Radiomic texture analysis based on neurite orientation dispersion and density imaging to differentiate glioblastoma from solitary brain metastasis

Bai, Jie; He, Mengyang; Gao, Eryuan; Yang, Guang; Yang, Hongxi; Dong, Jie; Ma, **aoyue; Gao, Yufei; Zhang, Huiting; Yan, Xu; Zhang, Yong; Cheng, **gliang; Zhao, Guohua

doi:10.1186/s12885-023-11718-0

Radiomic texture analysis based on neurite orientation dispersion and density imaging to differentiate glioblastoma from solitary brain metastasis

Research
Open access
Published: 14 December 2023

Volume 23, article number 1231, (2023)
Cite this article

Download PDF

You have full access to this open access article

BMC Cancer Aims and scope Submit manuscript

Radiomic texture analysis based on neurite orientation dispersion and density imaging to differentiate glioblastoma from solitary brain metastasis

Download PDF

Jie Bai¹^na1,
Mengyang He²^na1,
Eryuan Gao¹,
Guang Yang³,
Hongxi Yang³,
Jie Dong⁴,
**aoyue Ma¹,
Yufei Gao²,
Huiting Zhang⁵,
Xu Yan⁵,
Yong Zhang¹,
**gliang Cheng¹ &
…
Guohua Zhao¹

902 Accesses
Explore all metrics

Abstract

Background

We created discriminative models of different regions of interest (ROIs) using radiomic texture features of neurite orientation dispersion and density imaging (NODDI) and evaluated the feasibility of each model in differentiating glioblastoma multiforme (GBM) from solitary brain metastasis (SBM).

Methods

We conducted a retrospective study of 204 patients with GBM (n = 146) or SBM (n = 58). Radiomic texture features were extracted from five ROIs based on three metric maps (intracellular volume fraction, orientation dispersion index, and isotropic volume fraction of NODDI), including necrosis, solid tumors, peritumoral edema, tumor bulk volume (TBV), and abnormal bulk volume. Four feature selection methods and eight classifiers were used for the radiomic texture feature selection and model construction. Receiver operating characteristic (ROC) curve analysis was used to evaluate the diagnostic performance of the models. Routine magnetic resonance imaging (MRI) radiomic texture feature models generated in the same manner were used for the horizontal comparison.

Results

NODDI-radiomic texture analysis based on TBV subregions exhibited the highest accuracy (although nonsignificant) in differentiating GBM from SBM, with area under the ROC curve (AUC) values of 0.918 and 0.882 in the training and test datasets, respectively, compared to necrosis (AUC_training:0.845, AUC_test:0.714), solid tumor (AUC_training:0.852, AUC_test:0.821), peritumoral edema (AUC_training:0.817, AUC_test:0.762), and ABV (AUC_training:0.834, AUC_test:0.779). The performance of the five ROI radiomic texture models in routine MRI was inferior to that of the NODDI-radiomic texture model.

Conclusion

Preoperative NODDI-radiomic texture analysis based on TBV subregions shows great potential for distinguishing GBM from SBM.

Key points

• NODDI-radiomic texture analysis can distinguish glioblastoma multiforme from solitary brain metastasis.

• ROI division affects efficiency, and tumor bulk volume outperformed other ROIs.

• NODDI-based outperforms routine magnetic resonance imaging-based radiomic texture analysis.

View this article's peer review reports

Extracted magnetic resonance texture features discriminate between phenotypes and are associated with overall survival in glioblastoma multiforme patients

Article 10 March 2016

Texture analysis on conventional MRI images accurately predicts early malignant transformation of low-grade gliomas

Article 07 January 2019

MRI radiomics to differentiate between low grade glioma and glioblastoma peritumoral region

Article 25 October 2021

Background

Glioblastoma multiforme (GBM) and solitary brain metastases (SBM) are the most common malignant brain tumors, and their correct identification is key for further diagnosis and treatment [1,2,3]. Although magnetic resonance imaging (MRI) is the main tool for differentiating between the two types of tumors, both GBM and SBM may show marked peritumoral edema and similar contrast-enhancement patterns on routine MRI, leading to great challenges in identification [4,5,6].

Previous studies reported that radiomics combined with routine MRI showed significant advantages in distinguishing GBM from SBM and suggested that specific imaging features are helpful in distinguishing between the two types of tumors [7, 8]. Currently, the acquisition of specific image features can be summarized into two trends: applying special MRI modalities or focusing on specific image feature types [9, 10].

