Risk of Peritumoral Invasion in Rat Glioblastoma: Nomogram-based Ultrasound Localization Microscopy

Risk of Peritumoral Invasion in Rat Glioblastoma: Nomogram-based Ultrasound Localization Microscopy | 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 Risk of Peritumoral Invasion in Rat Glioblastoma: Nomogram-based Ultrasound Localization Microscopy Xing Hu, Gaobo Zhang, Xiandi Zhang, Yong Wang, Xin Liu, Hong Ding This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6464545/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Background This study aims to develop and assess a nomogram based on multiparametric ultrasound localization microscopy to evaluate the risk of peritumoral invasion. Methods Thirty-six in situ rat glioblastoma models were created. After craniotomy, ultrasound localization microscopy was used to quantify microvascular morphology and hemodynamics, which were combined with multimodal magnetic resonance imaging to manually delineate the invasive and normal brain regions. The least absolute shrinkage and selection operator regression algorithm was applied to select ultrasound localization microscopy parameters, followed by multivariable logistic regression to identify significant variables. A nomogram to predict peritumoral invasion risk was constructed using R software, and its diagnostic performance was evaluated. Results Vascularity (p < 0.001), orientation variance (p = 0.013), and diameter (p = 0.002) were identified as independent predictors of peritumoral invasion. The prediction model demonstrated strong discriminatory power, with an area under the curve of 0.964 (0.933–0.994) for the training set and 0.995 (0.984–1.000) for the validation set. The goodness-of-fit Hosmer-Lemeshow test statistics were 5.135 (p = 0.702) and 3.163 (p = 0.237), indicating that the predicted invasion risk closely matched the actual risk. Decision curve analysis revealed that when the invasion incidence ranged from 1–99% in the training set and from 5–94% in the validation set, the nomogram provided clinical benefit, demonstrating good generalizability. Conclusions We developed and validated a nomogram to predict peritumoral invasion in glioblastoma, enabling clinicians to perform preoperative risk assessments and implement personalized surgical strategies to improve resection rates. Ultrasound localization microscopy Glioblastoma Visualization Invasion Microvascular Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Glioblastoma (GBM) is a highly malignant and aggressive primary brain tumor. Standard treatment generally involves maximal safe resection, followed by radiotherapy and chemotherapy [ 1 ]. A major challenge is the diffuse invasion of surrounding tissue. Incomplete resection of invasive areas during surgery is a major cause of high recurrence and mortality rates due to residual tumor cells. Notably, 85% of recurrences occur at the surgical margins. Maximizing tumor removal while preserving neurological function is crucial [ 2 , 3 ]. This presents significant challenges due to unclear imaging boundaries between invasive and normal tissues. Achieving a precise preoperative invasion risk assessment to guide intraoperative tumor resection, while preserving normal neurological function, is a key focus in glioma surgery [ 4 , 5 ]. Current technologies are limited in accurately assessing the pathological features of noncontrast-enhanced tumor components, often resulting in missed small invasive foci during surgical resection or radiotherapy [ 6 ]. Although MRI and CT provide good resolution and reduce brain shift effects, they require specific equipment standards and intraoperative conditions, which increase anesthesia time and safety risks. Furthermore, they do not provide real-time multiparametric imaging [ 7 ]. Invasive finger-like projections may not always be visible [ 8 ]. Fluorescence navigation uses selectively accumulating photosensitive agents to highlight tumor borders, improving resection rates and reducing residuals, while offering simplicity and repeatability in administration. However, agents such as 5-aminolevulinic acid rely on blood-brain barrier (BBB) disruption, limiting the discernibility of deeper structures. Indocyanine green bypasses the BBB but exhibits low signal intensity, and its spectral strength assessment is subjective. However, standardized quantitative measures for weakly fluorescent surrounding tissue are lacking, potentially leading to false negatives when distinguishing between low-grade tumors, inflammatory tissue, and edema [ 9 ]. Emerging technologies, including Raman imaging, terahertz imaging, two-dimensional local correlation spectroscopy, ion mass spectrometry, and inflammation-specific imaging, demonstrate high sensitivity in detecting molecular markers or metabolites to delineate tumor boundaries. However, these techniques are operationally complex, involve small sample sizes, are time-consuming, and suffer from low signal-to-noise ratios [ 10 , 11 ]. There is a need for highly accurate, rapid, noninvasive intraoperative tools to guide maximal tumor resection. Tumor cells frequently invade from enhanced regions to distant areas, with marginal zones exhibiting characteristics ranging from sparse tumor cell invasion to invasive islands. However, current imaging modalities typically reveal only dense tumor regions and fail to detect areas of low-density invasion [ 1 ]. Microvascular pathology is an early and common feature