Multimodal MRI lesion habitat-based radiomics analysis for preoperative prediction of spatial pattern in locally recurrent high-grade gliomas

Multimodal MRI lesion habitat-based radiomics analysis for preoperative prediction of spatial pattern in locally recurrent high-grade gliomas | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multimodal MRI lesion habitat-based radiomics analysis for preoperative prediction of spatial pattern in locally recurrent high-grade gliomas Han-wei Wang, Lin-lan Zeng, Xiao-guang Li, Mi-mi Zhao, Xuan Li, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3870027/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Purpose This study aims to preoperatively predict spatial patterns in locally recurrent high-grade gliomas (HGGs) based on lesion habitat radiomics analysis of multimodal MRI and to evaluate the predictive performance of this approach. Methods Our study included 121 patients with locally recurrent HGGs after maximum safe surgical resections and radiotherapy combined with temozolomide (training set, n = 84; validation set, n = 37). Local recurrence was divided into intra-resection cavity recurrence (ICR) and extra-resection cavity recurrence (ECR), according to the distance between the recurrent tumor and the surgical area or resection cavity. Radiomic features were extracted from the lesion habitat (T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region) on contrast-enhanced T1WI and FLAIR, respectively. The LASSO was used to select radiomic features and calculate radiomics score. Logistic regression analysis was used to construct a predictive radiomics model, which was evaluated using calibration curves and the area under the receiver operating characteristic curve (AUC). Results Seven features with nonzero coefficients related to spatial recurrence patterns were selected. The radiomics score of patients with ECR was higher than that of patients with ICR in the training set [0.424 (0.278–0.573) vs. -0.030 (-0.226-0.248), p < 0.001] and in the validation set [0.369 (0.258–0.487) vs. 0.277 (0.103–0.322), p = 0.033]. The radiomics model demonstrated good calibration and performed well in predicting ECR, with AUC values of 0.844 in the training set and 0.706 in the validation set. Conclusion Radiomics analysis of lesion habitat can preoperatively predict spatial patterns in locally recurrent HGGs, providing a basis for determining personalized treatment strategies for HGGs. glioma recurrence habitat magnetic resonance imaging radiomics Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction High-grade gliomas (HGGs, WHO grade Ⅲ-Ⅳ) are highly heterogeneous tumors of the central nervous system. Despite standardized treatments, the prognosis of HGGs patients has not improved, with more than half of patients relapsing within 6–8 months after surgery [ 1 , 2 ]. The high recurrence rate is a critical factor affecting the prognosis of the patients. Recurrence of HGGs most commonly occurs locally within 2–3 cm of the resection cavity, and in a few cases far from the original resection cavity or surgical area [ 3 – 6 ]. What’s more, locally recurrent HGGs exhibits different spatial patterns. Based on postoperative magnetic resonance imaging (MRI), local recurrence can be divided into intra-resection cavity recurrence (ICR) and extra-resection cavity recurrence (ECR), according to the distances between the recurrent tumor and the resection cavity [ 3 ]. Wang et al. analyzed that the spatial recurrence patterns and prognosis in 69 patients with locally recurrent HGGs and revealed patients with ICR had longer prognosis than patients with ECR [ 7 ]. Currently, the treatment of recurrent HGGs is dominated by individualized therapy, and it has been reported that surgical resection of recurrent glioblastoma may help to prolong patients’ overall survival [ 8 ]. Thus, early prediction of the spatial recurrence patterns may help to individualize patient treatment and improve prognosis. Previous studies have shown that molecular pathological markers and clinical factors were related to spatial recurrence patterns [ 5 , 9 – 13 ]. However, there is a lack of imaging markers for preoperative non-invasive prediction of spatial recurrence patterns of HGGs. MRI is the main non-invasive imaging tool for the HGGs management, which plays an important role in diagnosis, efficacy evaluation and follow-up. Previous studies have shown that certain MRI features, such as subventricular zone (SVZ) involvement, ventricular entry, and T2-weighted-fluid-attenuated inversion recovery (T2WI/FLAIR) abnormal transformation, are associated with recurrence patterns [ 12 , 14 – 16 ]. However, subjective judgment by radiologists is susceptible to interobserver variability, which affects the accuracy and reliability of the assessment results. In contrast, radiomics overcomes this problem by high-throughput extraction and analysis of quantitative image features from many medical images to correlate with clinical outcomes [ 17 , 18 ]. In addition, radiomics can more accurately reflect tumor heterogeneity by segmenting tumor subregion regions of interest (ROI). Ismail et al. [ 19 ] segmented the lesion habitat and found that 3D shape features of the enhancing lesion on T1WI, and T2WI/FLAIR hyperintensities could distinguish tumor progression and pseudoprogression. Habitat-based MRI radiomics have also been shown to predict MGMT promoter methylation and prognosis in astrocytoma [ 20 ]. To the best of our knowledge, a few studies have reported on the prediction of recurrence patterns in HGGs based on radiomic features [ 21 , 22 ]. These studies suggest that habitat-based radiomics may be equally applicable to preoperative prediction of spatial recurrence patterns in HGGs. Therefore, this study aims to preoperatively predict spatial patterns in locally recurrent high-grade gliomas (HGGs) based on lesion habitat radiomics analysis of multimodal magnetic resonance imaging (MRI) and to evaluate the predictive performance of this approach. Materials and Methods Patients This study was performed in line with the principles of the Declaration of Helsinki. Approved by the institutional review board of the Army Medical Center of the PLA, and requirement of written informed consent was waived (Date November 9, 2022 /No.2022 − 325). We performed this study on patients with locally recurrent HGGs after maximum safe surgical resections and radiotherapy combined with temozolomide between January 2012 and October 2021. Patients who met the following inclusion criteria were enrolled in our study: (1) had histopathological confirmation of newly diagnosed HGGs according to the 2016 WHO classification criteria without evidence of subarachnoid dissemination on MRI; (2) had no history of chemotherapy or radiotherapy treatment before preoperative imaging; (3) underwent preoperative MRI, including T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), T2WI/FLAIR imaging, and contrast-enhanced (CE) T1WI; (4) underwent maximum safe surgical resection and radiotherapy combined with temozolomide; and (5) aged > 18 years. Exclusion criteria were as follows: (1) infratentorial HGGs; (2) low-quality MR images, including noise and artifacts; (3) loss of follow-up; (4) no recurrence during the follow-up; and (5) non-locally recurrent HGGs. Follow-up All patients underwent CE-MRI or CT or both 24–72 h postoperatively and were followed up with CE-MRI every 3–6 months. The patients were followed until death or the study cutoff time (October 2023). The evidence of tumor recurrence included (1) patients who underwent reoperation to confirm recurrence, (2) continuous CE-MRI according to the Response Assessment in Neuro-Oncology criteria, or (3) multimodal MRI to rule out radiation necrosis or pseudoprogression. Magnetic resonance imaging protocols MRI was performed using a 3.0-T MR scanner (Verio, Siemens Healthcare) with an 8-channel head coil. The MRI sequence included T1WI (repetition time (TR), 250 ms, time of echo (TE), 2.67 ms, matrix, 320 × 256, slice thickness, 5 mm, and field of view (FOV), 230 mm × 230 mm), T2WI (TR, 4 900 ms, TE, 100 ms, matrix, 320 × 320, slice thickness, 5 mm, FOV, 230 mm × 230 mm), T2WI/FLAIR (TR, 8 000 ms, TE, 94 ms, matrix, 256 × 256, slice thickness, 5 mm, FOV, 230 mm × 230 mm, inversion time (TI), 2 370 ms), CE-T1WI (the parameters were consistent with those of T1WI). Gadolinium contrast agent (gadopentetate glucosamine, Gd-DTPA) was used for the enhancement sequence, and the dose was 0.2 mL/kg by intravenous injection with a flow rate of 3 mL/s. Qualitative analysis of magnetic resonance images Two radiologists (with 5 and 11 years of work experience, respectively) independently evaluated spatial recurrence patterns and radiological features. If there were any disputes, a consensus was reached after discussing with another radiologist (with 14 years of work experience). The spatial recurrence patterns assessment process was as follows: (1) The surgical region was evaluated using CE-MRI and CT at 24–72 h postoperatively, and changes in the surgical region and formation of the resection cavity were observed during the follow-up CE-MRI. (2) The spatial pattern of locally recurrent HGGs was assessed based on the distance between the location of the recurrent tumor and the surgical region or surgical resection cavity [ 3 ]. The spatial pattern of local recurrence was divided into ICR and ECR [ 7 ]. ICR was defined as recurrence in the resection cavity or the surgical region. ECR was defined as recurrence at the edge or within 2 cm of the resection cavity or surgical region. (3) For excessively extensive recurrent lesions, we determined the spatial recurrence pattern by sequential CE-MRI, which showed the tendency of recurrent lesions to grow into or out of the resection cavity. Ventricular entry was defined as intraoperative access to the ventricle and was confirmed after seeing that the ventricle was connected to the resection cavity on the postoperative MRI. SVZ involvement was defined as enhanced tumor contact with the lateral ventricular edge. Cortex infiltration was defined as enhanced tumor involvement in the cerebral cortex. Radiomics analysis The workflow of the radiomics analysis included image segmentation, radiomics feature extraction, feature selection, and model construction and evaluation (Fig. 1 a). Image segmentation All digital imaging and communication in medicine (DICOM) data were converted to the neuroimaging informatics technology initiative format using MRIcroGL ( https://www.nitrc.org/projects/mricrogl/ ). Initial segmentation was performed using an unsupervised automated segmentation method provided by ONCOhabitats [ 23 ] ( https://www.oncohabitats.upv.es/ ). The automated segmentation process of brain tumors was as follows: (1) Preprocessing: Images were corrected for magnetic bias field inhomogeneities, noise, or spike artifacts. Consistent multiparameter high-quality brain images were generated through skull stripping, automated registration, brain extraction, and intensity normalization, (2) Segmentation: A 3D Convolutional Neural Network classifier based on the U-Net architecture was used to segment the intratumoral region. Then, a radiologist (11 years of work experience) used ITK-SNAP software (version 3.8, www.itksnap.org ) to re-evaluate and validate the images and generate the final volume of interest. Radiomics feature extraction Before feature extraction, all images were resampled to a voxel size of 1.0 mm × 1.0 mm × 1.0 mm, and the bin width of the grey level histogram was fixed at 25 to discretize the images. This was done to avoid the influence of different MR reconstruction parameters. Feature extraction was performed using the pyradiomics within Python (version 3.7.0) [ 24 ]. The extracted radiomics features included the following seven classes: first order statistics, shape-based 3D and 2D, grey level cooccurrence matrix (GLCM), grey level run length matrix (GLRLM), grey level size zone matrix (GLSZM), neighboring gray tone difference matrix (NGTDM), and gray level dependence matrix (GLDM). A total of 2446 radiomic features were extracted from the T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and FLAIR, respectively. The features complied with feature definitions described by the Imaging Biomarker Standardization Initiative. Feature selection All radiomic features were