Diffusion-weighted imaging (DWI) can provide a class of microscopic features related to the movement of water molecules in tissues, such as the current advanced diffusion imaging model and neurite orientation dispersion and density imaging (NODDI) [11, 12]. NODDI is a multi-spherical shell diffusion model based on the difference in the diffusion of water molecules inside and outside the cell and is more often used to characterize the difference in water diffusion between tumor infiltration and vasogenic edema [Full size image

Radiomic texture extraction and model construction

Feature extraction, feature selection, and model building were performed using the open-source software FeAture Explorer (FAE, version 0.5.2) [23]. Based on the automatically segmented ROIs, radiomic texture features were extracted using first-order statistical functions, gray-level co-occurrence matrix (GLCM) functions, and gray-level run-length matrix (GLRLM) functions on the original NODDI parametric maps, as well as eight sub-bands of its wavelet transformation. As controls, the features of routine MR images (T2WI, T2-dark-fluid, T1WI, and CE-T1 MPRAGE) were extracted in the same manner. Radiomic texture analysis for each ROI was based on a combination of three parametric map features from the NODDI or a combination of four routine MRI features. Finally, 234 features were extracted from each parameter map of NODDI (or each type of routine MRI). Details of the extracted features are presented in Supplementary Appendix E2.

Because of the imbalanced GBM-to-SBM sample ratio (2.5:1), we applied upsampling to the training dataset. After feature extraction, all radiomic texture feature values were normalized using the min-max or Z-score method. Four feature selection methods—Pearson’s correlation coefficient (PCC), analysis of variance, recursive feature elimination, and the Kruskal-Wallis test—as well as eight classifiers—support vector machine, linear discriminant analysis, auto-encoder, random forest, logistic regression, logistic regression via Lasso, ada-boost, and decision tree—were utilized to construct texture feature prediction models for each ROI. When the PCC value of a feature pair was greater than 0.90, only one of the features was randomly retained. Five-fold cross-validation was used to determine the hyperparameters of each model. After determining the hyperparameters, all training data were retrained for the final models. The maximum number of features included in the radiomic texture analysis model construction was four. Details of the sample size and feature number estimates are displayed in Supplementary Appendix E3. The final models were determined based on the highest area under the receiver operating characteristic (ROC) curve (AUC) value in the cross-validation, and a time-independent test dataset was used to evaluate the performance of the final model. The performance of the test dataset was determined through ROC curve analysis and evaluations of accuracy, AUC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Statistical analysis

Statistical analyses were performed using SPSS (version 21.0) and MedCalc (version 20.015) software. Differences in clinical characteristics between GBM and SBM were assessed using chi-square tests (or the Mann–Whitney U test, depending on the results of normality and homoscedasticity tests) and independent t-tests, as appropriate. DeLong’s test was used to assess differences in AUC values between models. Statistical significance was set at a two-sided p value < 0.05.

Results

Patient clinical characteristics

The patients’ clinical characteristics are summarized in Table 1. No significant differences were found in clinical characteristics between the patients in the training and test datasets (all p > 0.05). A total of 146 (71.5%) patients had GBMs, and 58 (28.5%) were diagnosed with SBM by pathological examination. The GBM rates were 71.3% (102/143) and 72.1% (44/61) for the training and test datasets, respectively, with no significant difference between the two (p = 0.907).

Performance of texture feature prediction models

The NODDI-radiomic texture model based on the five ROIs performed differently when discriminating between GBM and SBM, and the TBV radiomic texture model exhibited the best performance. In the training set, we determined AUCs for the necrosis, solid tumor, peritumoral edema, TBV, and ABV texture models of 0.845, 0.852, 0.817, 0.918, and 0.834; the same values for the five models in the test set were 0.714, 0.821, 0.762, 0.882, and 0.779. Figure 3 shows the cross-validation, training set, and test set AUCs of the texture model for the five ROIs.

Table 3 Performance on the test dataset: NODDI and routine MRI radiomic texture model

Full size table

The performance of the five ROI texture models on routine MRI was inferior to that of the NODDI-radiomic texture model. The AUCs of the necrosis, solid tumor, peritumoral edema, TBV, and ABV texture models were 0.712, 0.836, 0.853, 0.770, and 0.813 in the training set and 0.651, 0.786, 0.806, 0.723, and 0.818 in the test set.