of GBM, characterized histologically by microvascular proliferation and necrosis [ 12 ]. Folkman (1971) hypothesized that tumor growth and spread depend on angiogenesis, suggesting that the formation of new blood vessels induced by tumors is essential for their invasion [ 13 ]. Cellular invasion often follows perivascular spaces, facilitating spread along blood vessels into surrounding tissue. The marginal zone is the primary site of tumor invasion. Developing invasive molecular models is challenging, primarily due to the brain’s intricate architecture and the diverse cellular composition in the tumor microenvironment, particularly considering the three-dimensional organization of tissue [ 14 ]. Therefore, detailed microanatomical studies, including thin-walled tissue and the vascular niche, are essential for understanding tumor biology [ 15 ]. Developing a reliable preoperative tool to visualize vascular anatomy and function, thus guiding surgical resection, is essential. Ultrasound localization microscopy (ULM) has recently emerged as a novel microvascular imaging technology with significant potential for both structural and functional visualization. Unlike single-plane waves, ULM uses multi-angle coherent superposition to merge images from different angles, achieving uniform focusing at various depths and directions, greatly enhancing blood flow detection sensitivity [ 16 , 17 ]. ULM quantifies microhemodynamics and perfusion across a broad dynamic range, effectively compensating for signal-to-noise ratio losses and resolution degradation caused by unfocused ultrasound beams. It addresses the trade-offs between spatial resolution and penetration depth in previous technologies [ 18 ]. Nomograms are practical tools for predicting event probabilities and guiding clinical decisions. Incorporating biomarkers enhances the precision and personalization of these predictive models [ 19 – 21 ]. However, few studies have reported using nomograms to predict the risk of peritumoral invasion in GBM. This study aims to create a nomogram using preoperative ULM data to help clinicians identify patients at high risk of GBM invasion. By enabling early detection and targeted interventions for residual lesions post-surgery, this approach aims to improve patient outcomes. Materials and Methods This study was conducted with approval from the Animal Ethics Committee of Fudan University (approval number: 202408008S) and was in accordance with the declaration of Helsinki. The research workflow is shown in Fig. 1 . Rat GBM Modeling Thirty-six male Sprague-Dawley rats (6 weeks old, weighing 200–250 g) were used to create the GBM models. The rats were housed under controlled conditions, including regulated temperature, humidity, and a 12-hour light/dark cycle. The rats had free access to rodent food and sterilized tap water. The C6 glioma cell line, obtained from the National Institutes of Health, was cultured in Dulbecco’s Modified Eagle Medium with 10% fetal bovine serum and 5% penicillin-streptomycin. A C6 cell suspension (5 × 10 5 cells per 10 µL) was injected into each rat to establish the model. Anesthesia was induced by intraperitoneally administering a ketamine-xylazine mixture (60 mg/kg). The rats were secured in a stereotactic frame and stabilized with ear bars and a nose clamp. A 25-gauge flat-tip syringe (25G; Hamilton Company, Reno, NV, USA) was used to inject 20 µL of the cell suspension into the caudate nucleus. A 0.5 cm longitudinal scalp incision was made at the intersection of the rat's medial canthus and sagittal suture. The skin was retracted to expose the bregma and sagittal suture. The needle was displaced laterally by 3 mm from the intersection point, then advanced 1 mm toward the head to reach a depth of approximately 5.5 mm, at an injection rate of 4 µL/min. After a 2-minute retention, the needle was slowly withdrawn (Fig. 2 a). The injection site was treated with 5% penicillin-streptomycin, and the bone defect was sealed with sterile bone wax. The scalp was sutured, and the rats were returned to their cages for recovery. MRI Pre-Scanning MRI scanning was performed one day before the super-resolution experiment. The rats were kept under light anesthesia with isoflurane (1-1.25% oxygen concentration) and placed on a heating pad for warmth. Imaging was performed via 3.0T MRI (uMR870, United Imaging Healthcare, 2022, China) in the coronal plane. The scanning sequences included T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR), and diffusion-weighted imaging (DWI). After a tail vein injection of 0.6 mmol/kg gadobutrol, contrast-enhanced scans were performed, with scanning parameters listed in Supplementary Table 1. Super-Resolution Data Acquisition Surgical Site Preparation. The rats were anesthetized in a gas induction chamber with 4% isoflurane (R510-22-10, RWD) and maintained at a 1% concentration. After hair removal from the neck, the skin was disinfected with iodine, followed by 70% ethanol. A PE10 medical-grade polyethylene catheter (inner diameter: 0.28 mm, outer diameter: 0.61 mm) was used to cannulate the right jugular vein. Ophthalmic solution was applied to the rats' eyes to prevent dryness while they were positioned prone, and their heads were secured in a stereotactic frame. After incising the scalp, the periosteum was carefully removed using a scalpel. A craniotome (78001, RWD) was used to thin the skull until the vascular system became visible. A microscope (DOM-1001, RWD) was used to preserve larger vessels and the superior sagittal sinus, while micro-forceps retracted the skull to expose the dura mater. Hemostatic sponges were applied to control minor bleeding from superficial injuries, and sterile saline was used to prevent thermal damage and reduce bleeding and swelling. Multi-Angle Plane Radiofrequency Data Acquisition. The L22-14vX LF transducer (15.625 MHz, MS200, Visualsonics Ltd., Toronto, ON, Canada) was securely attached to the stereotactic imaging frame using a high-precision linear stage (VT-80, Physik Instrumente, Auburn, MA, USA). The transducer was oriented to produce coronal anatomical brain sections. The tumor site was identified and aligned with the MRI coronal plane using the Mindray M10 Portable Ultrasound System (Mindray, Shenzhen, China). The Verasonics Vantage 256 system (Verasonics Ltd., Kirkland, WA, USA) and the L22-14vX-LF transducer were aligned for motion adjustments in 1 mm increments. A programmable microinjection pump (R462, RWD) infused diluted SonoVue (Bracco Imaging, Massy, France) at 80 µL/min. The imaging sequence included multiple 5-angle plane waves (-5°, -2.5°, 0°, 2.5°, 5°) [ 22 ]. Data collection was synchronized with a function generator, and all processing was performed in the MATLAB environment (The MathWorks, Natick, Massachusetts, USA, R2020a). The number of scanned sections per rat was determined by tumor size, as shown in Supplementary Table 2. Singular value decomposition filtering was used to extract microbubble signals from the tissue background of each in-phase/quadrature dataset. A threshold was chosen at the inflection point of the singular value curve to optimize clutter filtering. The centroid of the microbubbles was determined using the radial symmetry algorithm in MATLAB. The Hungarian algorithm was used for frame-to-frame centroid pairing and trajectory estimation. A minimum trajectory length of 15 frames for microbubbles was set, followed by accumulating each acquisition to reconstruct the vascular system in ULM (Fig. 2 b). Vessel reconstruction in superficial cortical areas was limited due to significant echo scattering from the ear bars and acoustic shadows from other skull regions [ 23 ]. Histopathology After collecting the ULM data, the rats were humanely euthanized. Their brains were quickly extracted and preserved in formalin. To minimize analytical errors from sample position deviations, tumors were coronally sectioned into 1 mm thick slices using the Rat Brain Matrix (WPI-RBMA-600S), and guided by the Mindray M10 Portable Ultrasound System. This approach ensured that the ULM sections precisely matched the sample positions for histological staining analysis. Hematoxylin-eosin (H&E) staining was used to clearly visualize tumor cells. Quantitative analysis of H&E-stained sections was performed using digital image analysis software (Fiji/Image J v.1.54f), which included automated cell counting and density distribution analysis. This analysis quantified variations in tumor cell density, distinguishing between tumor regions (≥ 90% tumor cell occupancy), invasive regions (< 90% tumor cell occupancy), and normal brain regions (no tumor cells) (Fig. 3 ). Image Post-Processing For each section, the T2, DWI, and T1CE sequences were co-registered with the FLAIR sequence. FSL bias field correction tools were used to correct magnetic field inhomogeneities, and the FLIRT linear registration tool enabled rigid image registration. In the T1CE sequence, the enhanced tumor region, including necrosis, was identified as the region of interest (ROI), while in the FLAIR sequence, high-signal and abnormal signals (such as tumors and edema) were designated as ROIs. Based on the image registration results, ROIs were manually delineated and mapped to the super-resolved images, encompassing both invasive regions and normal brain tissue (Fig. 2 c). Image features were extracted using scale-invariant feature transform and wavelet transform algorithms in Fiji/ImageJ v.1.54f (Wayne Rasband, NIH, USA), with segmentation performed using the Trainable Weka Segmentation tool. Vascular segmentation was carried out in a supervised, semi-automated manner. Random forest classifiers were used to train models for image pre-collection, deep learning-based vascular segmentation, feature extraction, and scanning modulation, effectively representing the three-dimensional volume of interest in two-dimensional depictions. Vascular features, including area, skeleton, and midpoint, were extracted from binarized images using deep learning methods. The out-of-bag (OOB) error was defined as the error rate of samples not used in training decision trees, enabling the random forest to use OOB samples for unbiased internal estimation. To improve model generalization, the OOB error was maintained below 5%. Curvature is defined as follows (1): Lc represents the actual path length of a vessel segment, and L represents the linear distance between the segment endpoints. The fractal dimension (FD) of vascular structures quantifies the complexity and branching patterns of vascular networks, reflecting the degree of branching and twisting. The box-counting method is applied for segmentation across multiple scales, followed by linear regression analysis to derive the FD. To accurately capture the fractal characteristics of the images, an appropriate two-dimensional Euclidean space (R 2 > 0.99) was selected, ensuring a good fit between the