normalized by Z-score normalization. The Pearson or Spearman correlation analysis and the least absolute shrinkage and selection operator (LASSO) method were used to select nonzero coefficients of ECR related features from the radiomics features of the training set. After that, a formula was constructed using the linear combination of features with nonzero coefficients weighted by their respective LASSO coefficients. The formula was used to calculate the radiomics score for each patient to reflect the risk of ECR, and the Mann-Whitney U test was used to compare the radiomics score for the two spatial recurrence patterns in the training and validation sets. Radiomics nomogram construction and evaluation The radiomics score were analyzed using logistic regression analysis to predict ECR in the training set. In addition, a radiomics nomogram was constructed, and a calibration curve was used to assess the calibration of the nomogram. The goodness-of-fit of the nomogram was assessed using the Hosmer-Lemeshow test, and the area under the receiver operating characteristic (ROC) curve (AUC) was calculated to quantify the predictive performance of the nomogram. The validation set was used for the internal validation of the nomogram. Calibration and Hosmer-Lemeshow tests were performed, and the AUC was calculated. Statistical analysis Statistical analyses were performed using the Statistical Package for the Social Sciences for Windows, version 26.0 (IBM SPSS Statistics, IBM Corp), R software, version 4.2.1 (R Foundation for Statistical Computing, Boston, MA, USA), and MedCalc software, version 20.116 (MedCalc Software Ltd, Belgium). The difference in continuous variables was first tested using the normal distribution and homogeneity of variance. If both were satisfied, the t -test was used; if not, the Mann-Whitney U test was used. The Chi-squared or Fisher’s exact test was used for categorical tests. Penalty parameter tuning in the LASSO logistic regression model was selected using minimum criteria with 10-fold cross-validation. The LASSO logistic regression model analysis was performed using the "glmnet" package, MedCalc software plotted the ROC curves, and the "car" package established the regression model. Logistic regression analysis to identify the predictors of spatial recurrence patterns. Radiomics nomogram construction and calibration plots were performed using the "rms" package. A two-tailed p value of < 0.05 was considered to be statistically significant. Results Patient characteristics A total of 121 patients with local recurrence (mean age, 50.84 ± 14.35 years, 72 men) were included after screening according to the inclusion and exclusion criteria (Fig. 1 b), comprising 54 patients with ICR (Fig. 2 a) and 67 patients with ECR (Fig. 2 b). The cohort was randomly divided into training ( n = 84, mean age, 50.31 ± 14.56 years, 49 men) and validation sets ( n = 37, mean age, 52.05 ± 13.95 years, 23 men) at a 7:3 ratio. Patients with ECR formed 56.0% (47/84) and 54.1% (20/37) of the training and validation sets, respectively, and there were no significant differences between the sets ( p = 0.847). Clinical information (age, sex, and preoperative Karnofsky Performance Scale score), pathological characteristics (WHO grade, isocitrate dehydrogenase mutation status, and Ki-67), and radiological features (tumor side, tumor location, ventricular entry, SVZ involvement, and cortex infiltrated) in the training and validation sets are summarized in Table 1 and Supplementary Table 1 . In the training and validation sets, there were no differences in patient characteristics between ICR and ECR, except for ventricular entry in the training sets ( p 0.05). Table 1 ༎ Patients’ clinical, pathological characteristics, and radiological features Training set p Validation set p ICR ( n = 37) ECR ( n = 47) ICR ( n = 17) ECR ( n = 20) Age*, years 50.46 ± 15.23 50.19 ± 14.18 0.934 a 53.00 ± 13.46 51.25 ± 14.66 0.709 a Sex 0.853 b 0.745 c Male 22(59.5) 27(57.4) 10(58.8) 13(65.0) Female 15(40.5) 20(42.6) 7(41.2) 7(35.0) Preoperative KPS score 0.960 b 0.157 c ≥70 25(67.6) 32(68.1) 9(52.9) 16(80.0) <70 12(32.4) 15(31.9) 8(47.1) 4(20.0) WHO grade 0.163 b 1.000 c Grade Ⅲ 12(32.4) 9(19.1) 6(35.3) 7(35.0) Grade Ⅳ 25(67.6) 38(80.9) 11(64.7) 13(65.0) IDH mutation status 0.930 c 0.531 c Mutant 2(5.4) 3(6.4) 2(11.8) 5(25.0) Wild-type 25(67.6) 33(70.2) 10(58.8) 12(60.0) Undetermined 10(27.0) 11(23.4) 5(29.4) 3(15.0) Ki-67 0.253 b 1.000 c High level (> 10%) 30(81.1) 33(70.2) 11(64.7) 14(70.0) Low level (≤ 10%) 7(18.9) 14(29.8) 6(35.3) 6(30.0) Tumor side 0.416 c 0.609 c Left 18(48.6) 18(38.3) 5(29.5) 8(40.0) Right 15(40.5) 19(40.4) 10(58.8) 8(40.0) Bilateral 4(10.8) 10(21.3) 2(11.8) 4(20.0) Tumor location 0.548 c 0.148 c Frontal 13(35.1) 21(44.7) 2(11.8) 9(45.0) Parietal 4(10.8) 9(19.1) 6(35.3) 4(20.0) Temporal 11(29.7) 11(23.4) 4(23.5) 2(10.0) Occipital 1(2.7) 1(2.1) 0(0) 1(5.0) Other 8(21.6) 5(10.6) 5(29.4) 4(20.0) Ventricular entry < 0.001 b 0.082 c No 33(89.2) 19(40.4) 14(82.4) 10(50.0) Yes 4(10.8) 28(59.6) 3(17.6) 10(50.0) SVZ involvement 0.451 b 1.000 c No 18(48.6) 19(40.4) 7(41.2) 8(40.0) Yes 19(51.4) 28(59.6) 10(58.8) 12(60.0) Cortex infiltrated 0.528 b 0.515 c No 14(37.8) 21(44.7) 7(41.2) 11(55.0) Yes 23(62.2) 26(55.3) 10(58.8) 9(45.0) Radiomics score, median (interquartile range) -0.030 (-0.226-0.248) 0.424 (0.278–0.573) < 0.001 d 0.277 (0.103–0.322) 0.369 (0.258–0.487) 0.033 d Note. Data are numbers of patients and data in parentheses are percentages. *Data are means ± standard deviation. a = Student’s t-test, b = Pearson chi squared test, c = Fisher’s exact test, d = Mann-Whitney U test. ICR = intra-resection cavity recurrence, ECR = extra-resection cavity recurrence, KPS = Karnofsky Performance Scale, IDH = isocitrate dehydrogenase, SVZ = subventricular zone. Feature selection and radiomics score The Pearson or Spearman correlation analysis and the LASSO logistic regression model was used in the training set to screen the seven features with nonzero coefficients related to spatial recurrence patterns (Figs. 3 a and 3 b). The weights of the nonzero coefficients and radiomics score calculation formula are shown in Supplementary Table 2 . Notably, there was a significant difference in the radiomics score between ICR and ECR ( p < 0.001), which was subsequently confirmed in the validation set. The radiomics score of patients with ECR were higher than those of patients with ICR (Figs. 3 c and 3 d). The radiomics score in the ICR and ECR patient groups in the training set were − 0.030 (-0.226-0.248) vs 0.424 (0.278–0.573), respectively, p < 0.001. The radiomics score in the ICR and ECR patient groups in the validation set were 0.277 (0.103–0.322) vs 0.369 (0.258–0.487), respectively, p = 0.033. The radiomics score of each patient in the training and validation sets is respectively shown in Figs. 3 e and 3 f. Construction and validation of radiomics nomogram for spatial recurrence patterns The logistic regression analysis results showed that radiomics score ( p < 0.001) were independent factors for predicting spatial recurrence patterns. Thus, we constructed a radiomics nomogram based on these findings (Fig. 4 a). The calibration curve and nonsignificant Hosmer-Lemeshow test ( p = 0.453) showed that the training set had good calibration (Fig. 4 b). Similarly, the predictive model showed performed well with an AUC of 0.844 (95%CI: 0.749–0.914, sensitivity: 100.00%; specificity: 59.46%) in the training set (Fig. 4 d). In particular, the good calibration and nonsignificant Hosmer-Lemeshow test ( p = 0.434) of the radiomics nomogram were validated by the validation set (Fig. 4 c). The AUC of the validation set was 0.706 (95%CI: 0.534–0.844, sensitivity: 70.00%; specificity: 76.47%) (Fig. 4 e). Discussion In this study, preoperative radiomics analysis was used to integrate radiomic features from T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and T2WI/FLAIR images to predict spatial patterns in locally recurrent HGGs. Moreover, we developed and validated a radiomics nomogram model for predicting spatial recurrence patterns of HGGs and found that the model had good calibration and predictive performance. Local recurrence is the most common pattern of recurrence in HGGs, accounting for approximately 80% of all recurrences. Petrecca et al. analyzed the recurrence patterns of 20 patients with glioblastoma after total resection plus standard treatment and found that 17 patients (17/20, 85%) had tumor recurrence at the resection margin [ 25 ]. Despite total resection of the enhanced tumor, it was observed that most recurrent tumors occur at the margin of surgical resection[ 26 ], suggesting that local recurrence still exhibits different spatial recurrence patterns. Therefore, in this study, we divided local recurrence into ICR and ECR based on the relationship between the recurrent tumor and the surgical area or surgical resection cavity. Wang et al. reported that spatial recurrence patterns (ICR and ECR) were associated with prognosis in patients with locally recurrent HGGs after maximum safety resection and chemoradiotherapy [ 7 ]. Moreover, developing individualized treatment plans based on recurrence patterns may improve the prognosis of patients with HGGs. Previous studies have reported that extended resection of FLAIR abnormalities in patients with glioblastoma could prolong survival [ 27 – 29 ]. Thus, if patients with high risk of ECR can be preoperatively predicted, expanding the extent of resection may benefit patient survival. Besides, if patients are at high risk of ICR, increasing the local radiotherapy dose may improve their prognosis. Preoperative non-invasive prediction of spatial recurrence patterns in patients with HGGs is important for clinical decision-making. Glioma recurrence is associated with tumor heterogeneity, which is manifested on imaging as enhanced tumor and T2WI/FLAIR hyperintense perilesional region [ 30 , 31 ]. However, it is not easy to accurately assess tumor aggressiveness using conventional imaging processing methods. Radiomics is an emerging field that transforms medical images into massive amounts of quantitative data to reveal disease information. Compared to traditional imaging methods, radiomics can provide objective and quantitative image information about the lesion, independent of the individual subjective judgment and experience of radiologist. Previous studies have shown that radiomics has high diagnostic performance in predicting the recurrence pattern of glioblastoma [ 21 , 22 ]. Among them, Shim et al. [ 22 ] established a model of dynamic sensitivity contrast-enhanced MRI using a radiomic-based neural network, and predicted the AUC of local recurrence to be 0.969. However, the study of a multimodal MRI radiomics model to predict the spatial recurrence pattern of local recurrence has not been reported. In this study, we investigated the spatial pattern of local recurrence and attempted to identify the value of radiomic features in predicting the spatial patterns of local recurrence. The integrated radiomics features of T1WI and T2WI/FLAIR can more comprehensively reflect the characteristics of the tumor microenvironment, predict tumor invasion and the recurrence patterns. In addition, as in previous studies, we segmented lesion habitats, including T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region, to better capture tumor heterogeneity [ 19 ]. Therefore, we developed a prediction of the spatial recurrence pattern based on radiomics score of enhance lesions and T2WI/FLAIR hyperintense perilesional region. This study demonstrated that in both the training and validation sets, the radiomics score for ECR was higher than for ICR and the radiomics score showed a good predictive performance. The radiomics score consisted of 7 radiomics features, including two first-order statistical and five texture features. Among these features, the one with the highest weight coefficient was GLCM, followed by GLSZM, both of which are texture features. Texture features detect potential tumor infiltration by reflecting the heterogeneity of gliomas [ 30 ]. This indicated that the preoperative tumor heterogeneity of ECR was higher than that of ICR, and that ECR tumors have a higher potential for outward invasion before surgery. We then used logistic regression analysis to construct a radiomics nomogram. The predictive