛用于脑疾病的检测和诊断。不幸的是,报道的CT分割工作并不是很重要。在本文中,通过可能的解决方案识别和解决自动脑分割中的差距,这些解决方案导致强大的自动计算机辅助诊断(CAD)系统。自适应阈值处理,自动种子点发现,知识驱动区域增长和多级掩蔽是这项工作的关键潜力。创建了两种类型的蒙版。一个面罩用于分割脑部物质,另一个面罩限制包含鼻部图像切片的非颅内区域。第二个掩码是数据集中所有切片的全局参考掩码,而脑强度掩码在相邻切片上实现,并自动更新为下一个切片。掩模在两个不同方向上独立传播,保持参考切片位于中心。脑屏蔽的连续传播和自适应修改已经证明在脑分割中具有非常高的潜力。所呈现的结果显示出最高的灵敏度,即在所有情况下均为1且超过96%的准确度。这些分割的图像可以用于任何脑部疾病诊断或进一步的图像分析,因为脑内的信息没有被改变或去除。
URL
https://arxiv.org/abs/1809.06215