Glioblastoma and primary central nervous system lymphoma: Preoperative differentiation by using MRI-based 3D texture analysis
Introduction
Glioblastomas(GBMs) and primary central nervous system lymphomas (PCNSLs) are common intracranial malignant tumors that share many visual characteristics on magnetic resonance imaging (MRI) [1,2]. Preoperative differentiation of GBMs and PCNSLs is of great clinical importance because the therapy and prognosis for the two entities are markedly different. For instance, GBMs are almost always treated by surgical intervention followed by radiation therapy and chemotherapy, whereas the treatment of choice for PCNSLs is methotrexate-based chemotherapy after a stereotactic biopsy [1,2]. In addition, preoperative steroid use usually complicates definitive histopathological diagnosis [1]. As a result, preoperative differentiation of GBMs and PCNSLs is a significant problem in need of a solution.
Although preliminary studies suggest that advanced MRI may be valuable in the task of distinguishing brain tumors, conventional MRI remains the preferred method for preoperative assessment in clinical practice [[3], [4], [5], [6], [7], [8], [9]]. However, the value of qualitative information obtained from conventional MRI is often limited. For example, atypical solid enhanced glioblastomas without visible necrosis may appear very similar to typical PCNSLs, making it difficult and sometimes even impossible to distinguish the two by qualitative methods (Fig. 1). Thus, objective quantitative methods based on conventional MRI may be clinically useful as a supplement to qualitative methods.
Texture analysis is a well-established technique to quantify parameters of a set of pixels (2D) or voxels (3D), i.e., the gray-level intensity and the distribution and relationship of image texture. Texture and shape analysis have been widely applied to medical images in recent years [10,11]. A considerable amount of research interest has been directed towards texture and shape analyses, which have attracted particular attention in the field of neurology, where these methods have demonstrated promising results [[12], [13], [14], [15], [16]]. Furthermore, some recent studies have concluded that heterogeneity information obtained from the total tumor mass (achieved by multiple MR image slices) is more valuable than that obtained from a single slice in texture analysis [[17], [18], [19]]. Considering the high rate of imaging-based misdiagnosis between GBM and PCNSL, we hypothesized that MR-based 3D texture and shape analysis would be able to aid in differentiating between GBM and PCNSL. Accordingly, machine-learning models were built to improve discrimination between these conditions using the results of this analysis.
Section snippets
Patients
We performed a database search of radiology and pathology reports from our Hospital between January 2012 and September 2017. Patients with solitary brain tumors shown on conventional MRI were identified by evaluating the MR image reports and postoperative pathological diagnoses. The histopathological diagnoses were determined according to the 2007 World Health Organization (WHO) classification criteria for glioblastoma and PCNSL [22,23]. The exclusion criteria included (1) non-enhancement or
Results
The three features with high predictive power were as follows: (1) Firstorder_Skewness, (2) Firstorder_Kurtosis and (3) Ngtdm_Busyness. The shape features showed no differences between glioblastomas and PCNSLs. However, according to the Mann-Whitney U test, the three above-mentioned parameters demonstrated statistically significant differences between GBMs and PCNSLs (p < 0.001, p < 0.001, and p < 0.001, respectively). Each individual ROC analysis of Firstorder_Skewness, Firstorder_Kurtosis and
Discussion
In this study, two common brain malignancies, GBMs and PCNSLs were differentiated by using MRI-based 3D texture and shape analysis. Three selected texture features with high predictive power were fed into the classification models. The results indicate that the texture feature analysis performs well and that the combination of the three machine-learning models aids in differentiation between the two tumors.
The first question in this study involves the efficacy of 3D texture analysis. According
Conclusions
MRI-based 3D texture analysis has potential utility in preoperative discrimination of glioblastomas and PCNSLs. In our study, the features with the highest predictive power in CE-T1WI were Firstorder_Skewness, Ngtdm_Busynessin, and Firstorder_Kurtosis, which had high sensitivity, specificity, accuracy and AUC. Machine-learning models using these three features also showed promising results, demonstrating their usefulness in the differentiation of glioblastomas and PCNSLs.
Funding sources
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Conflicts of interest
None.
Acknowledgements
None.
References (45)
- et al.
Lymphomas and glioblastomas: differences in the apparent diffusion coefficient evaluated with high b-value diffusion-weighted magnetic resonance imaging at 3 T
Eur. J. Radiol.
(2012) - et al.
Distinguishing of primary cerebral lymphoma from high-grade glioma with perfusion-weighted magnetic resonance imaging
Neurosci. Lett.
(2003) - et al.
Texture analysis of medical images
Clin. Radiol.
(2004) - et al.
Characterization of breast cancer types by texture analysis of magnetic resonance images
Acad. Radiol.
(2010) - et al.
The potential value of preoperative mri texture and shape analysis in grading meningiomas: a preliminary investigation
Transl. Oncol.
(2017) - et al.
Assessment of tumor heterogeneity by CT texture analysis: can the largest cross-sectional area be used as an alternative to whole tumor analysis?
Eur. J. Radiol.
(2013) - et al.
A functional role for tumor cell heterogeneity in a mouse model of small cell lung cancer
Cancer Cell
(2011) - et al.
Three dimensional texture analysis in MRI: a preliminary evaluation in gliomas
Magn. Reson. Imaging
(2003) - et al.
Radiomics: the process and the challenges
Magn. Reson. Imaging
(2012) - et al.
Improved lesion detection in MR mammography: three-dimensional segmentation, moving voxel sampling, and normalized maximum intensity–time ratio entropy
Acad. Radiol.
(2007)
Three-dimensional solid texture analysis in biomedical imaging: review and opportunities
Med. Image Anal.
Primary CNS lymphoma
J. Clin. Oncol.
Diagnosis and treatment of recurrent high-grade astrocytoma
J. Clin. Oncol.
Patterns of contrast enhancement in the brain and meninges
Radiographics
Primary central nervous system lymphoma and atypical glioblastoma: multiparametric differentiation by using diffusion-, perfusion-, and susceptibility-weighted MR imaging
Radiology
Lymphomas and high-grade astrocytomas: comparison of water diffusibility and histologic characteristics
Radiology
Evaluation of different cerebral mass lesions by perfusion-weighted MR imaging
J. Magn. Reson. Imaging
High-grade gliomas and solitary metastases: differentiation by using perfusion and proton spectroscopic MR imaging
Radiology
Texture analysis in magnetic resonance imaging: review and considerations for future applications
Assessment of Cellular and Organ Function and Dysfunction using Direct and Derived MRI Methodologies
Texture-based classification of focal liver lesions on MRI at 3.0 Tesla: a feasibility study in cysts and hemangiomas
J. Magn. Reson. Imaging
Magnetic resonance imaging texture analysis classification of primary breast cancer
Eur. Radiol.
Neurogenic and myogenic diseases: quantitative texture analysis of muscle US data for differentiation
Radiology
Cited by (0)
- 1
These authors contributed equally to the work.