performance of the constructed radiomics nomogram in the training and validation sets were good. The radiomics analysis of multimodal MRI based on lesion habitat can reflect the preoperative tumor heterogeneity of HGGs with different spatial recurrence patterns, and predict ECR and ICR. Limitations This study has some limitations. First, this study was based on conventional MRI, and future studies may apply more advanced MR sequences to improve the model performance. Second, the model has only been validated internally and lacks external validation. In order to promote the clinical application of our model, it is necessary to conduct multi-center studies with larger sample sizes for validation. Finally, we use logistic regression to construct predictive model, while other machine learning methods, such as support vector machine, K-nearest neighbor classifier, and Gaussian naive Bayes can also be used to filter the best model. In the future, more machine learning methods need to be established to explore the optimal radiomics model. Conclusion Our study showed that radiomic features extracted from T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and FLAIR images can be used noninvasively to preoperatively predict spatial patterns in locally recurrent HGGs, providing a basis for determining personalized treatment strategies for HGGs. Abbreviations AUC Area under the receiver operating characteristic curve CE Contrast-enhanced ECR Extra-resection cavity recurrence FOV Field of view Gd-DTPA Gadopentetate glucosamine, HGGs High-grade gliomas ICR Intra-resection cavity recurrence MRI Magnetic resonance imaging RANO Response Assessment in Neuro-Oncology ROI Regions of interest SVZ Subventricular zone T2WI/FLAIR Fluid-attenuated inversion recovery T1WI T1-weighted imaging TE Time of echo TI Inversion time TR Repetition time Declarations Funding: This work was supported by Chongqing medical scientific research project (No.2023MSXM009), by the National Natural Science Foundation of China (No.81701661) and by Science and Technology planning project of Chongqing Clinical Research Centre of Imaging and Nuclear Medicine (No. CSTC2015YFPT-gcjsyjzx0175). Declaration of Competing Interest: The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Author Contributions All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Han-wei Wang, Lin-lan Zeng, Xiao-guang Li, Mi-mi Zhao, Xuan Li, Ling Feng, Ping Xiang, Li-zhao Chen, Jing Tian, Qi-sheng Ran. The first draft of the manuscript was written by Han-wei Wang, Lin-lan Zeng. Liang Yi, Shu-nan Wang revised the manuscript critically for important intellectual content and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Data Availability The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Ethics approval This study was performed in line with the principles of the Declaration of Helsinki. Approved by the institutional review board of the Army Medical Center of the PLA, and requirement of written informed consent was waived (Date November 9, 2022 /No.2022-325). Consent to publish The authors affirm that human research participants provided informed consent for publication of the images in Figure(s) 1a, 2a and 2b. Acknowledgements : None. References Tan AC, Ashley DM, López GY, Malinzak M, Friedman HS, Khasraw M (2020) Management of glioblastoma: State of the art and future directions. CA: a cancer journal for clinicians 70: 299-312 doi:10.3322/caac.21613 McKinnon C, Nandhabalan M, Murray SA, Plaha P (2021) Glioblastoma: clinical presentation, diagnosis, and management. BMJ (Clinical research ed) 374: n1560 doi:10.1136/bmj.n1560 Konishi Y, Muragaki Y, Iseki H, Mitsuhashi N, Okada Y (2012) Patterns of intracranial glioblastoma recurrence after aggressive surgical resection and adjuvant management: retrospective analysis of 43 cases. Neurologia medico-chirurgica 52: 577-586 doi:10.2176/nmc.52.577 Piper RJ, Senthil KK, Yan JL, Price SJ (2018) Neuroimaging classification of progression patterns in glioblastoma: a systematic review. Journal of neuro-oncology 139: 77-88 doi:10.1007/s11060-018-2843-3 Ge S, Shi Y, Zhu G, Li S, Cai Y, Ji P, Liu J, Guo W, Gong L, Lou M, Feng F, Wang Y, Zhai Y, Qu Y, Wang L (2020) Molecular Pathological Markers Correlated With the Recurrence Patterns of Glioma. Frontiers in oncology 10: 565045 doi:10.3389/fonc.2020.565045 Cho SJ, Kim HS, Suh CH, Park JE (2020) Radiological Recurrence Patterns after Bevacizumab Treatment of Recurrent High-Grade Glioma: A Systematic Review and Meta-Analysis. Korean journal of radiology 21: 908-918 doi:10.3348/kjr.2019.0898 Wang H, Zeng L, Wu H, Tian J, Xie H, Zhang L, Ran Q, Zhong P, Chen L, Yi L, Wang S (2023) Preoperative vascular heterogeneity based on dynamic susceptibility contrast MRI in predicting spatial pattern of locally recurrent high-grade gliomas. European radiology doi:10.1007/s00330-023-10149-6 Ringel F, Pape H, Sabel M, Krex D, Bock HC, Misch M, Weyerbrock A, Westermaier T, Senft C, Schucht P, Meyer B, Simon M (2016) Clinical benefit from resection of recurrent glioblastomas: results of a multicenter study including 503 patients with recurrent glioblastomas undergoing surgical resection. Neuro-oncology 18: 96-104 doi:10.1093/neuonc/nov145 Shibahara I, Sonoda Y, Saito R, Kanamori M, Yamashita Y, Kumabe T, Watanabe M, Suzuki H, Watanabe T, Ishioka C, Tominaga T (2013) The expression status of CD133 is associated with the pattern and timing of primary glioblastoma recurrence. Neuro-oncology 15: 1151-1159 doi:10.1093/neuonc/not066 Jungk C, Warta R, Mock A, Friauf S, Hug B, Capper D, Abdollahi A, Debus J, Bendszus M, von Deimling A, Unterberg A, Herold-Mende C (2019) Location-Dependent Patient Outcome and Recurrence Patterns in IDH1-Wildtype Glioblastoma. Cancers 11 doi:10.3390/cancers11010122 Yoon N, Kim HS, Lee JW, Lee EJ, Maeng LS, Yoon WS (2020) Targeted Genomic Sequencing Reveals Different Evolutionary Patterns Between Locally and Distally Recurrent Glioblastomas. Cancer genomics & proteomics 17: 803-812 doi:10.21873/cgp.20234 Jiang H, Yu K, Li M, Cui Y, Ren X, Yang C, Zhao X, Lin S (2020) Classification of Progression Patterns in Glioblastoma: Analysis of Predictive Factors and Clinical Implications. Frontiers in oncology 10: 590648 doi:10.3389/fonc.2020.590648 Tang W, Wang X, Chen Y, Zhang J, Chen Y, Lin Z (2015) CXCL12 and CXCR4 as predictive biomarkers of glioma recurrence pattern after total resection. Pathologie-biologie 63: 190-198 doi:10.1016/j.patbio.2015.07.002 Mistry AM, Kelly PD, Gallant JN, Mummareddy N, Mobley BC, Thompson RC, Chambless LB (2019) Comparative Analysis of Subventricular Zone Glioblastoma Contact and Ventricular Entry During Resection in Predicting Dissemination, Hydrocephalus, and Survival. Neurosurgery 85: E924-e932 doi:10.1093/neuros/nyz144 Comas S, Luguera E, Molero J, Balaña C, Estival A, Castañer S, Carrato C, Hostalot C, Teixidor P, Villà S (2021) Influence of glioblastoma contact with the subventricular zone on survival and recurrence patterns. Clinical & translational oncology : official publication of the Federation of Spanish Oncology Societies and of the National Cancer Institute of Mexico 23: 554-564 doi:10.1007/s12094-020-02448-x Li M, Huang W, Chen H, Jiang H, Yang C, Shen S, Cui Y, Dong G, Ren X, Lin S (2022) T2/FLAIR Abnormity Could be the Sign of Glioblastoma Dissemination. Frontiers in neurology 13: 819216 doi:10.3389/fneur.2022.819216 Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout RG, Granton P, Zegers CM, Gillies R, Boellard R, Dekker A, Aerts HJ (2012) Radiomics: extracting more information from medical images using advanced feature analysis. European journal of cancer (Oxford, England : 1990) 48: 441-446 doi:10.1016/j.ejca.2011.11.036 Aerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S, Bussink J, Monshouwer R, Haibe-Kains B, Rietveld D, Hoebers F, Rietbergen MM, Leemans CR, Dekker A, Quackenbush J, Gillies RJ, Lambin P (2014) Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nature communications 5: 4006 doi:10.1038/ncomms5006 Ismail M, Hill V, Statsevych V, Huang R, Prasanna P, Correa R, Singh G, Bera K, Beig N, Thawani R, Madabhushi A, Aahluwalia M, Tiwari P (2018) Shape Features of the Lesion Habitat to Differentiate Brain Tumor Progression from Pseudoprogression on Routine Multiparametric MRI: A Multisite Study. AJNR American journal of neuroradiology 39: 2187-2193 doi:10.3174/ajnr.A5858 Wei J, Yang G, Hao X, Gu D, Tan Y, Wang X, Dong D, Zhang S, Wang L, Zhang H, Tian J (2019) A multi-sequence and habitat-based MRI radiomics signature for preoperative prediction of MGMT promoter methylation in astrocytomas with prognostic implication. European radiology 29: 877-888 doi:10.1007/s00330-018-5575-z Yan JL, Toh CH, Ko L, Wei KC, Chen PY (2021) A Neural Network Approach to Identify Glioblastoma Progression Phenotype from Multimodal MRI. Cancers 13 doi:10.3390/cancers13092006 Shim KY, Chung SW, Jeong JH, Hwang I, Park CK, Kim TM, Park SH, Won JK, Lee JH, Lee ST, Yoo RE, Kang KM, Yun TJ, Kim JH, Sohn CH, Choi KS, Choi SH (2021) Radiomics-based neural network predicts recurrence patterns in glioblastoma using dynamic susceptibility contrast-enhanced MRI. Scientific reports 11: 9974 doi:10.1038/s41598-021-89218-z Juan-Albarracín J, Fuster-Garcia E, Manjón JV, Robles M, Aparici F, Martí-Bonmatí L, García-Gómez JM (2015) Automated glioblastoma segmentation based on a multiparametric structured unsupervised classification. PloS one 10: e0125143 doi:10.1371/journal.pone.0125143 van Griethuysen JJM, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, Beets-Tan RGH, Fillion-Robin JC, Pieper S, Aerts H (2017) Computational Radiomics System to Decode the Radiographic Phenotype. Cancer research 77: e104-e107 doi:10.1158/0008-5472.Can-17-0339 Petrecca K, Guiot MC, Panet-Raymond V, Souhami L (2013) Failure pattern following complete resection plus radiotherapy and temozolomide is at the resection margin in patients with glioblastoma. Journal of neuro-oncology 111: 19-23 doi:10.1007/s11060-012-0983-4 Lemée JM, Clavreul A, Menei P (2015) Intratumoral heterogeneity in glioblastoma: don't forget the peritumoral brain zone. Neuro-oncology 17: 1322-1332 doi:10.1093/neuonc/nov119 Li YM, Suki D, Hess K, Sawaya R (2016) The influence of maximum safe resection of glioblastoma on survival in 1229 patients: Can we do better than gross-total resection? Journal of neurosurgery 124: 977-988 doi:10.3171/2015.5.Jns142087 Pessina F, Navarria P, Cozzi L, Ascolese AM, Simonelli M, Santoro A, Clerici E, Rossi M, Scorsetti M, Bello L (2017) Maximize surgical resection beyond contrast-enhancing boundaries in newly diagnosed glioblastoma multiforme: is it useful and safe? A single institution retrospective experience. Journal of neuro-oncology 135: 129-139 doi:10.1007/s11060-017-2559-9 Lu M, Fu ZH, He XJ, Lu JK, Deng XQ, Lin DL, Gu YM, Fan YF, Lai MY, Li J, Yang MM, Chen ZP (2020) T2 Fluid-Attenuated Inversion Recovery Resection for Glioblastoma Involving Eloquent Brain Areas Facilitated Through Awake Craniotomy and Clinical Outcome. World neurosurgery 135: e738-e747 doi:10.1016/j.wneu.2019.12.130 Hu LS, Hawkins-Daarud A, Wang L, Li J, Swanson KR (2020) Imaging of intratumoral heterogeneity in high-grade glioma. Cancer letters 477: 97-106 doi:10.1016/j.canlet.2020.02.025 Lemée JM, Clavreul A, Aubry M, Com E, de Tayrac M, Eliat PA, Henry C, Rousseau A, Mosser J, Menei P (2015) Characterizing the peritumoral brain zone in glioblastoma: a multidisciplinary analysis. Journal of neuro-oncology 122: 53-61 doi:10.1007/s11060-014-1695-8 Additional Declarations No competing interests reported. Supplementary Files 5Supplementarymaterial.