fitted model and the actual data (2). Here, r represents the grid scale, and M(r) denotes the minimum number of grids required to cover the fractal. Assuming the mean vector of the vessel orientation is , the orientation variance (OV) of the vessels is calculated as follows (3): Let N represent the total count of blood vessels, θ i denote the direction of the i-th data point, and be the average angle across all directions. Blood flow reflects the functional perfusion status based on Poiseuille’s flow principles (4): Q = A×V (4) Where Q represents blood flow, A denotes the vessel’s cross-sectional area, and V indicates blood flow velocity. Statistical Analysis The samples were randomly divided into two groups: one for training the predictive model and the other for evaluation. Continuous variables were analyzed using a t-test. LASSO regression, conducted with the 'glmnet' package in R, was used to select relevant microvascular parameters. A multifactorial regression model was then developed using the Akaike Information Criterion (AIC) for model selection, with significant parameters chosen as input variables. A personalized nomogram based on ULM scores was created from the training set to predict the risk of tumor invasion by aggregating scores and aligning them with a predefined scale. The model’s calibration was tested using the Hosmer-Lemeshow method, and its clinical significance was further evaluated using decision curve analysis. Classification performance was evaluated using metrics from the receiver operating characteristic (ROC) curve, including accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC). The optimal critical values for each indicator were determined using the Youden index (Youden index = sensitivity + specificity − 1) and validated in the validation group. The model’s validation was conducted using the validation set, with statistical significance set at P = 0.05. Results ULM Baseline Features This study analyzed 208 sections from rat GBM models. The training dataset consisted of 146 sections: 72 from invasion areas and 74 from normal brain areas. The validation dataset included 62 sections: 32 from invasion areas and 30 from normal brain areas. Table 1 shows the baseline characteristics of ULM for both datasets. Table 1 Baseline characteristics of the ULM cohort Characteristic Training set (95% CI) Validation set (95% CI) Invasion Normal p-value Invasion Normal p-value No. of examinations 72 74 --- 32 30 --- Diameter (µm) 47.12 (43.47, 50.76) 29.26 (26.48, 32.03) < 0.001 47.97 (42.83, 53.10) 29.55 (25.53, 33.56) < 0.001 Vascularity (%) 24.25 (22.79, 25.71) 14.97 (14.04, 15.90) < 0.001 24.67 (22.82, 26.52) 11.67 (9.92, 13.41) < 0.001 Curvature 1.1837 (1.1753, 1.1921) 1.1531 (1.1479, 1.1583) < 0.001 1.1804 (1.1711, 1.1897) 1.1169 (1.1331, 1.1606) < 0.001 Branch points (/mm²) 74.03 (64.80, 83.25) 42.31 (36.97, 47.65) < 0.001 72.56 (59.84, 85.27) 27.17 (20.45, 33.90) < 0.001 Branches (/mm²) 189.23 (167.97, 210.50) 112.48 (99.94, 125.03) < 0.001 187.11 (157.73, 216.49) 72.17 (56.90−87.44) < 0.001 FD 15.4132 (15.2398–15.5864) 14.2518 (13.9825–14.5211) < 0.001 15.4877 (15.2535, 15.7219) 13.5946 (13.0999, 14.0894) < 0.001 Velocity (mm/s) 29.92 (28.60, 31.24) 27.06 (26.27, 29.85) 0.001 30.72 (28.78, 32.66) 28.29 (26.85, 29.72) 0.060 OV 24.57 (22.47, 26.67) 16.91 (15.75, 18.07) < 0.001 25.80 (23.29, 27.70) 16.02 (13.74, 18.30) < 0.001 Blood flow (µL) 51.50 (43.19, 59.81) 33.77 (23.38, 44.15) < 0.001 57.63 (44.36, 70.91) 40.05 (31.55, 48.55) < 0.001 CI, confidence interval; FD, fractal dimension; OV, orientation variance Parameter filtering In the training set, GBM invasion (presence or absence) was used as the dependent variable to determine the optimal λ value. Six features with non-zero coefficients were selected: diameter, vascularity, curvature, FD, velocity, and OV (Fig. 4 ). This approach enables variable selection in datasets with multiple features, retaining the most explanatory variables and creating a more interpretable model. The final model optimally balances generalizability and complexity. Analysis of Peritumoral Invasion Risk Factors Univariate analysis showed a positive correlation between peritumoral invasion and several variables, including diameter (odds ratios [OR], 1.081; 95% confidence interval [CI], 1.006 to 1.161; P = 0.003), vascularity (OR, 1.289; 95% CI, 1.109 to 1.500; P = 0.001), curvature (OR, 1.210; 95% CI, 0.949 to 1.543; P = 0.125), FD (OR, 1.723; 95% CI, 0.773 to 3.838; P = 0.183), velocity (OR, 1.041; 95% CI, 0.912 to 1.188; P = 0.556), and OV (OR, 1.185; 95% CI, 1.043 to 1.347; P = 0.009). The corresponding forest plot visualizes the 95% CI for these ORs (Supplementary Fig. 1). Multivariate analysis using AIC identified diameter (OR, 1.109; 95% CI, 1.039 to 1.184; P = 0.002), vascularity (OR, 1.348; 95% CI, 1.180 to 1.541; P < 0.001), and OV (OR, 1.167; 95% CI, 1.033 to 1.318; P = 0.013) as independent predictors of peritumoral invasion (Table 2 ). Table 2 Univariate and multivariate logistic regression analysis of risk factors associated with invasion Characteristic Univariable Analysis Multivariable Model β-value SE P-value OR (95% CI) β-value SE P-value OR (95% CI) Diameter (µm) 0.078 0.037 0.003 1.081 (1.006–1.161) 0.104 0.033 0.002 1.109 (1.039–1.184) Vascularity (%) 0.254 0.077 0.001 1.289 (1.109−1.500) 0.299 0.068 < 0.001 1.348 (1.180–1.541) Curvature (%) 0.191 0.124 0.125 1.210 (0.949–1.543) FD 0.544 0.409 0.183 1.723 (0.773–3.838) Velocity (mm/s) 0.040 0.068 0.556 1.041 (0.912–1.188) OV 0.170 0.065 0.009 1.185 (1.043–1.347) 0.154 0.062 0.013 1.167 (1.033–1.318) FD, fractal dimension; SE, standard error; OR, odds ratio; OV, orientation variance Construction of the Prediction Model Risk factors from the training set, identified through multivariable logistic regression with AIC, were used to create a nomogram (Fig. 5 a). The results indicated that vascularity had the greatest impact on peritumoral invasion, followed by OV and diameter. Each parameter was assigned a score, which was then summed to calculate a cumulative risk score. A straight line was used to visually estimate the peritumoral invasion probability for each slice, with predictions closely aligning with pathological findings (Fig. 3 ). Evaluation and Validation of the Model The training set achieved an AUC of 0.964 (95% CI, 0.933–0.994), while the validation set achieved an AUC of 0.995 (95% CI, 0.984-1.000), as shown in the ROC curve (Fig. 5 b-c). The calibration curve shows that the model’s GBM peritumoral invasion risk predictions closely align with actual values in both the training and validation sets, as confirmed by the Hosmer-Lemeshow test (P > 0.05) (Fig. 5 d). Decision curve analysis shows that the training set’s net benefit is higher across the 1–99% threshold probability range, while the validation set shows a higher net benefit from 5–94% (Fig. 5 e). The model’s diagnostic performance is detailed in Table 3 . Table 3 Assessment of model diagnostic performance Group Parameter Accuracy Sensitivity Specificity PPV NPV AUC (95% CI) Optimal cutoff value Training set Diameter (µm) 0.897 0.923 0.877 0.857 0.934 0.941 (0.900−0.981) 34.881 Vascularity (%) 0.845 0.852 0.788 0.743 0.882 0.899 (0.850–0.948) 20.703 OV 0.747 0.746 0.729 0.671 0.816 0.806 (0.736–0.875) 20.520 Nomogram 0.884 0.843 0.921 0.908 0.864 0.964 (0.933, 0.994) -- Validation set Diameter (µm) 0.887 0.865 0.920 0.941 0.821 0.868 (0.774–0.961) 36.024 Vascularity (%) 0.935 0.917 0.962 0.971 0.893 0.963 (0.919−1.000) 16.523 OV 0.790 0.889 0.714 0.706 0.893 0.870 (0.783–0.956) 22.702 Nomogram 0.936 0.941 0.929 0.941 0.929 0.995 (0.984, 1.000) -- AUC, area under the curve; NPV, negative predictive value; OV, orientation variance; PPV, positive predictive value Discussion Surgery remains the primary treatment for GBM; however, residual lesions frequently occur as postoperative complications. Retrospective studies indicate that patients with residual lesions after surgery have a 5-year mortality rate 10%-20% higher than those with negative margins [ 24 – 26 ]. Therefore, identifying patients at higher risk for residual lesions following surgery and implementing early intervention measures is critical. GBM’s invasive growth is associated with microvascular heterogeneity, and microvascular pathology is closely linked to the tumor’s invasive biological behavior. Newly formed microvessels exhibit abnormal morphological and functional characteristics, including increased branching, curvature, and enhanced perfusion [ 27 ]. Previous studies suggest that changes in microvascular structure and function are early events in tumor progression, underscoring the need for early monitoring [ 28 ]. This emphasizes the potential to improve the visualization of early invasion and enhance surgical resection precision. ULM provides in vivo, real-time, high-resolution imaging of microvessels, enabling the precise characterization of microvascular features in invasive regions [ 18 ]. This capability is crucial for improving the accuracy of early predictions regarding invasion. Our study employs a microbubble tracking algorithm to quantitatively analyze microvascular structure and function, enabling clearer differentiation of microvessels smaller than 50 µm compared to traditional imaging techniques. This approach overcomes the limitations of conventional imaging in capturing sufficient hemodynamic information, enabling more accurate tumor invasion risk assessments and supporting precise surgical resection. Most research on postoperative residual lesions depends on risk factor analyses. To visualize and quantify the risk of peritumoral invasion, we developed a risk assessment model for gliomas. Nomograms are useful tools for predicting peritumoral invasion and identifying high-risk patients, guiding treatment decisions. Consequently, the model, based on ULM’s multivariable logistic regression, enables accurate predictions of event probabilities and offers personalized risk assessments for potential invasion [ 20 , 29 , 30 ]. Our results closely align with pathological findings, confirming the model’s utility in guiding precise management of residual lesions while minimizing damage to normal brain tissue. In this study, 208 samples were randomly split into training and validation sets. We identified vascularity, OV, and diameter as independent risk factors for peritumoral invasion using both univariate and multivariate logistic regression analyses. We developed a personalized nomogram incorporating multidimensional imaging scores based on these risk factors. Vascularity is crucial in peritumoral invasion, as GBM induces angiogenesis through the secretion of vascular endothelial growth factor (VEGF), promoting invasive growth. Additionally, the high permeability of newly formed vessels facilitates the spread of tumor cells into surrounding tissues, increasing the risk of invasion and metastasis [ 