A more detailed comparison of the evaluation indicators between the NODDI and conventional MRI radiomic texture models for the five ROIs in the test set (accuracy, sensitivity, specificity, PPV, and NPV) is provided in Table 3. The NODDI TBV texture model achieved better sensitivity (84.1%) and specificity (82.3%), proving its excellent performance on imbalanced datasets. More details on the model construction are provided in Fig. 4, showing the cross-validation set, training set, and test set AUCs for different classifiers in the determination of optimal model performance. Figures 5 and 6, and Supplementary Table 1 show the feature values, distributions and statistical correlations, contribution of features, and methods used in the key modeling steps for each model. The DeLong test results for each model are presented in Supplementary Table 2.

Discussion

To the best of our knowledge, this is the first study that comprehensively explores the value of radiomic texture analysis based on preoperative NODDI in differentiating GBM from SBM. Our findings reveal the excellent potential and application value of NODDI-radiomic texture analysis based on the TBV for differentiating between the two types of tumors.

Analogous to histogram analysis, radiomic texture analysis is a nimble and interpretable method. Prior research has showcased the efficacy of MRI texture analysis in differentiating GBM from SBM [24, 25]. Ortiz-Ramón et al. proposed a method that combines two-dimensional texture analysis and machine learning technology to differentiate between GBM and SBM. They obtained an AUC of 0.896 by constructing a model using 32 texture features [24]. Our research, based on texture analysis, attained comparable discriminatory capability utilizing four texture features. Moreover, preceding studies proposed that routine MRI possesses restricted capacity to differentiate between the two tumor types. Han et al. developed a radiomics model on T2WI images to distinguish GBM and SBM, achieving an AUC of 0.696 [25]. Our results corroborate this assertion; although four routine sequences were combined to construct the model, only moderate discriminatory capability was ultimately achieved. Our results indicate that the NODDI texture model outperforms routine MRI texture models. The clinical benefits of NODDI over MRI include the elimination of risks associated with the use of contrast agents and a reduction in the number of parameters needed. We speculate that texture analysis based on NODDI can quantify the extent of axonal dispersion or damage within the lesion, which is very effective in distinguishing between GBM and SBM.

Various ROI settings lead to the difference in the identification ability of the model. Although specific texture features can be found in tumors and peritumoral edema, the discrimination ability of these features shows two trends. The model containing tumor regions showed a stronger ability to distinguish GBM from SBM. In fact, the tumor parenchymal area in GBM exhibits dense cell growth, heterogeneity, and nuclear atypia morphology. The tumor parenchyma of SBM exhibits relative homogeneity. This difference can be attributed to the different growth patterns of the two types of tumors. The radiomics texture analysis model based on NODDI seems to be able to distinguish the growth patterns of GBM and SBM. The pathological differences manifested by the peritumoral edema area of GBM (comprising tumor cells and new blood vessels) lead to the formation of invasive edema. In SBM, normal brain tissue is compressed by tumors, and simple vasogenic edema forms around the tumor. However, the identification efficiency of the edema area model is significantly reduced, and the radiomics texture analysis model based on NODDI cannot well distinguish invasive edema from angiogenic edema. In general, among the models based on five different ROI subregions, our NODDI-radiomic texture model based on TBV performed the best. TBV is a combination of necrosis and tumor parenchyma, and we speculate that relying on TBV guarantees the continuity and integrity of tumor texture features, leading to stronger identification ability.

Previous studies have reported progress in the ability of DWI and diffusion tensor imaging (DTI) to differentiate between GBM and SBM [26,27,28]. The meta-analysis conducted by Suh et al. suggests that the diagnostic abilities of DWI and DTI are limited (i.e., a pooled sensitivity of 79.8% [95% CI, 70.9–86.4%] and a pooled specificity of 80.9% [95% CI, 75.1–85.5%]), despite their wide individual sensitivity and specificity [26]. DWI and DTI are better suited as part of a multiparametric MRI protocol than as single sequences [25]. NODDI, an extension of DWI, is the only imaging modality with the potential to accurately differentiate GBM from SBM [29].

