当前位置: X-MOL 学术Trans. Emerg. Telecommun. Technol. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A brain tumor detection system using gradient based watershed marked active contours and curvelet transform
Transactions on Emerging Telecommunications Technologies ( IF 2.5 ) Pub Date : 2020-11-09 , DOI: 10.1002/ett.4170
Pelin Görgel 1
Affiliation  

Computer aided brain tumor detection is an efficient research area in brain image processing. In this study, a methodology called GWMAC-CT (gradient based watershed marked active contours and curvelet transform) is proposed to detect the brain tumors in magnetic resonance (MR) images. The implemented system is based on skull removing, segmentation of region of interest (ROI), feature extraction, and ROI classification as tumor or nontumor. The proposed GWMAC is a two-stage segmentation method which includes gradient based watershed transform (GWT) and improved active contours. The rough ROIs obtained with GWT are utilized as initial contours for the improved active contours method instead of marking initial contours manually. Curvelet transform-based features of the exact ROI contours are classified via well-known classification methods such as support vector machine (SVM), K-nearest neighbors, random forest tree, and Naïve Bayes. Experiments are carried out on a set of brain MR images from BRATS database to demonstrate the effectiveness of the proposed method. The performance evaluators such as accuracy, kappa statistics, false positive rate, precision, F1-measure, and area under ROC curve are calculated as 96.81%, 0.927, 0.046, 0.905, 0.95, and 0.977, respectively with SVM.

中文翻译:

使用基于梯度的分水岭标记活动轮廓和曲波变换的脑肿瘤检测系统

计算机辅助脑肿瘤检测是脑图像处理中一个有效的研究领域。在这项研究中,提出了一种称为 GWMAC-CT(基于梯度的分水岭标记活动轮廓和曲线波变换)的方法来检测磁共振 (MR) 图像中的脑肿瘤。实施的系统基于颅骨去除、感兴趣区域 (ROI) 的分割、特征提取和 ROI 分类为肿瘤或非肿瘤。所提出的 GWMAC 是一种两阶段分割方法,包括基于梯度的分水岭变换 (GWT) 和改进的活动轮廓。用 GWT 获得的粗略 ROI 被用作改进的活动轮廓方法的初始轮廓,而不是手动标记初始轮廓。精确 ROI 轮廓的基于 Curvelet 变换的特征通过众所周知的分类方法进行分类,例如支持向量机 (SVM)、K-最近邻、随机森林树和朴素贝叶斯。对来自 BRATS 数据库的一组脑 MR 图像进行了实验,以证明所提出方法的有效性。使用 SVM 计算准确率、kappa 统计量、误报率、精度、F1-measure 和 ROC 曲线下面积等性能评估器分别为 96.81%、0.927、0.046、0.905、0.95 和 0.977。
更新日期:2020-11-09
down
wechat
bug