31 , 32 ]. Furthermore, the abnormal distribution of blood vessels is a defining characteristic of tumor invasion. Irregular vascular morphology reflects structural instability, creating a turbulent blood flow microenvironment that promotes tumor growth and migration, which encourages tumor cell invasion in these areas [ 33 ]. Uneven oxygen supply and microenvironmental changes due to abnormal blood flow directionality may induce adaptive mutations in tumor cells, further enhancing their invasive and metastatic potential [ 34 ]. Finally, changes in vascular dilation provide a basis for tumor invasion, as larger vessel diameters increase the likelihood of tumor cells invading along vascular walls. Continuous VEGF stimulation promotes vascular dilation and enhances vascular permeability and plasticity, thereby creating migration pathways for tumor cell invasion [ 27 , 35 , 36 ]. However, the limited sample size may have led to some potential significances going undetected due to insufficient statistical power. LASSO regression simplifies the model by selecting features, thus helping prevent overfitting. However, it may overlook less influential factors that could be valuable in some biological contexts [ 37 ]. For example, branch number and branch point count may reflect vascular complexity or irregularity. Since OV partially captures this information, multicollinearity may reduce their role as independent predictors, thus obscuring their biological significance [ 38 ]. The microvascular network in the invasive region exhibits high biological heterogeneity, which may reduce the impact of parameters that vary significantly among individuals (e.g., FD and curvature) in multivariate analyses, potentially causing them to fail to reach significance [ 39 ]. The heterogeneity and complexity of the microenvironment pose challenges to model construction in statistical analyses. However, novel analytical methods, such as nonlinear regression or Bayesian models, may better capture the subtle effects of these variables in future studies. Overall, this model assists clinicians in personalizing and accurately identifying peritumoral invasion risks, guiding GBM surgical resection with real-time visualization while minimizing damage. It has practical implications for developing personalized treatment plans, reducing postoperative residual risks, and improving survival outcomes [ 40 ]. For patients identified as high risk for invasion through the nomogram, maximally safe resection combined with postoperative radiotherapy, chemotherapy, tumor-treating fields, or molecularly targeted therapies may be considered to extend survival. Conversely, for low-risk patients, limited surgical resection and reduced radiation dosages could be used to manage the disease while minimizing side effects [ 41 , 42 ]. With technological advancements and broader dissemination, ULM is poised to become a routine tool in diagnosing and treating GBM, ultimately improving the quality of life for patients. Although this study focuses on GBM, the methodology has broader applicability and could be extended to other brain tumors and invasive systemic malignancies. This study also identifies a few limitations. First, despite ULM’s ability to discern fine structures, its complexity and high operational requirements necessitate specialized training and experienced personnel, limiting its widespread application, particularly in resource-limited healthcare settings. Second, although the rat model simulates human GBM, physiological differences remain, requiring further clinical trials to validate and refine its diagnostic thresholds. Third, while ULM provides micro-level advantages, its limited imaging range makes it difficult to cover the entire brain in a single scan, necessitating multiple imaging sessions that increase time costs and present challenges for data processing. Finally, this study lacks additional imaging and clinical data beyond ULM data. Future studies should integrate multimodal parameters for a more comprehensive assessment to enhance the model’s robustness [ 23 , 43 , 44 ]. Conclusion In summary, this study presents a prospective evaluation of microcirculation quantification indices in a rat glioma model using ULM and establishes a nomogram based on three risk factors to predict preoperative peritumoral invasion risk, validated as a more accurate tool for assessing surgical resection. In the future, ULM is expected to become a key tool in diagnosing and treating GBM, with broad applications in intraoperative navigation and radiotherapy planning. Abbreviations AIC Akaike information criterion AUC Area under the curve BBB Blood - brain barrier CI Confidence interval DWI Diffusion-weighted imaging FLAIR Fluid-attenuated inversion recovery GBM Glioblastoma H&E Hematoxylin-eosin LASSO Least absolute shrinkage and selection operator NPV Negative predictive value OOB Out-of-bag OR Odds ratios PPV Positive predictive value ROC Receiver operating characteristic ROI Region of interest T2WI T2-weighted imaging ULM Ultrasound localization microscopy VEGF Vascular endothelial growth factor Declarations Ethics approval and consent to participate This study was conducted with approval from the Animal Ethics Committee of Fudan University (approval number: 202408008S) and was in accordance with the declaration of Helsinki. Consent for publication Not applicable. Data Availability Statement The datasets generated and analyzed during the current study are not publicly available but available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Funding This study was supported by National Natural Science Foundation of China (grant number 82272017). Authors’ Contributions All authors contribute to the study’s conception and design. Conception and design of the study: Hu X, Zhang GB, Liu X, Ding H. Data acquisition: Hu X, Zhang GB, Zhang XD. Data analysis or interpretation: Hu X, Zhang GB, Yong Wang. Writing of the first draft of the manuscript: Hu X, Zhang GB, Zhang XD. Critical revision of the manuscript: Yong Wang, Liu X, Ding H. All authors read and approved the final manuscript. Acknowledgments Not applicable. Author details 1 Department of Ultrasound, Huashan Hospital, Fudan University, No. 12 Middle Urumqi Road, Shanghai 200040, China. 