First-order statistical features are considered low-order features, whereas the GLCM and GLRLM are considered high-order features and are usually extracted from the original image or the wavelet image derived from the original image [10]. These features describe the statistical relationships between image pixels (voxels) from different perspectives and are often highly correlated and redundant [30]. Feature selection is therefore necessary. Feature selection simplifies the features used for texture analysis, excludes non-contributing and highly correlated features, reduces the redundancy and multicollinearity of candidate texture features, and facilitates the use of machine learning models for evaluation. In this study, four candidate features for building radiomic texture analysis models were identified, and the values of these reproducible features were independent of the traditional clinicopathological features.

The fifth edition of the World Health Organization classification of central nervous system tumors (WHO CNS5) emphasizes the value of molecular pathology in the diagnosis of GBM [31]. Most pathology centers need time to adapt to changes in the new WHO CNS5, thereby delaying the update of classification standards. Notably, all patients whose data were included in the present study were classified based on the 2016 WHO guidelines.

This study has, however, several limitations. First, the relatively small sample size was adequate for our tentative exploration but limits statistical power as well as the generalizability of our results. Second, samples need to be divided into training and testing datasets for internal validation. Unfortunately, we were not able to externally validate our results. Finally, it is imperative to enhance the transparency surrounding the associations between biological interpretability and advanced imaging parameters. This will enable the provision of substantial evidence for the treatment and care of tumors, as well as microscopic studies such as single-cell analysis [32, 33].

Conclusions

Preoperative NODDI-radiomic texture analysis based on TBV shows great potential for distinguishing GBM from SBM. Further studies are required to explore the generalizability of our findings through external validation and to apply these results to clinical practice.

Data Availability

The datasets generated and/or analyzed during the current study are not publicly available due to the ongoing further studies but are available from the corresponding author (ghzhao@ha.edu.cn) on reasonable request.

Abbreviations

ABV:: Abnormal bulk volume
AUC:: Area under the ROC curve
DTI:: Diffusion tensor imaging
DWI:: Diffusion-weighted imaging
GBM:: Glioblastoma multiforme
GLCM:: Gray-level co-occurrence matrix
GLRLM:: Gray-level run-length matrix
MRI:: Magnetic resonance imaging
NODDI:: Neurite orientation dispersion and density imaging
NPV:: Negative predictive value
PCC:: Pearson’s correlation coefficient
PPV:: Positive predictive value
ROC:: Receiver operating characteristic
ROI:: Region of interest
SBM:: Solitary brain metastasis
T2WI:: T2-weighted imaging
TBV:: Tumor bulk volume