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3870027","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":267776801,"identity":"ceb79a59-fdc6-4981-8e6c-76b84eb16c5c","order_by":0,"name":"Han-wei Wang","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Han-wei","middleName":"","lastName":"Wang","suffix":""},{"id":267776802,"identity":"a4113cc2-1744-4318-bc11-5f2857f2555e","order_by":1,"name":"Lin-lan Zeng","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Lin-lan","middleName":"","lastName":"Zeng","suffix":""},{"id":267776803,"identity":"a492f517-c1a5-4964-9a4f-bfeab5085a2e","order_by":2,"name":"Xiao-guang Li","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiao-guang","middleName":"","lastName":"Li","suffix":""},{"id":267776804,"identity":"5ef65b4c-aa35-43ed-99b2-c7240db3d8bc","order_by":3,"name":"Mi-mi Zhao","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Mi-mi","middleName":"","lastName":"Zhao","suffix":""},{"id":267776805,"identity":"e69b933a-2b66-4d7e-b0ea-68ec6f221710","order_by":4,"name":"Xuan Li","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xuan","middleName":"","lastName":"Li","suffix":""},{"id":267776806,"identity":"89ccd0ff-7865-4fe6-a7b2-f23c053c09be","order_by":5,"name":"Ling Feng","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ling","middleName":"","lastName":"Feng","suffix":""},{"id":267776807,"identity":"2e6fa0be-7c62-4bed-b3f2-fb824ba41d81","order_by":6,"name":"Ping Xiang","email":"","orcid":"","institution":"958 Hospital of Army, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Ping","middleName":"","lastName":"Xiang","suffix":""},{"id":267776808,"identity":"2b1ee068-627f-40da-b03b-5353c13a6dd4","order_by":7,"name":"Li-zhao Chen","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Li-zhao","middleName":"","lastName":"Chen","suffix":""},{"id":267776809,"identity":"0b6fb346-5f4c-4b60-a923-ab1ff1931f88","order_by":8,"name":"Jing Tian","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Tian","suffix":""},{"id":267776810,"identity":"22b486fe-2b28-4bb7-92b4-968393d895e9","order_by":9,"name":"Qi-sheng Ran","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Qi-sheng","middleName":"","lastName":"Ran","suffix":""},{"id":267776811,"identity":"409ac2f0-062c-47d7-a43b-9e7a9bc0f19f","order_by":10,"name":"Liang Yi","email":"","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liang","middleName":"","lastName":"Yi","suffix":""},{"id":267776812,"identity":"0f6c9a17-71ea-417c-b104-6d6578f13a70","order_by":11,"name":"Shu-nan Wang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYLACxgYQyXwAyk0gWgtbYgOpWngMidNicPzs4Zc/dxzO45/d8/3RzZzDDPzsOQYMP3fg0XImL82a98zhYok7Zzc25247zCDZ88aAsfcMbi1mB3LMjBnbDic23MiFaDG4kWPAzNiGR8v5N2aGP4Fa5t/IeQjWYk9Qy40c4we8QC0bbuQwQmyRIKDF/sYbM2betvTEjTfSDGfnbkvnkTjzrOBgLx4tkv05xh9/tlknzruR/OBz7jZrOf725I0PfuLRAgRsEsg8HhBxAK8GYEL5QEDBKBgFo2AUjHQAAEbgWza44acQAAAAAElFTkSuQmCC","orcid":"","institution":"Daping Hospital, Army Medical University","correspondingAuthor":true,"prefix":"","firstName":"Shu-nan","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2024-01-16 14:00:00","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3870027/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3870027/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":49893517,"identity":"94d5a020-5b03-4cfe-a523-03934e271a43","added_by":"auto","created_at":"2024-01-19 21:15:23","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":952313,"visible":true,"origin":"","legend":"\u003cp\u003e(\u003cstrong\u003ea\u003c/strong\u003e) Workflow of radiomics analysis.\u003cstrong\u003e (b)\u003c/strong\u003e Flow chart of patient selection. ROC = receiver operating characteristic, HGGs = high-grade gliomas, RANO = Response Assessment in Neuro-Oncology, ICR = intra-resection cavity recurrence, ECR = extra-resection cavity recurrence.\u003c/p\u003e","description":"","filename":"figure1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/d70fc4a480addf16f0a7d7ee.jpg"},{"id":49893522,"identity":"ff1303ce-99e7-4ad5-8141-12ca5d76f788","added_by":"auto","created_at":"2024-01-19 21:15:23","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":715590,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003e(a)\u003c/strong\u003e MRI scans in 29-year-old woman with glioblastoma, WHO grade Ⅳ, with tumor recurrence in the intra-resection cavity. (a) contrast-enhanced T1-weighted imaging (CE-T1WI) showed tumor in the right frontal parietal lobe (red dashed line). (b) 24 - 72 h postoperatively MRI. CE-T1WI displayed gross total resection of the tumor. (c) Follow-up MRI at 7 months. CE-T1WI showed the tumor recurrence in the intra-resection cavity. \u003cstrong\u003e(b) \u003c/strong\u003eMRI scans in a 49-year-old woman with glioblastoma, WHO grade Ⅳ, with tumor recurrence in the extra-resection cavity. (a) contrast-enhanced T1-weighted imaging (CE-T1WI) showed tumor in the right frontal lobe, which grew across the cerebral falx to contralateral side (red dashed line). (b) 24 - 72 h postoperatively MRI. CE-T1WI displayed gross total resection of the tumor and a little enhancement in the primary tumor area. (c) Follow-up MRI at 4 months. CE-T1WI showed new nodular enhancement outside the resection cavity (white arrow).\u003c/p\u003e","description":"","filename":"figure2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/2fcf99f0c5893610d54da621.jpg"},{"id":49893891,"identity":"7eb4a234-5fca-4ebe-952f-4cf3df92f70f","added_by":"auto","created_at":"2024-01-19 21:23:23","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":868418,"visible":true,"origin":"","legend":"\u003cp\u003eRadiomic features selection and the performance assessment. (a) Selection of tuning parameter (λ). The tuning parameter (λ) in LASSO method for 10-fold cross validation based on minimum criteria. Binomial deviations were plotted as a function of logarithm (λ). According to the minimum criteria, the calculated optimal value is plotted as a dashed line. The optimal λ value of 0.089 with log(λ) of -2.409 was selected. (b) LASSO coefficient profiles of the radiomics feature. According to the dash line plotted at the optimal λ value, seven feature with nonzero coefficients were selected. (c) and (d) showed boxplot of radiomics score in the training and validation sets, respectively. (e) and (f) showed waterfall plot of radiomics score for each patient in the training and validation sets, respectively.\u003c/p\u003e","description":"","filename":"figure3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/c64aed6550124b38d6fed81e.jpg"},{"id":49893518,"identity":"3752e8b5-de77-43bd-97cf-97c7a6475422","added_by":"auto","created_at":"2024-01-19 21:15:23","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":789400,"visible":true,"origin":"","legend":"\u003cp\u003eThe radiomics nomogram and its performance. (a) The radiomics nomogram was constructed to predict spatial recurrence patterns of HGGs. (b) and (c) showed the calibration curves of the nomogram in training and validation sets, respectively. The x-axis represents the nomogram predicted probability and y-axis represents the actual probability of extra-resection cavity recurrence. The diagonal dashed line represents a perfect prediction. The dotted line represents the entire cohort, and the solid line is bias-corrected by bootstrapping. The closer the solid line is to the diagonal dashed line; the better predictive accuracy of the nomogram is. (d) and (e) indicated the ROC curves of the nomogram in training and validation sets, respectively.\u003c/p\u003e","description":"","filename":"figure4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/176e3bf4f0701e230a69da9f.jpg"},{"id":50273812,"identity":"f40f4ce5-fdfe-47ac-9931-0de36408c21e","added_by":"auto","created_at":"2024-01-28 21:22:16","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":893197,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/ff35df11-6da2-49fc-85b3-657f5a1d423e.pdf"},{"id":49893520,"identity":"7da52e0d-2d52-483a-9636-984036632e64","added_by":"auto","created_at":"2024-01-19 21:15:23","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":45230,"visible":true,"origin":"","legend":"","description":"","filename":"5Supplementarymaterial.docx","url":"https://assets-eu.researchsquare.com/files/rs-3870027/v1/90a749d343d6260805f9c659.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Multimodal MRI lesion habitat-based radiomics analysis for preoperative prediction of spatial pattern in locally recurrent high-grade gliomas","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHigh-grade gliomas (HGGs, WHO grade Ⅲ-Ⅳ) are highly heterogeneous tumors of the central nervous system. Despite standardized treatments, the prognosis of HGGs patients has not improved, with more than half of patients relapsing within 6\u0026ndash;8 months after surgery [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. The high recurrence rate is a critical factor affecting the prognosis of the patients.\u003c/p\u003e \u003cp\u003eRecurrence of HGGs most commonly occurs locally within 2\u0026ndash;3 cm of the resection cavity, and in a few cases far from the original resection cavity or surgical area [\u003cspan additionalcitationids=\"CR4 CR5\" citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. What\u0026rsquo;s more, locally recurrent HGGs exhibits different spatial patterns. Based on postoperative magnetic resonance imaging (MRI), local recurrence can be divided into intra-resection cavity recurrence (ICR) and extra-resection cavity recurrence (ECR), according to the distances between the recurrent tumor and the resection cavity [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Wang et al. analyzed that the spatial recurrence patterns and prognosis in 69 patients with locally recurrent HGGs and revealed patients with ICR had longer prognosis than patients with ECR [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Currently, the treatment of recurrent HGGs is dominated by individualized therapy, and it has been reported that surgical resection of recurrent glioblastoma may help to prolong patients\u0026rsquo; overall survival [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Thus, early prediction of the spatial recurrence patterns may help to individualize patient treatment and improve prognosis. Previous studies have shown that molecular pathological markers and clinical factors were related to spatial recurrence patterns [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e, \u003cspan additionalcitationids=\"CR10 CR11 CR12\" citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. However, there is a lack of imaging markers for preoperative non-invasive prediction of spatial recurrence patterns of HGGs.\u003c/p\u003e \u003cp\u003eMRI is the main non-invasive imaging tool for the HGGs management, which plays an important role in diagnosis, efficacy evaluation and follow-up. Previous studies have shown that certain MRI features, such as subventricular zone (SVZ) involvement, ventricular entry, and T2-weighted-fluid-attenuated inversion recovery (T2WI/FLAIR) abnormal transformation, are associated with recurrence patterns [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan additionalcitationids=\"CR15\" citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. However, subjective judgment by radiologists is susceptible to interobserver variability, which affects the accuracy and reliability of the assessment results. In contrast, radiomics overcomes this problem by high-throughput extraction and analysis of quantitative image features from many medical images to correlate with clinical outcomes [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. In addition, radiomics can more accurately reflect tumor heterogeneity by segmenting tumor subregion regions of interest (ROI). Ismail et al. [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e] segmented the lesion habitat and found that 3D shape features of the enhancing lesion on T1WI, and T2WI/FLAIR hyperintensities could distinguish tumor progression and pseudoprogression. Habitat-based MRI radiomics have also been shown to predict MGMT promoter methylation and prognosis in astrocytoma [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. To the best of our knowledge, a few studies have reported on the prediction of recurrence patterns in HGGs based on radiomic features [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. These studies suggest that habitat-based radiomics may be equally applicable to preoperative prediction of spatial recurrence patterns in HGGs.\u003c/p\u003e \u003cp\u003eTherefore, this study aims to preoperatively predict spatial patterns in locally recurrent high-grade gliomas (HGGs) based on lesion habitat radiomics analysis of multimodal magnetic resonance imaging (MRI) and to evaluate the predictive performance of this approach.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003ePatients\u003c/h2\u003e\n \u003cp\u003eThis study was performed in line with the principles of the Declaration of Helsinki. Approved by the institutional review board of the Army Medical Center of the PLA, and requirement of written informed consent was waived (Date November 9, 2022 /No.2022\u0026thinsp;\u0026minus;\u0026thinsp;325). We performed this study on patients with locally recurrent HGGs after maximum safe surgical resections and radiotherapy combined with temozolomide between January 2012 and October 2021.