2 Department of Biomedical Engineering, School of Information Science and Technology, Fudan University, No. 2005 Songhu Road, Shanghai 200438, China. 3 Academy for Engineering and Technology, Fudan University, No. 2005 Songhu Road, Shanghai 200438, China. References Cheng M, Hu C, Yao Z, et al. Harnessing Reconstructed Macrophage Modulation of Infiltration-Excluded Immune Microenvironments To Delineate Glioma Infiltrative Region. ACS Appl Mater Interfaces. 2023. https://doi.org/10.1021/acsami.2c16586 Nie S, Zhu Y, Yang J, et al. Clinicopathologic analysis of microscopic tumor extension in glioma for external beam radiotherapy planning. BMC Med. 2021;19(1):269. https://doi.org/10.1186/s12916-021-02143-w Karschnia P, Smits M, Reifenberger G, et al. A framework for standardised tissue sampling and processing during resection of diffuse intracranial glioma: joint recommendations from four RANO groups. Lancet Oncol. 2023;24(11):438-50. https://doi.org/10.1016/S1470-2045(23)00453-9 Lee E, Lee EA, Kong E, et al. An agonistic anti-Tie2 antibody suppresses the normal-to-tumor vascular transition in the glioblastoma invasion zone. Exp Mol Med. 2023;55(2):470-84. https://doi.org/10.1038/s12276-023-00939-9 Segura-Collar B, Garranzo-Asensio M, Herranz B, et al. Tumor-Derived Pericytes Driven by EGFR Mutations Govern the Vascular and Immune Microenvironment of Gliomas. Cancer Res. 2021;81(8):2142-56. https://doi.org/10.1158/0008-5472 Chai R, Fang S, Pang B, et al. Molecular pathology and clinical implications of diffuse glioma. Chin Med J (Engl). 2022;135(24):2914-25. https://doi.org/10.1097/CM9.0000000000002446 Bliesener Y, Lebel RM, Acharya J, Frayne R, Nayak KS. Pseudo Test-Retest Evaluation of Millimeter-Resolution Whole-Brain Dynamic Contrast-enhanced MRI in Patients with High-Grade Glioma. Radiology. 2021;300(2):410-20. https://doi.org/10.1148/radiol.2021203628 Gerritsen JKW, Zwarthoed RH, Kilgallon JL, et al. Impact of maximal extent of resection on postoperative deficits, patient functioning, and survival within clinically important glioblastoma subgroups. Neuro Oncol. 2023;25(5):958-72. https://doi.org/10.1093/neuonc/noac255 Della Pepa GM, Ius T, La Rocca G, et al. 5-Aminolevulinic Acid and Contrast-Enhanced Ultrasound: The Combination of the Two Techniques to Optimize the Extent of Resection in Glioblastoma Surgery. Neurosurgery. 2020;86(6):529-40. https://doi.org/10.1093/neuros/nyaa037 Wang Z, Zhang M, Chi S, Zhu M, Wang C, Liu Z. Brain Tumor Cell Membrane-Coated Lanthanide-Doped Nanoparticles for NIR-IIb Luminescence Imaging and Surgical Navigation of Glioma. Adv Healthc Mater. 2022;11(16):e2200521. https://doi.org/10.1002/adhm.202200521 Croci D, Santalla Méndez R, Temme S, et al. Multispectral fluorine-19 MRI enables longitudinal and noninvasive monitoring of tumor-associated macrophages. Sci Transl Med. 2022;14(667):eabo2952. https://doi.org/10.1126/scitranslmed.abo2952 Do MH, Shi W, Ji L, et al. Reprogramming tumor-associated macrophages to outcompete endovascular endothelial progenitor cells and suppress tumor neoangiogenesis. Immunity. 2023;56(11):2555-69.e5. https://doi.org/10.1016/j.immuni.2023.10.010 Folkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med. 1971;285(21):1182-6. https://doi.org/10.1056/NEJM197111182852108 Mathur R, Wang Q, Schupp PG, et al. Glioblastoma evolution and heterogeneity from a 3D whole-tumor perspective. Cell. 2024;187(2):446-63.e16. https://doi.org/10.1016/j.cell.2023.12.013 Wang X, Xu T, Niu C. Vascularization ability of glioma stem cells in different three-dimensional microenvironments. Regen Biomater. 2023;11:rbad094. https://doi.org/10.1093/rb/rbad094 Hingot V, Brodin C, Lebrun F, et al. Early Ultrafast Ultrasound Imaging of Cerebral Perfusion correlates with Ischemic Stroke outcomes and responses to treatment in Mice. Theranostics. 2020;10(17):7480-91. https://doi.org/10.7150/thno.44233 Errico C, Pierre J, Pezet S, et al. Ultrafast ultrasound localization microscopy for deep super-resolution vascular imaging. Nature. 2015;527(7579):499-502. https://doi.org/10.1038/nature16066 Demené C, Robin J, Dizeux A, et al. Transcranial ultrafast ultrasound localization microscopy of brain vasculature in patients. Nat Biomed Eng. 2021;5(3):219-28. https://doi.org/10.1038/s41551-021-00697-x Wang R, Dai W, Gong J, et al. Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients. J Hematol Oncol. 2022;15(1):11. https://doi.org/10.1186/s13045-022-01225-3 Niu X, Yang Y, Zhou X, et al. A prognostic nomogram for patients with newly diagnosed adult thalamic glioma in a surgical cohort. Neuro Oncol. 