References

Fordham AJ, Hacherl CC, Patel N, et al. Differentiating glioblastomas from solitary brain metastases: an update on the current literature of advanced imaging modalities. Cancers. 2021;13(12):2960. https://doi.org/10.3390/cancers13122960
Article CAS PubMed PubMed Central Google Scholar
Boire A, Brastianos PK, Garzia L, Valiente M. Brain metastasis. Nat Rev Cancer. 2020;20(1):4–11. https://doi.org/10.1038/s41568-019-0220-y
Article CAS PubMed Google Scholar
Bette S, Huber T, Wiestler B, et al. Analysis of fractional anisotropy facilitates differentiation of glioblastoma and brain metastases in a clinical setting. Eur J Radiol. 2016;85(12):2182–7. https://doi.org/10.1016/j.ejrad.2016.10.002
Article PubMed Google Scholar
Weller M, van den Bent M, Hopkins K, et al. EANO guideline for the diagnosis and treatment of anaplastic gliomas and glioblastoma. Lancet Oncol. 2014;15(9):e395–403. https://doi.org/10.1016/S1470-2045(14)70011-7
Article PubMed Google Scholar
Dong F, Li Q, Jiang B, et al. Differentiation of supratentorial single brain metastasis and glioblastoma by using peri-enhancing oedema region-derived radiomic features and multiple classifiers. Eur Radiol. 2020;30(5):3015–22. https://doi.org/10.1007/s00330-019-06460-w
Article PubMed Google Scholar
Lah TT, Novak M, Breznik B. Brain malignancies: Glioblastoma and brain metastases. Semin Cancer Biol. 2020;60:262–73. https://doi.org/10.1016/j.semcancer.2019.10.010
Article CAS PubMed Google Scholar
Artzi M, Bressler I, Ben Bashat D. Differentiation between glioblastoma, brain metastasis and subtypes using radiomics analysis. J Magn Reson Imaging: JMRI. 2019;50(2):519–28. https://doi.org/10.1002/jmri.26643
Article PubMed Google Scholar
Huang Y, Huang S, Liu Z. Multi-task learning-based feature selection and classification models for glioblastoma and solitary brain metastases. Front Oncol. 2022;12:1000471. https://doi.org/10.3389/fonc.2022.1000471
Article PubMed PubMed Central Google Scholar
Lai PH, Chung HW, Chang HC, et al. Susceptibility-weighted imaging provides complementary value to diffusion-weighted imaging in the differentiation between pyogenic brain abscesses, necrotic glioblastomas, and necrotic metastatic brain tumors. Eur J Radiol. 2019;117:56–61. https://doi.org/10.1016/j.ejrad.2019.05.021
Article PubMed Google Scholar
Romano A, Moltoni G, Guarnera A, et al. Single brain metastasis versus glioblastoma multiforme: a VOI-based multiparametric analysis for differential diagnosis. Radiol Med. 2022;127(5):490–7. https://doi.org/10.1007/s11547-022-01480-x
Article PubMed PubMed Central Google Scholar
Zhao J, Li JB, Wang JY, et al. Quantitative analysis of neurite orientation dispersion and density imaging in grading gliomas and detecting IDH-1 gene mutation status. Neuroimage Clin. 2018;19:174–81. https://doi.org/10.1016/j.nicl.2018.04.011
Article PubMed PubMed Central Google Scholar
Zhang H, Schneider T, Wheeler-Kingshott CA, Alexander DC. NODDI: practical in vivo neurite orientation dispersion and density imaging of the human brain. NeuroImage. 2012;61(4):1000–16. https://doi.org/10.1016/j.neuroimage.2012.03.072
Article PubMed Google Scholar
**e Y, Li S, Shen N, et al. Assessment of Isocitrate dehydrogenase 1 genotype and cell proliferation in gliomas using multiple diffusion magnetic resonance imaging. Front Neurosci. 2021;22(15):783361. https://doi.org/10.3389/fnins.2021.783361
Article Google Scholar
Gao E, Gao A, Kit Kung W, et al. Histogram analysis based on diffusion kurtosis imaging: differentiating glioblastoma multiforme from single brain metastasis and comparing the diagnostic performance of two region of interest placements. Eur J Radiol. 2022;147:110104. https://doi.org/10.1016/j.ejrad.2021.110104
Article PubMed Google Scholar
Qi J, Wang P, Zhao G et al. (2022) Histogram analysis based on neurite orientation dispersion and density MR imaging for differentiation between glioblastoma multiforme and solitary brain metastasis and comparison of the diagnostic performance of two ROI placements. J Magn Reson Imaging. https://doi.org/10.1002/jmri.28419
Castellano G, Bonilha L, Li LM, Cendes F. Texture analysis of medical images. Clin Radiol. 2004;59:1061–9. https://doi.org/10.1016/j.crad.2004.07.008
Article CAS PubMed Google Scholar