\u003c/p\u003e\n \u003cp\u003ePatients who met the following inclusion criteria were enrolled in our study: (1) had histopathological confirmation of newly diagnosed HGGs according to the 2016 WHO classification criteria without evidence of subarachnoid dissemination on MRI; (2) had no history of chemotherapy or radiotherapy treatment before preoperative imaging; (3) underwent preoperative MRI, including T1-weighted imaging (T1WI), T2-weighted imaging (T2WI), T2WI/FLAIR imaging, and contrast-enhanced (CE) T1WI; (4) underwent maximum safe surgical resection and radiotherapy combined with temozolomide; and (5) aged\u0026thinsp;\u0026gt;\u0026thinsp;18 years. Exclusion criteria were as follows: (1) infratentorial HGGs; (2) low-quality MR images, including noise and artifacts; (3) loss of follow-up; (4) no recurrence during the follow-up; and (5) non-locally recurrent HGGs.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003eFollow-up\u003c/h2\u003e\n \u003cp\u003eAll patients underwent CE-MRI or CT or both 24\u0026ndash;72 h postoperatively and were followed up with CE-MRI every 3\u0026ndash;6 months. The patients were followed until death or the study cutoff time (October 2023). The evidence of tumor recurrence included (1) patients who underwent reoperation to confirm recurrence, (2) continuous CE-MRI according to the Response Assessment in Neuro-Oncology criteria, or (3) multimodal MRI to rule out radiation necrosis or pseudoprogression.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003eMagnetic resonance imaging protocols\u003c/h2\u003e\n \u003cp\u003eMRI was performed using a 3.0-T MR scanner (Verio, Siemens Healthcare) with an 8-channel head coil. The MRI sequence included T1WI (repetition time (TR), 250 ms, time of echo (TE), 2.67 ms, matrix, 320 \u0026times; 256, slice thickness, 5 mm, and field of view (FOV), 230 mm \u0026times; 230 mm), T2WI (TR, 4 900 ms, TE, 100 ms, matrix, 320 \u0026times; 320, slice thickness, 5 mm, FOV, 230 mm \u0026times; 230 mm), T2WI/FLAIR (TR, 8 000 ms, TE, 94 ms, matrix, 256 \u0026times; 256, slice thickness, 5 mm, FOV, 230 mm \u0026times; 230 mm, inversion time (TI), 2 370 ms), CE-T1WI (the parameters were consistent with those of T1WI). Gadolinium contrast agent (gadopentetate glucosamine, Gd-DTPA) was used for the enhancement sequence, and the dose was 0.2 mL/kg by intravenous injection with a flow rate of 3 mL/s.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003eQualitative analysis of magnetic resonance images\u003c/h2\u003e\n \u003cp\u003eTwo radiologists (with 5 and 11 years of work experience, respectively) independently evaluated spatial recurrence patterns and radiological features. If there were any disputes, a consensus was reached after discussing with another radiologist (with 14 years of work experience).\u003c/p\u003e\n \u003cp\u003eThe spatial recurrence patterns assessment process was as follows: (1) The surgical region was evaluated using CE-MRI and CT at 24\u0026ndash;72 h postoperatively, and changes in the surgical region and formation of the resection cavity were observed during the follow-up CE-MRI. (2) The spatial pattern of locally recurrent HGGs was assessed based on the distance between the location of the recurrent tumor and the surgical region or surgical resection cavity [\u003cspan class=\"CitationRef\"\u003e3\u003c/span\u003e]. The spatial pattern of local recurrence was divided into ICR and ECR [\u003cspan class=\"CitationRef\"\u003e7\u003c/span\u003e]. ICR was defined as recurrence in the resection cavity or the surgical region. ECR was defined as recurrence at the edge or within 2 cm of the resection cavity or surgical region. (3) For excessively extensive recurrent lesions, we determined the spatial recurrence pattern by sequential CE-MRI, which showed the tendency of recurrent lesions to grow into or out of the resection cavity.\u003c/p\u003e\n \u003cp\u003eVentricular entry was defined as intraoperative access to the ventricle and was confirmed after seeing that the ventricle was connected to the resection cavity on the postoperative MRI. SVZ involvement was defined as enhanced tumor contact with the lateral ventricular edge. Cortex infiltration was defined as enhanced tumor involvement in the cerebral cortex.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\n \u003ch2\u003eRadiomics analysis\u003c/h2\u003e\n \u003cp\u003eThe workflow of the radiomics analysis included image segmentation, radiomics feature extraction, feature selection, and model construction and evaluation (Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea).\u003c/p\u003e\n \u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003eImage segmentation\u003c/h2\u003e\n \u003cp\u003eAll digital imaging and communication in medicine (DICOM) data were converted to the neuroimaging informatics technology initiative format using MRIcroGL (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.nitrc.org/projects/mricrogl/\u003c/span\u003e\u003c/span\u003e). Initial segmentation was performed using an unsupervised automated segmentation method provided by ONCOhabitats [\u003cspan class=\"CitationRef\"\u003e23\u003c/span\u003e] (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.oncohabitats.upv.es/\u003c/span\u003e\u003c/span\u003e). The automated segmentation process of brain tumors was as follows: (1) Preprocessing: Images were corrected for magnetic bias field inhomogeneities, noise, or spike artifacts. Consistent multiparameter high-quality brain images were generated through skull stripping, automated registration, brain extraction, and intensity normalization, (2) Segmentation: A 3D Convolutional Neural Network classifier based on the U-Net architecture was used to segment the intratumoral region. Then, a radiologist (11 years of work experience) used ITK-SNAP software (version 3.8, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ewww.itksnap.org\u003c/span\u003e\u003c/span\u003e) to re-evaluate and validate the images and generate the final volume of interest.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003eRadiomics feature extraction\u003c/h2\u003e\n \u003cp\u003eBefore feature extraction, all images were resampled to a voxel size of 1.0 mm \u0026times; 1.0 mm \u0026times; 1.0 mm, and the bin width of the grey level histogram was fixed at 25 to discretize the images. This was done to avoid the influence of different MR reconstruction parameters. Feature extraction was performed using the pyradiomics within Python (version 3.7.0) [\u003cspan class=\"CitationRef\"\u003e24\u003c/span\u003e]. The extracted radiomics features included the following seven classes: first order statistics, shape-based 3D and 2D, grey level cooccurrence matrix (GLCM), grey level run length matrix (GLRLM), grey level size zone matrix (GLSZM), neighboring gray tone difference matrix (NGTDM), and gray level dependence matrix (GLDM). A total of 2446 radiomic features were extracted from the T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and FLAIR, respectively. The features complied with feature definitions described by the Imaging Biomarker Standardization Initiative.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003eFeature selection\u003c/h2\u003e\n \u003cp\u003eAll radiomic features were normalized by Z-score normalization. The Pearson or Spearman correlation analysis and the least absolute shrinkage and selection operator (LASSO) method were used to select nonzero coefficients of ECR related features from the radiomics features of the training set. After that, a formula was constructed using the linear combination of features with nonzero coefficients weighted by their respective LASSO coefficients. The formula was used to calculate the radiomics score for each patient to reflect the risk of ECR, and the Mann-Whitney \u003cem\u003eU\u003c/em\u003e test was used to compare the radiomics score for the two spatial recurrence patterns in the training and validation sets.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\n \u003ch2\u003eRadiomics nomogram construction and evaluation\u003c/h2\u003e\n \u003cp\u003eThe radiomics score were analyzed using logistic regression analysis to predict ECR in the training set. In addition, a radiomics nomogram was constructed, and a calibration curve was used to assess the calibration of the nomogram. The goodness-of-fit of the nomogram was assessed using the Hosmer-Lemeshow test, and the area under the receiver operating characteristic (ROC) curve (AUC) was calculated to quantify the predictive performance of the nomogram. The validation set was used for the internal validation of the nomogram. Calibration and Hosmer-Lemeshow tests were performed, and the AUC was calculated.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003eStatistical analysis\u003c/h2\u003e\n \u003cp\u003eStatistical analyses were performed using the Statistical Package for the Social Sciences for Windows, version 26.0 (IBM SPSS Statistics, IBM Corp), R software, version 4.2.1 (R Foundation for Statistical Computing, Boston, MA, USA), and MedCalc software, version 20.116 (MedCalc Software Ltd, Belgium).\u003c/p\u003e\n \u003cp\u003eThe difference in continuous variables was first tested using the normal distribution and homogeneity of variance. If both were satisfied, the \u003cem\u003et\u003c/em\u003e-test was used; if not, the Mann-Whitney \u003cem\u003eU\u003c/em\u003e test was used. The Chi-squared or Fisher\u0026rsquo;s exact test was used for categorical tests. Penalty parameter tuning in the LASSO logistic regression model was selected using minimum criteria with 10-fold cross-validation. The LASSO logistic regression model analysis was performed using the \u0026quot;glmnet\u0026quot; package, MedCalc software plotted the ROC curves, and the \u0026quot;car\u0026quot; package established the regression model. Logistic regression analysis to identify the predictors of spatial recurrence patterns. Radiomics nomogram construction and calibration plots were performed using the \u0026quot;rms\u0026quot; package. A two-tailed \u003cem\u003ep\u003c/em\u003e value of \u0026lt;\u0026thinsp;0.05 was considered to be statistically significant.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003ePatient characteristics\u003c/h2\u003e \u003cp\u003eA total of 121 patients with local recurrence (mean age, 50.84\u0026thinsp;\u0026plusmn;\u0026thinsp;14.35 years, 72 men) were included after screening according to the inclusion and exclusion criteria (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003eb), comprising 54 patients with ICR (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea) and 67 patients with ECR (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The cohort was randomly divided into training (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;84, mean age, 50.31\u0026thinsp;\u0026plusmn;\u0026thinsp;14.56 years, 49 men) and validation sets (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;37, mean age, 52.05\u0026thinsp;\u0026plusmn;\u0026thinsp;13.95 years, 23 men) at a 7:3 ratio. Patients with ECR formed 56.0% (47/84) and 54.1% (20/37) of the training and validation sets, respectively, and there were no significant differences between the sets (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.847).