Spiculation Quantification Method Based on Edge Gradient Orientation
Histogram
Guodong Zhang
School of Computer
Shenyang Aerospace University
Shenyang , China
E-mail:zhanggd@sau.edu.cn
Nan Xiao
School of Computer
Shenyang Aerospace University
Shenyang , China
E-mail:ln758342@163.com
Wei Guo*
School of Computer
Shenyang Aerospace University
Shenyang , China
E-mail: guowei@sau.edu.cn
Abstract—A new approach to quantify the lung nodule
spiculation levels in CT (computed tomography)
images was proposed. Firstly, two-dimensional image
of nodule was generated by using the spiral scan
technology. Secondly, the lung nodule was segmented
in the two-dimensional image using the dynamic
programming. Thirdly, based on the expanded regions
from the segmented boundary, the spiculation was
segmented by use of threshold segmentation. Fourthly,
the feature region where spiculation may exist was
extracted on the region of spiculation boundary.
Finally, the edge gradient orientation histogram as a
new quantitative index was extracted to quantify the
spiculation levels of the lung nodule. The experimental
results indicate that this index can quantify the
spiculation levels accurately with better classification
results.
Keywords: lung nodules; CT (computed tomography);
spiculation features; edge gradient orientation histogram
I. I
NTRODUCTION
The lung nodule spiculation in CT images show a
radial, non-branched, straight and short line shadow on the
extend region along the segmented boundary. Spiculation
is a primary sign of diagnosing malignant lung nodules.
The research that quantifying the lung nodule spiculation
levels is popular on computer-aided diagnosis.
Quantification methods can be classified into three
categories: the ones based on the degree of surface rules
[1-2], the ones based on the texture feature [3-10] and the
ones based on gradient feature [11-13].
In this paper, a new computerized quantification
method was proposed to evaluate the lung nodule
spiculation levels in CT images. Spiral Scan Technology
was used to transform a three-dimensional nodule into a
two-dimension image [14]. The nodule boundary was
segmented on two-dimension image. Based on segmented
boundary, the extend region where spiculation may exist
was extracted by expanding certain range outward along
the segmented boundary. The spiculation was segmented
on extend region using threshold segmentation algorithm.
Edge gradient orientation histogram was used as a new
quantitative index to analysis the extend region, using this
new index to evaluate the lung nodule spiculation levels in
CT images [15-18]. The CT images used in this study
were obtained from the standard CT lung nodule database
provided by the Lung Image Database Consortium
(LIDC).
II.
METHODS
A. Spiculation segmentation
Most existing methods for extracting spiculation focus
on the center layer of the nodule and did not take into
account three dimensional structure of the nodule. Two-
dimensional image with three-dimensional information
was generated by using Spiral Scan Technology (see
Fig.1). The dynamic programming algorithm was
employed for accurate segmentation of nodules (see
Fig.2). The spiculation was not segmented on the above
process and the spiculation was located in annular region
outside the nodule boundary. Based on accurate
segmentation of nodules, the extend region where
spiculation may exist was generated by expanding the
certain range outward along the nodule boundary (see
Fig.3). Generally, the empirical value of extend distance is
10 pixels. The threshold segmentation algorithm was
employed for accurate segmentation of nodule spiculation
(see Fig.4). The average gray value of the bottom line on
extend region was used as the threshold value. The
specific process was described as follows: searching each
column from the bottom of extend region, the first point
which was less than the threshold was selected as the
boundary point. These boundary points constitute a
spiculation boundary set B.
2014 International Conference on Virtual Reality and Visualization
978-1-4799-6854-1/14 $31.00 © 2014 IEEE
DOI 10.1109/ICVRV.2014.26
86
资源评论
weixin_38606300
- 粉丝: 4
- 资源: 829
最新资源
- Linux环境下Nginx服务器的源码安装与自动启动配置指南
- 【Unity 插件】DLSS - Upscaling for Unity 将低分辨率图像提升为高分辨率图像，接近或超越原生分辨率
- 基于角色访问控制的Linux安全模块+项目源码+文档说明
- 基于uniapp构建的顺风车、约车、拼车、通勤、滴滴微信小程序（源码+文档说明）
- 【Unity 插件】Invector FSM AI Template 易用的 AI 模板, 轻松实现敌人的巡逻、追击、攻击等动作
- 文本数据可视化tocsv.csv
- 基于HSV色彩空间和樽海鞘群优化算法的低照度图像增强
- Fine-BI考试全量题库（含答案）
- yolo的xtx数据集增强
- 技术资料分享RDA5820很好的技术资料.zip
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