Ruiz-España S, Ortiz-Ramón R, Pérez-Ramírez Ú, et al. MRI texture-based radiomics analysis for the identification of altered functional networks in alcoholic patients and animal models. Comput Med Imaging Graph. 2023;104:102187. https://doi.org/10.1016/j.compmedimag.2023.102187
Article PubMed Google Scholar
Mo X, Chen W, Chen S, et al. MRI texture-based machine learning models for the evaluation of renal function on different segmentations: a proof-of-concept study. Insights into Imaging. 2023;14(1):28. https://doi.org/10.1186/s13244-023-01370-4
Article PubMed PubMed Central Google Scholar
Wang X, Dai Y, Lin H, et al. Shape and texture analyses based on conventional MRI for the preoperative prediction of the aggressiveness of pituitary adenomas. Eur Radiol. 2023;33(5):3312–21. https://doi.org/10.1007/s00330-023-09412-7
Article PubMed Google Scholar
Li J, Fu S, Gong Z, et al. MRI-based texture analysis of infrapatellar fat pad to predict knee osteoarthritis incidence. Radiology. 2022;304:611–21. https://doi.org/10.1148/radiol.212009
Article PubMed Google Scholar
Cui Z, Ren G, Cai R, et al. MRI-based texture analysis for differentiate between pediatric posterior fossa ependymoma type A and B. Eur J Radiol. 2022;152:110288. https://doi.org/10.1016/j.ejrad.2022.110288
Article PubMed Google Scholar
Isensee F, Jaeger PF, Kohl SAA, Petersen J, Maier-Hein KH. nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nat Methods. 2021;18(2):203–11. https://doi.org/10.1038/s41592-020-01008-z
Article CAS PubMed Google Scholar
Song Y, Zhang J, Zhang YD, et al. FeAture Explorer (FAE): a tool for develo** and comparing radiomics models. PLoS ONE. 2020;15(8):e0237587. https://doi.org/10.1371/journal.pone.0237587
Article CAS PubMed PubMed Central Google Scholar
Ortiz-Ramón R, Ruiz-España S, Mollá-Olmos E, Moratal D. Glioblastomas and brain metastases differentiation following an MRI texture analysis-based radiomics. Approach Phys Med. 2020;76:44–54. https://doi.org/10.1016/j.ejmp.2020.06.016
Article PubMed Google Scholar
Han Y, Zhang L, Niu S, et al. Differentiation between glioblastoma multiforme and metastasis from the lungs and other sites using combined clinical/routine MRI radiomics. Front cell Dev Biology. 2021;9:710461. https://doi.org/10.3389/fcell.2021.710461
Article Google Scholar
Suh CH, Kim HS, Jung SC, Kim SJ. Diffusion-weighted imaging and diffusion tensor imaging for differentiating high-grade glioma from solitary brain metastasis: a systematic review and meta-analysis. AJNR Am J Neuroradiol. 2018;39(7):1208–14. https://doi.org/10.3174/ajnr.A5650
Article CAS PubMed PubMed Central Google Scholar
Skogen K, Schulz A, Helseth E, et al. Texture analysis on diffusion tensor imaging: discriminating glioblastoma from single brain metastasis. Acta Radiol. 2019;60(3):356–66. https://doi.org/10.1177/0284185118780889
Article PubMed Google Scholar
Mao J, Zeng W, Zhang Q, et al. Differentiation between high-grade gliomas and solitary brain metastases: a comparison of five diffusion-weighted MRI models. BMC Med Imaging. 2020;20(1):124. https://doi.org/10.1186/s12880-020-00524-w
Article PubMed PubMed Central Google Scholar
Kadota Y, Hirai T, Azuma M, et al. Differentiation between glioblastoma and solitary brain metastasis using neurite orientation dispersion and density imaging. J Neuroradiol. 2020;47(3):197–202. https://doi.org/10.1016/j.neurad.2018.10.005
Article PubMed Google Scholar
Demircioğlu A. Evaluation of the dependence of radiomic features on the machine learning model. Insights into Imaging. 2022;13(1):28. https://doi.org/10.1186/s13244-022-01170-2
Article PubMed PubMed Central Google Scholar
Louis DN, Perry A, Wesseling P, et al. The 2021 WHO classification of tumors of the central nervous system: a summary. Neuro Oncol. 2021;23(8):1231–51. https://doi.org/10.1093/neuonc/noab106
Article CAS PubMed PubMed Central Google Scholar
Liu C, Zhang M, Yan X, et al. Single-cell dissection of cellular and molecular features underlying human cervical squamous cell carcinoma initiation and progression. Sci Adv. 2023;9(4):eadd8977. https://doi.org/10.1126/sciadv.add8977
Article CAS PubMed PubMed Central Google Scholar
Liu C, Wang P, Sun Y, et al. Neoadjuvant chemoradiotherapy changes the landscape of soluble immune checkpoint molecules in patients with locally advanced rectal cancer. Front Oncol. 2022;12:756811. https://doi.org/10.3389/fonc.2022.756811
Article CAS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