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eClinical information (age, sex, and preoperative Karnofsky Performance Scale score), pathological characteristics (WHO grade, isocitrate dehydrogenase mutation status, and Ki-67), and radiological features (tumor side, tumor location, ventricular entry, SVZ involvement, and cortex infiltrated) in the training and validation sets are summarized in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e \u003cb\u003eand Supplementary Table\u0026nbsp;1\u003c/b\u003e. In the training and validation sets, there were no differences in patient characteristics between ICR and ECR, except for ventricular entry in the training sets (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.05). In addition, there were no statistically significant differences in patient characteristics between the training and validation sets (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003e\u003cb\u003e༎\u003c/b\u003ePatients\u0026rsquo; clinical, pathological characteristics, and radiological features\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c3\" namest=\"c2\"\u003e \u003cp\u003eTraining set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eValidation set\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003e\u003cem\u003ep\u003c/em\u003e\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eICR\u003c/p\u003e \u003cp\u003e(\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;37)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eECR\u003c/p\u003e \u003cp\u003e(\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;47)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eICR\u003c/p\u003e \u003cp\u003e(\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;17)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eECR\u003c/p\u003e \u003cp\u003e(\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;20)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge*, years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e50.46\u0026thinsp;\u0026plusmn;\u0026thinsp;15.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e50.19\u0026thinsp;\u0026plusmn;\u0026thinsp;14.18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.934\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e53.00\u0026thinsp;\u0026plusmn;\u0026thinsp;13.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e51.25\u0026thinsp;\u0026plusmn;\u0026thinsp;14.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.709\u003csup\u003ea\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSex\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.853\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.745\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e22(59.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27(57.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10(58.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13(65.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15(40.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e20(42.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7(41.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7(35.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePreoperative KPS score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.960\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.157\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026ge;70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25(67.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e32(68.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e9(52.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e16(80.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u0026lt;70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12(32.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e15(31.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e8(47.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4(20.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWHO grade\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.163\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade Ⅲ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e12(32.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9(19.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6(35.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e7(35.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGrade Ⅳ\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25(67.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e38(80.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11(64.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e13(65.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIDH mutation status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.930\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.531\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMutant\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2(5.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e3(6.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2(11.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e5(25.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWild-type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25(67.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33(70.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10(58.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12(60.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUndetermined\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e10(27.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11(23.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5(29.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e3(15.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKi-67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.253\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHigh level (\u0026gt;\u0026thinsp;10%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e30(81.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33(70.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e11(64.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e14(70.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLow level (\u0026le;\u0026thinsp;10%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e7(18.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14(29.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6(35.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e6(30.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor side\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.416\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.609\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLeft\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18(48.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18(38.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5(29.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8(40.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRight\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15(40.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(40.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10(58.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8(40.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBilateral\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4(10.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e10(21.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2(11.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4(20.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTumor location\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.548\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.148\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFrontal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e13(35.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21(44.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e2(11.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9(45.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eParietal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4(10.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e9(19.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e6(35.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4(20.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTemporal\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11(29.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e11(23.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e4(23.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e2(10.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOccipital\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1(2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1(2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0(0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e1(5.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8(21.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e5(10.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e5(29.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e4(20.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVentricular entry\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.082\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e33(89.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(40.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e14(82.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10(50.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4(10.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e28(59.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e3(17.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e10(50.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSVZ involvement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.451\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.000\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e18(48.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19(40.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7(41.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e8(40.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e19(51.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e28(59.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10(58.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e12(60.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCortex infiltrated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.528\u003csup\u003eb\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.515\u003csup\u003ec\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e14(37.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e21(44.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e7(41.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e11(55.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e23(62.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e26(55.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e10(58.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e9(45.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRadiomics score, median (interquartile range)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e-0.030 (-0.226-0.248)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.424 (0.278\u0026ndash;0.573)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;0.001\u003c/b\u003e\u003csup\u003e\u003cb\u003ed\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.277 (0.103\u0026ndash;0.322)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.369 (0.258\u0026ndash;0.487)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u003cb\u003e0.033\u003c/b\u003e\u003csup\u003e\u003cb\u003ed\u003c/b\u003e\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003eNote. Data are numbers of patients and data in parentheses are percentages. *Data are means\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation. a\u0026thinsp;=\u0026thinsp;Student\u0026rsquo;s t-test, b\u0026thinsp;=\u0026thinsp;Pearson chi squared test, c\u0026thinsp;=\u0026thinsp;Fisher\u0026rsquo;s exact test, d\u0026thinsp;=\u0026thinsp;Mann-Whitney U test. ICR\u0026thinsp;=\u0026thinsp;intra-resection cavity recurrence, ECR\u0026thinsp;=\u0026thinsp;extra-resection cavity recurrence, KPS\u0026thinsp;=\u0026thinsp;Karnofsky Performance Scale, IDH\u0026thinsp;=\u0026thinsp;isocitrate dehydrogenase, SVZ\u0026thinsp;=\u0026thinsp;subventricular zone.