当前位置: X-MOL 学术Pattern Recogn. Lett. › 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.)
Two-stage multi-modal MR images fusion method based on Parametric Logarithmic Image Processing (PLIP) Model
Pattern Recognition Letters ( IF 5.1 ) Pub Date : 2020-05-27 , DOI: 10.1016/j.patrec.2020.05.027
Vikrant Bhateja , Mansi Nigam , Anuj Singh Bhadauria , Anu Arya

MRI is one of the most compliant technique that is used for the screening of Brain Tumor. MRI can be acquired in four available modalities which are MR-T1, MR-T2, MR-PD and MR-Gad; among these MR-T2 comprises of most of the detailed information of the tumorous tissues. However, the accuracy and reliability of the diagnosis may be affected due to lack of sufficient details in each modality (as different MRI modalities highlight different set of tissues). Therefore, MR Image(s) fusion is essential to obtain a more illustrative image containing the requisite complementary details of each modality. For this purpose, multi-modal fusion of MR-T2 with MR-T1, MR-PD and MR-Gad have been dealt in this work using the proposed fusion method. This paper presents a two-stage fusion method using Stationary Wavelet Transform (SWT) in combination with Parameterized Logarithmic Image Processing (PLIP) model. At Stage-I of sub-band decomposition: the first level SWT coefficients contain large amount of noise thus suppressing the necessary edge information. This aspect has been resolved at Stage-II by employing second level SWT decomposition along with Principal Component Analysis (PCA). The fusion coefficients from both the stages are finally fused using PLIP operators (prior to reconstruction). The obtained results are compared qualitatively as well as quantitatively using fusion metrics like Entropy, Fusion Factor, Standard Deviation and Edge Strength. Noteworthy visual response is obtained with PLIP fusion model in coherence with Human Visual System (HVS) characteristics.



中文翻译:

基于参数对数图像处理模型的两阶段多模态MR图像融合方法

MRI是用于筛查脑肿瘤的最合规技术之一。可以通过四种可用的方式获取MRI:MR-T1,MR-T2,MR-PD和MR-Gad。在这些MR-T2中,大多数包含肿瘤组织的详细信息。但是,由于每种方式缺乏足够的细节,诊断的准确性和可靠性可能会受到影响(因为不同的MRI方式会突出显示不同的组织集)。因此,一个或多个MR图像融合对于获得包含每个模态的必要补充细节的更具说明性的图像至关重要。为此,在这项工作中,使用提出的融合方法处理了MR-T2与MR-T1,MR-PD和MR-Gad的多模式融合。本文提出了一种结合平稳小波变换(SWT)和参数化对数图像处理(PLIP)模型的两阶段融合方法。在子带分解的阶段I:第一级SWT系数包含大量噪声,因此抑制了必要的边缘信息。通过使用二级SWT分解和主成分分析(PCA),已在第二阶段解决了这一方面。最后,使用PLIP运算符对两个阶段的融合系数进行融合(在重建之前)。使用诸如熵,融合因子,标准偏差和边缘强度之类的融合指标对获得的结果进行定性和定量比较。与人的视觉系统(HVS)特征一致的PLIP融合模型获得了值得注意的视觉响应。

更新日期:2020-05-27
down
wechat
bug