The authors thank all colleagues who participated in this study for their support and help.

Funding

This work is supported in part by the National Natural Science Foundation of China (82202270), and in part by the Natural Science Foundation of Henan Province (232300421298), and in part by the Scientific and Technological Research Project of Henan Province (LHGJ20220403).

Author information

Jie Bai and Mengyang He contributed equally to this work.

Authors and Affiliations

Department of Magnetic Resonance Imaging, The First Affiliated Hospital of Zhengzhou University, No. Jianshe Dong Road, Zhengzhou, 450052, China
Jie Bai, Eryuan Gao, **aoyue Ma, Yong Zhang, **gliang Cheng & Guohua Zhao
School of Cyber Science and Engineering, Zhengzhou University, Zhengzhou, 450001, China
Mengyang He & Yufei Gao
Shanghai Key Laboratory of Magnetic Resonance, East China Normal University, Shanghai, 200062, China
Guang Yang & Hongxi Yang
School of Information Engineering, North China University of Water Resources and Electric Power, Zhengzhou, 450046, China
Jie Dong
MR Research Collaboration, Siemens Healthineers, Wuhan, 201318, China
Huiting Zhang & Xu Yan

Authors

Jie Bai
View author publications
You can also search for this author in PubMed Google Scholar
Mengyang He
View author publications
You can also search for this author in PubMed Google Scholar
Eryuan Gao
View author publications
You can also search for this author in PubMed Google Scholar
Guang Yang
View author publications
You can also search for this author in PubMed Google Scholar
Hongxi Yang
View author publications
You can also search for this author in PubMed Google Scholar
Jie Dong
View author publications
You can also search for this author in PubMed Google Scholar
**aoyue Ma
View author publications
You can also search for this author in PubMed Google Scholar
Yufei Gao
View author publications
You can also search for this author in PubMed Google Scholar
Huiting Zhang
View author publications
You can also search for this author in PubMed Google Scholar
Xu Yan
View author publications
You can also search for this author in PubMed Google Scholar
Yong Zhang
View author publications
You can also search for this author in PubMed Google Scholar
**gliang Cheng
View author publications
You can also search for this author in PubMed Google Scholar
Guohua Zhao
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

JB, MH and GZ was involved in conceptualization. EG, YZ, JC, and GZ were involved in data curation. GY, CZ and HY performed the imaging texture analysis. JD, XM and YG were involved in methodology. YZ, JC and GZ were involved in project administration and supervision. JB, HZ, XY and GZ was involved in resources, visualization, and writing—original draft. GY and HY were involved in software. MH, JD and YG were involved in validation. MH, GY, JC and GZ were involved in writing—review and editing. All authors approved for publication.

Corresponding author

Correspondence to Guohua Zhao.

Ethics declarations

Ethics approval and consent to participate

All methods were performed in accordance with the Declaration of Helsinki. The study was approved by the ethics committee of the First Affiliated Hospital of Zhengzhou University (2019-KY-231). Informed consent was waived by the institutional review board of the First Affiliated Hospital of Zhengzhou University, because this was a retrospective study and the privacy and identity information of the subjects were guaranteed.

Consent for publication

This study has obtained publishing licenses from all study participants.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1: Supplemental Appendix

. E1: ROI segmentation. E2: List of extracted radiomic texture features. E3: Sample size and featyre number estimation. Supplemental Table 1. Radiomic texture analysis models and features on different ROIs, and key steps in modeling. Supplemental Table 2. The DeLong test result in test datasets

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article

Bai, J., He, M., Gao, E. et al. Radiomic texture analysis based on neurite orientation dispersion and density imaging to differentiate glioblastoma from solitary brain metastasis. BMC Cancer 23, 1231 (2023). https://doi.org/10.1186/s12885-023-11718-0

Download citation

Received: 14 September 2023
Accepted: 05 December 2023
Published: 14 December 2023
DOI: https://doi.org/10.1186/s12885-023-11718-0

Radiomic texture analysis based on neurite orientation dispersion and density imaging to differentiate glioblastoma from solitary brain metastasis

Abstract

Background

Methods

Results

Conclusion

Key points

Similar content being viewed by others

Extracted magnetic resonance texture features discriminate between phenotypes and are associated with overall survival in glioblastoma multiforme patients

Texture analysis on conventional MRI images accurately predicts early malignant transformation of low-grade gliomas

MRI radiomics to differentiate between low grade glioma and glioblastoma peritumoral region

Background

Radiomic texture extraction and model construction

Statistical analysis

Results

Patient clinical characteristics

Performance of texture feature prediction models

Discussion

Conclusions

Data Availability

Abbreviations

References

Acknowledgements

Funding

Author information

Authors and Affiliations

Contributions

Corresponding author

Ethics declarations

Ethics approval and consent to participate

Consent for publication

Competing interests

Additional information

Publisher’s Note

Electronic supplementary material

Supplementary Material 1: Supplemental Appendix

Rights and permissions

About this article

Cite this article

Share this article

Keywords

Search

Navigation