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eFeature selection and radiomics score\u003c/h2\u003e \u003cp\u003eThe Pearson or Spearman correlation analysis and the LASSO logistic regression model was used in the training set to screen the seven features with nonzero coefficients related to spatial recurrence patterns (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). The weights of the nonzero coefficients and radiomics score calculation formula are shown \u003cb\u003ein Supplementary Table\u0026nbsp;2\u003c/b\u003e. Notably, there was a significant difference in the radiomics score between ICR and ECR (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001), which was subsequently confirmed in the validation set. The radiomics score of patients with ECR were higher than those of patients with ICR (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). The radiomics score in the ICR and ECR patient groups in the training set were \u0026minus;\u0026thinsp;0.030 (-0.226-0.248) vs 0.424 (0.278\u0026ndash;0.573), respectively, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001. The radiomics score in the ICR and ECR patient groups in the validation set were 0.277 (0.103\u0026ndash;0.322) vs 0.369 (0.258\u0026ndash;0.487), respectively, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.033. The radiomics score of each patient in the training and validation sets is respectively shown in Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee and \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eConstruction and validation of radiomics nomogram for spatial recurrence patterns\u003c/h2\u003e \u003cp\u003eThe logistic regression analysis results showed that radiomics score (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001) were independent factors for predicting spatial recurrence patterns. Thus, we constructed a radiomics nomogram based on these findings (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ea). The calibration curve and nonsignificant Hosmer-Lemeshow test (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.453) showed that the training set had good calibration (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb). Similarly, the predictive model showed performed well with an AUC of 0.844 (95%CI: 0.749\u0026ndash;0.914, sensitivity: 100.00%; specificity: 59.46%) in the training set (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ed). In particular, the good calibration and nonsignificant Hosmer-Lemeshow test (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.434) of the radiomics nomogram were validated by the validation set (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec). The AUC of the validation set was 0.706 (95%CI: 0.534\u0026ndash;0.844, sensitivity: 70.00%; specificity: 76.47%) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this study, preoperative radiomics analysis was used to integrate radiomic features from T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and T2WI/FLAIR images to predict spatial patterns in locally recurrent HGGs. Moreover, we developed and validated a radiomics nomogram model for predicting spatial recurrence patterns of HGGs and found that the model had good calibration and predictive performance.\u003c/p\u003e \u003cp\u003eLocal recurrence is the most common pattern of recurrence in HGGs, accounting for approximately 80% of all recurrences. Petrecca et al. analyzed the recurrence patterns of 20 patients with glioblastoma after total resection plus standard treatment and found that 17 patients (17/20, 85%) had tumor recurrence at the resection margin [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Despite total resection of the enhanced tumor, it was observed that most recurrent tumors occur at the margin of surgical resection[\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e], suggesting that local recurrence still exhibits different spatial recurrence patterns. Therefore, in this study, we divided local recurrence into ICR and ECR based on the relationship between the recurrent tumor and the surgical area or surgical resection cavity. Wang et al. reported that spatial recurrence patterns (ICR and ECR) were associated with prognosis in patients with locally recurrent HGGs after maximum safety resection and chemoradiotherapy [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Moreover, developing individualized treatment plans based on recurrence patterns may improve the prognosis of patients with HGGs. Previous studies have reported that extended resection of FLAIR abnormalities in patients with glioblastoma could prolong survival [\u003cspan additionalcitationids=\"CR28\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e]. Thus, if patients with high risk of ECR can be preoperatively predicted, expanding the extent of resection may benefit patient survival. Besides, if patients are at high risk of ICR, increasing the local radiotherapy dose may improve their prognosis. Preoperative non-invasive prediction of spatial recurrence patterns in patients with HGGs is important for clinical decision-making.\u003c/p\u003e \u003cp\u003eGlioma recurrence is associated with tumor heterogeneity, which is manifested on imaging as enhanced tumor and T2WI/FLAIR hyperintense perilesional region [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. However, it is not easy to accurately assess tumor aggressiveness using conventional imaging processing methods. Radiomics is an emerging field that transforms medical images into massive amounts of quantitative data to reveal disease information. Compared to traditional imaging methods, radiomics can provide objective and quantitative image information about the lesion, independent of the individual subjective judgment and experience of radiologist. Previous studies have shown that radiomics has high diagnostic performance in predicting the recurrence pattern of glioblastoma [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Among them, Shim et al. [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e] established a model of dynamic sensitivity contrast-enhanced MRI using a radiomic-based neural network, and predicted the AUC of local recurrence to be 0.969. However, the study of a multimodal MRI radiomics model to predict the spatial recurrence pattern of local recurrence has not been reported. In this study, we investigated the spatial pattern of local recurrence and attempted to identify the value of radiomic features in predicting the spatial patterns of local recurrence. The integrated radiomics features of T1WI and T2WI/FLAIR can more comprehensively reflect the characteristics of the tumor microenvironment, predict tumor invasion and the recurrence patterns. In addition, as in previous studies, we segmented lesion habitats, including T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region, to better capture tumor heterogeneity [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. Therefore, we developed a prediction of the spatial recurrence pattern based on radiomics score of enhance lesions and T2WI/FLAIR hyperintense perilesional region. This study demonstrated that in both the training and validation sets, the radiomics score for ECR was higher than for ICR and the radiomics score showed a good predictive performance.\u003c/p\u003e \u003cp\u003eThe radiomics score consisted of 7 radiomics features, including two first-order statistical and five texture features. Among these features, the one with the highest weight coefficient was GLCM, followed by GLSZM, both of which are texture features. Texture features detect potential tumor infiltration by reflecting the heterogeneity of gliomas [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. This indicated that the preoperative tumor heterogeneity of ECR was higher than that of ICR, and that ECR tumors have a higher potential for outward invasion before surgery. We then used logistic regression analysis to construct a radiomics nomogram. The predictive performance of the constructed radiomics nomogram in the training and validation sets were good. The radiomics analysis of multimodal MRI based on lesion habitat can reflect the preoperative tumor heterogeneity of HGGs with different spatial recurrence patterns, and predict ECR and ICR.\u003c/p\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eThis study has some limitations. First, this study was based on conventional MRI, and future studies may apply more advanced MR sequences to improve the model performance. Second, the model has only been validated internally and lacks external validation. In order to promote the clinical application of our model, it is necessary to conduct multi-center studies with larger sample sizes for validation. Finally, we use logistic regression to construct predictive model, while other machine learning methods, such as support vector machine, K-nearest neighbor classifier, and Gaussian naive Bayes can also be used to filter the best model. In the future, more machine learning methods need to be established to explore the optimal radiomics model.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur study showed that radiomic features extracted from T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region on CE-T1WI and FLAIR images can be used noninvasively to preoperatively predict spatial patterns in locally recurrent HGGs, providing a basis for determining personalized treatment strategies for HGGs.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eAUC \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Area under the receiver operating characteristic curve\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCE \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Contrast-enhanced\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eECR \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Extra-resection cavity recurrence\u003c/p\u003e\n\u003cp\u003eFOV \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Field of view\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eGd-DTPA \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Gadopentetate glucosamine,\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHGGs \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; High-grade gliomas\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eICR \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Intra-resection cavity recurrence\u003c/p\u003e\n\u003cp\u003eMRI \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Magnetic resonance imaging\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eRANO \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Response Assessment in Neuro-Oncology\u003c/p\u003e\n\u003cp\u003eROI \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Regions of interest\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSVZ \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Subventricular zone\u003c/p\u003e\n\u003cp\u003eT2WI/FLAIR \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Fluid-attenuated inversion recovery\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eT1WI \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;T1-weighted imaging\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTE \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Time of echo\u003c/p\u003e\n\u003cp\u003eTI \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Inversion time\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTR \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Repetition time\u0026nbsp;\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e This work was supported by Chongqing medical scientific research project (No.2023MSXM009), by the National Natural Science Foundation of China (No.81701661) and by Science and Technology planning project of Chongqing Clinical Research Centre of Imaging and Nuclear Medicine (No. CSTC2015YFPT-gcjsyjzx0175).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclaration of Competing Interest:\u003c/strong\u003e The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Han-wei Wang, Lin-lan Zeng, Xiao-guang Li, Mi-mi Zhao, Xuan Li, Ling Feng, Ping Xiang, Li-zhao Chen, Jing Tian, Qi-sheng Ran. The first draft of the manuscript was written by Han-wei Wang, Lin-lan Zeng. Liang Yi, Shu-nan Wang revised the manuscript critically for important intellectual content and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was performed in line with the principles of the Declaration of Helsinki. Approved by the institutional review board of the Army Medical Center of the PLA, and requirement of written informed consent was waived (Date November 9, 2022 /No.2022-325).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent to publish\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors affirm that human research participants provided informed consent for publication of the images in Figure(s) 1a, 2a and 2b.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAcknowledgements\u003c/em\u003e\u003c/strong\u003e: None.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eTan AC, Ashley DM, L\u0026oacute;pez GY, Malinzak M, Friedman HS, Khasraw M (2020) Management of glioblastoma: State of the art and future directions. CA: a cancer journal for clinicians 70: 299-312 doi:10.3322/caac.21613\u003c/li\u003e\n\u003cli\u003eMcKinnon C, Nandhabalan M, Murray SA, Plaha P (2021) Glioblastoma: clinical presentation, diagnosis, and management. BMJ (Clinical research ed) 374: n1560 doi:10.1136/bmj.n1560\u003c/li\u003e\n\u003cli\u003eKonishi Y, Muragaki Y, Iseki H, Mitsuhashi N, Okada Y (2012) Patterns of intracranial glioblastoma recurrence after aggressive surgical resection and adjuvant management: retrospective analysis of 43 cases. Neurologia medico-chirurgica 52: 577-586 doi:10.2176/nmc.52.577\u003c/li\u003e\n\u003cli\u003ePiper RJ, Senthil KK, Yan JL, Price SJ (2018) Neuroimaging classification of progression patterns in glioblastoma: a systematic review. Journal of neuro-oncology 139: 77-88 doi:10.1007/s11060-018-2843-3\u003c/li\u003e\n\u003cli\u003eGe S, Shi Y, Zhu G, Li S, Cai Y, Ji P, Liu J, Guo W, Gong L, Lou M, Feng F, Wang Y, Zhai Y, Qu Y, Wang L (2020) Molecular Pathological Markers Correlated With the Recurrence Patterns of Glioma. Frontiers in oncology 10: 565045 doi:10.3389/fonc.2020.565045\u003c/li\u003e\n\u003cli\u003eCho SJ, Kim HS, Suh CH, Park JE (2020) Radiological Recurrence Patterns after Bevacizumab Treatment of Recurrent High-Grade Glioma: A Systematic Review and Meta-Analysis. Korean journal of radiology 21: 908-918 doi:10.3348/kjr.2019.0898\u003c/li\u003e\n\u003cli\u003eWang H, Zeng L, Wu H, Tian J, Xie H, Zhang L, Ran Q, Zhong P, Chen L, Yi L, Wang S (2023) Preoperative vascular heterogeneity based on dynamic susceptibility contrast MRI in predicting spatial pattern of locally recurrent high-grade gliomas. European radiology doi:10.1007/s00330-023-10149-6\u003c/li\u003e\n\u003cli\u003eRingel F, Pape H, Sabel M, Krex D, Bock HC, Misch M, Weyerbrock A, Westermaier T, Senft C, Schucht P, Meyer B, Simon M (2016) Clinical benefit from resection of recurrent glioblastomas: results of a multicenter study including 503 patients with recurrent glioblastomas undergoing surgical resection. Neuro-oncology 18: 96-104 doi:10.1093/neuonc/nov145\u003c/li\u003e\n\u003cli\u003eShibahara I, Sonoda Y, Saito R, Kanamori M, Yamashita Y, Kumabe T, Watanabe M, Suzuki H, Watanabe T, Ishioka C, Tominaga T (2013) The expression status of CD133 is associated with the pattern and timing of primary glioblastoma recurrence. Neuro-oncology 15: 1151-1159 doi:10.1093/neuonc/not066\u003c/li\u003e\n\u003cli\u003eJungk C, Warta R, Mock A, Friauf S, Hug B, Capper D, Abdollahi A, Debus J, Bendszus M, von Deimling A, Unterberg A, Herold-Mende C (2019) Location-Dependent Patient Outcome and Recurrence Patterns in IDH1-Wildtype Glioblastoma. Cancers 11 doi:10.3390/cancers11010122\u003c/li\u003e\n\u003cli\u003eYoon N, Kim HS, Lee JW, Lee EJ, Maeng LS, Yoon WS (2020) Targeted Genomic Sequencing Reveals Different Evolutionary Patterns Between Locally and Distally Recurrent Glioblastomas. Cancer genomics \u0026amp; proteomics 17: 803-812 doi:10.21873/cgp.20234\u003c/li\u003e\n\u003cli\u003eJiang H, Yu K, Li M, Cui Y, Ren X, Yang C, Zhao X, Lin S (2020) Classification of Progression Patterns in Glioblastoma: Analysis of Predictive Factors and Clinical Implications. Frontiers in oncology 10: 590648 doi:10.3389/fonc.2020.590648\u003c/li\u003e\n\u003cli\u003eTang W, Wang X, Chen Y, Zhang J, Chen Y, Lin Z (2015) CXCL12 and CXCR4 as predictive biomarkers of glioma recurrence pattern after total resection. Pathologie-biologie 63: 190-198 doi:10.1016/j.patbio.2015.07.002\u003c/li\u003e\n\u003cli\u003eMistry AM, Kelly PD, Gallant JN, Mummareddy N, Mobley BC, Thompson RC, Chambless LB (2019) Comparative Analysis of Subventricular Zone Glioblastoma Contact and Ventricular Entry During Resection in Predicting Dissemination, Hydrocephalus, and Survival. Neurosurgery 85: E924-e932 doi:10.1093/neuros/nyz144\u003c/li\u003e\n\u003cli\u003eComas S, Luguera E, Molero J, Bala\u0026ntilde;a C, Estival A, Casta\u0026ntilde;er S, Carrato C, Hostalot C, Teixidor P, Vill\u0026agrave; S (2021) Influence of glioblastoma contact with the subventricular zone on survival and recurrence patterns. Clinical \u0026amp; translational oncology : official publication of the Federation of Spanish Oncology Societies and of the National Cancer Institute of Mexico 23: 554-564 doi:10.1007/s12094-020-02448-x\u003c/li\u003e\n\u003cli\u003eLi M, Huang W, Chen H, Jiang H, Yang C, Shen S, Cui Y, Dong G, Ren X, Lin S (2022) T2/FLAIR Abnormity Could be the Sign of Glioblastoma Dissemination. Frontiers in neurology 13: 819216 doi:10.3389/fneur.2022.819216\u003c/li\u003e\n\u003cli\u003eLambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout RG, Granton P, Zegers CM, Gillies R, Boellard R, Dekker A, Aerts HJ (2012) Radiomics: extracting more information from medical images using advanced feature analysis. European journal of cancer (Oxford, England : 1990) 48: 441-446 doi:10.1016/j.ejca.2011.11.036\u003c/li\u003e\n\u003cli\u003eAerts HJ, Velazquez ER, Leijenaar RT, Parmar C, Grossmann P, Carvalho S, Bussink J, Monshouwer R, Haibe-Kains B, Rietveld D, Hoebers F, Rietbergen MM, Leemans CR, Dekker A, Quackenbush J, Gillies RJ, Lambin P (2014) Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach. Nature communications 5: 4006 doi:10.1038/ncomms5006\u003c/li\u003e\n\u003cli\u003eIsmail M, Hill V, Statsevych V, Huang R, Prasanna P, Correa R, Singh G, Bera K, Beig N, Thawani R, Madabhushi A, Aahluwalia M, Tiwari P (2018) Shape Features of the Lesion Habitat to Differentiate Brain Tumor Progression from Pseudoprogression on Routine Multiparametric MRI: A Multisite Study. AJNR American journal of neuroradiology 39: 2187-2193 doi:10.3174/ajnr.A5858\u003c/li\u003e\n\u003cli\u003eWei J, Yang G, Hao X, Gu D, Tan Y, Wang X, Dong D, Zhang S, Wang L, Zhang H, Tian J (2019) A multi-sequence and habitat-based MRI radiomics signature for preoperative prediction of MGMT promoter methylation in astrocytomas with prognostic implication. European radiology 29: 877-888 doi:10.1007/s00330-018-5575-z\u003c/li\u003e\n\u003cli\u003eYan JL, Toh CH, Ko L, Wei KC, Chen PY (2021) A Neural Network Approach to Identify Glioblastoma Progression Phenotype from Multimodal MRI. Cancers 13 doi:10.3390/cancers13092006\u003c/li\u003e\n\u003cli\u003eShim KY, Chung SW, Jeong JH, Hwang I, Park CK, Kim TM, Park SH, Won JK, Lee JH, Lee ST, Yoo RE, Kang KM, Yun TJ, Kim JH, Sohn CH, Choi KS, Choi SH (2021) Radiomics-based neural network predicts recurrence patterns in glioblastoma using dynamic susceptibility contrast-enhanced MRI. Scientific reports 11: 9974 doi:10.1038/s41598-021-89218-z\u003c/li\u003e\n\u003cli\u003eJuan-Albarrac\u0026iacute;n J, Fuster-Garcia E, Manj\u0026oacute;n JV, Robles M, Aparici F, Mart\u0026iacute;-Bonmat\u0026iacute; L, Garc\u0026iacute;a-G\u0026oacute;mez JM (2015) Automated glioblastoma segmentation based on a multiparametric structured unsupervised classification. PloS one 10: e0125143 doi:10.1371/journal.pone.0125143\u003c/li\u003e\n\u003cli\u003evan Griethuysen JJM, Fedorov A, Parmar C, Hosny A, Aucoin N, Narayan V, Beets-Tan RGH, Fillion-Robin JC, Pieper S, Aerts H (2017) Computational Radiomics System to Decode the Radiographic Phenotype. Cancer research 77: e104-e107 doi:10.1158/0008-5472.Can-17-0339\u003c/li\u003e\n\u003cli\u003ePetrecca K, Guiot MC, Panet-Raymond V, Souhami L (2013) Failure pattern following complete resection plus radiotherapy and temozolomide is at the resection margin in patients with glioblastoma. Journal of neuro-oncology 111: 19-23 doi:10.1007/s11060-012-0983-4\u003c/li\u003e\n\u003cli\u003eLem\u0026eacute;e JM, Clavreul A, Menei P (2015) Intratumoral heterogeneity in glioblastoma: don\u0026apos;t forget the peritumoral brain zone. Neuro-oncology 17: 1322-1332 doi:10.1093/neuonc/nov119\u003c/li\u003e\n\u003cli\u003eLi YM, Suki D, Hess K, Sawaya R (2016) The influence of maximum safe resection of glioblastoma on survival in 1229 patients: Can we do better than gross-total resection? Journal of neurosurgery 124: 977-988 doi:10.3171/2015.5.Jns142087\u003c/li\u003e\n\u003cli\u003ePessina F, Navarria P, Cozzi L, Ascolese AM, Simonelli M, Santoro A, Clerici E, Rossi M, Scorsetti M, Bello L (2017) Maximize surgical resection beyond contrast-enhancing boundaries in newly diagnosed glioblastoma multiforme: is it useful and safe? A single institution retrospective experience. Journal of neuro-oncology 135: 129-139 doi:10.1007/s11060-017-2559-9\u003c/li\u003e\n\u003cli\u003eLu M, Fu ZH, He XJ, Lu JK, Deng XQ, Lin DL, Gu YM, Fan YF, Lai MY, Li J, Yang MM, Chen ZP (2020) T2 Fluid-Attenuated Inversion Recovery Resection for Glioblastoma Involving Eloquent Brain Areas Facilitated Through Awake Craniotomy and Clinical Outcome. World neurosurgery 135: e738-e747 doi:10.1016/j.wneu.2019.12.130\u003c/li\u003e\n\u003cli\u003eHu LS, Hawkins-Daarud A, Wang L, Li J, Swanson KR (2020) Imaging of intratumoral heterogeneity in high-grade glioma. Cancer letters 477: 97-106 doi:10.1016/j.canlet.2020.02.025\u003c/li\u003e\n\u003cli\u003eLem\u0026eacute;e JM, Clavreul A, Aubry M, Com E, de Tayrac M, Eliat PA, Henry C, Rousseau A, Mosser J, Menei P (2015) Characterizing the peritumoral brain zone in glioblastoma: a multidisciplinary analysis. Journal of neuro-oncology 122: 53-61 doi:10.1007/s11060-014-1695-8\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"glioma, recurrence, habitat, magnetic resonance imaging, radiomics","lastPublishedDoi":"10.21203/rs.3.rs-3870027/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3870027/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eThis study aims to preoperatively predict spatial patterns in locally recurrent high-grade gliomas (HGGs) based on lesion habitat radiomics analysis of multimodal MRI and to evaluate the predictive performance of this approach.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eOur study included 121 patients with locally recurrent HGGs after maximum safe surgical resections and radiotherapy combined with temozolomide (training set, \u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;84; validation set, \u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;37). Local recurrence was divided into intra-resection cavity recurrence (ICR) and extra-resection cavity recurrence (ECR), according to the distance between the recurrent tumor and the surgical area or resection cavity. Radiomic features were extracted from the lesion habitat (T1WI enhancing lesion and T2WI/FLAIR hyperintense perilesional region) on contrast-enhanced T1WI and FLAIR, respectively. The LASSO was used to select radiomic features and calculate radiomics score. Logistic regression analysis was used to construct a predictive radiomics model, which was evaluated using calibration curves and the area under the receiver operating characteristic curve (AUC).\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eSeven features with nonzero coefficients related to spatial recurrence patterns were selected. The radiomics score of patients with ECR was higher than that of patients with ICR in the training set [0.424 (0.278\u0026ndash;0.573) vs. -0.030 (-0.226-0.248), \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;0.001] and in the validation set [0.369 (0.258\u0026ndash;0.487) vs. 0.277 (0.103\u0026ndash;0.322), \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.033]. The radiomics model demonstrated good calibration and performed well in predicting ECR, with AUC values of 0.844 in the training set and 0.706 in the validation set.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eRadiomics analysis of lesion habitat can preoperatively predict spatial patterns in locally recurrent HGGs, providing a basis for determining personalized treatment strategies for HGGs.\u003c/p\u003e","manuscriptTitle":"Multimodal MRI lesion habitat-based radiomics analysis for preoperative prediction of spatial pattern in locally recurrent high-grade gliomas","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-19 21:15:18","doi":"10.21203/rs.3.rs-3870027/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3569f2d5-898a-4181-83ed-8b47be6b190c","owner":[],"postedDate":"January 19th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-01-28T21:14:09+00:00","versionOfRecord":[],"versionCreatedAt":"2024-01-19 21:15:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3870027","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-3870027","identity":"rs-3870027","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}