2021;23(2):337-8. https://doi.org/10.1093/neuonc/noaa268 Linazi G, Maimaiti A, Abulaiti Z, et al. Prognostic value of anoikis-related genes revealed using multi-omics analysis and machine learning based on lower-grade glioma features and tumor immune microenvironment. Heliyon. 2024;10(17):e36989. https://doi.org/10.1016/j.heliyon.2024.e36989 Lowerison MR, Vaithiyalingam Chandra Sekaran N, Dong Z, et al. Super-Resolution Ultrasound Reveals Cerebrovascular Impairment in a Mouse Model of Alzheimer’s Disease. J Neurosci. 2024;44(9):e1251232024. https://doi.org/10.1523/JNEUROSCI.1251-23.2024 Nyúl-Tóth Á, Negri S, Sanford M, et al. Novel intravital approaches to quantify deep vascular structure and perfusion in the aging mouse brain using ultrasound localization microscopy (ULM). J Cereb Blood Flow Metab. 2024;44(11):1378-96. https://doi.org/10.1177/0271678X241260526 Sun J, Cheng W, Guo S, et al. A ratiometric SERS strategy for the prediction of cancer cell proportion and guidance of glioma surgical resection. Biosens Bioelectron. 2024;261:116475. https://doi.org/10.1016/j.bios.2024.116475 He A, Xu L, Yang X, Gu Z, Cai Y, Zhou H. Risk factors for surgical compliance and impact on the survival of patients with glioma: a population-based propensity score-matched study. J Cancer Res Clin Oncol. 2023;149(16):14797-815. https://doi.org/10.1007/s00432-023-05261-5 Del Bene M, Perin A, Casali C, et al. Advanced Ultrasound Imaging in Glioma Surgery: Beyond Gray-Scale B-mode. Front Oncol. 2018;8:576. https://doi.org/10.3389/fonc.2018.00576 Mei X, Chen YS, Zhang QP, et al. Association between glioblastoma cell-derived vessels and poor prognosis of the patients. Cancer Commun (Lond). 2020;40(5):211-21. https://doi.org/10.1002/cac2.12026 Yin J, Dong F, An J, et al. Pattern recognition of microcirculation with super-resolution ultrasound imaging provides markers for early tumor response to anti-angiogenic therapy. Theranostics. 2024;14(3):1312-24. https://doi.org/10.7150/thno.89306 Jiang H, Zuo M, Li W, Zhuo S, Wu P, An C. Multimodal imaging-based prediction of recurrence for unresectable HCC after downstage and resection-cohort study. Int J Surg. 2024;110(9):5672-84. https://doi.org/10.1097/JS9.0000000000001752 Zhu C, Zou C, Guan G, et al. Development and validation of an interferon signature predicting prognosis and treatment response for glioblastoma. Oncoimmunology. 2019;8(9):e1621677. https://doi.org/10.1080/2162402X.2019.1621677 Ellingson BM, Hagiwara A, Morris CJ, et al. Depth of Radiographic Response and Time to Tumor Regrowth Predicts Overall Survival Following Anti-VEGF Therapy in Recurrent Glioblastoma. Clin Cancer Res. 2023;29(20):4186-95. https://doi.org/10.1158/1078-0432.CCR-23-1235 Spinelli C, Adnani L, Meehan B, et al. Mesenchymal glioma stem cells trigger vasectasia-distinct neovascularization process stimulated by extracellular vesicles carrying EGFR. Nat Commun. 2024;15(1):2865. https://doi.org/10.1038/s41467-024-46597-x Zhu X, Huang Q, DiSpirito A, et al. Real-time whole-brain imaging of hemodynamics and oxygenation at micro-vessel resolution with ultrafast wide-field photoacoustic microscopy. Light Sci Appl. 2022;11(1):138. https://doi.org/10.1038/s41377-022-00836-2 Rosberg R, Smolag KI, Sjölund J, et al. Hypoxia-induced complement component 3 promotes aggressive tumor growth in the glioblastoma microenvironment. JCI Insight. 2024;9(19):e179854. https://doi.org/10.1172/jci.insight.179854 Stadlbauer A, Zimmermann M, Kitzwögerer M, et al. MR Imaging-derived Oxygen Metabolism and Neovascularization Characterization for Grading and IDH Gene Mutation Detection of Gliomas. Radiology. 2017;283(3):799-809. https://doi.org/10.1148/radiol.2016161422 McCall JR, DeRuiter R, Ross M, et al. Longitudinal 3-D Visualization of Microvascular Disruption and Perfusion Changes in Mice During the Evolution of Glioblastoma Using Super-Resolution Ultrasound. IEEE Trans Ultrason Ferroelectr Freq Control. 2023;70(11):1401-16. https://doi.org/10.1109/TUFFC.2023.3320034 Frost HR, Amos CI. Gene set selection via LASSO penalized regression (SLPR). Nucleic Acids Res. 2017;45(12):e114. https://doi.org/10.1093/nar/gkx291 Hormuth DA, Wong AH, Yankeelov TE. Mathematical modeling of tumor vascularization: understanding the interplay of angiogenesis and vascular remodeling. Front Physiol. 2021;12:725-36. Liu G, Yang J, Wang J, et al. Extended axial imaging range: Widefield swept-source optical coherence tomography angiography. J Biophotonics. 2017;10(11):1464-72. https://doi.org/10.1002/jbio.201600325 Schupper AJ, Hadjipanayis CG. Novel approaches to targeting gliomas at the leading/cutting edge. J Neurosurg. 2023;139(3):760-8. https://doi.org/10.3171/2023.1.JNS221798 Huang R, Wu H, Lu X, Sun X. Clinical characteristics and prognostic factors of solitary and multiple adult gliomas: a retrospective study based on propensity score matching. Eur Rev Med Pharmacol Sci. 2023;27(21):10481-98. https://doi.org/10.26355/eurrev_202311_34325 Staub-Bartelt F, Suresh Babu MP, Szelényi A, Rapp M, Sabel M. Establishment of Different Intraoperative Monitoring and Mapping Techniques and Their Impact on Survival, Extent of Resection, and Clinical Outcome in Patients with High-Grade Gliomas-A Series of 631 Patients in 14 Years. Cancers (Basel). 2024;16(5):926. https://doi.org/10.3390/cancers16050926 Yan J, Huang B, Tonko J, et al. Transthoracic ultrasound localization microscopy of myocardial vasculature in patients. Nat Biomed Eng. 2024;8(6):689-700. https://doi.org/10.1038/s41551-024-01206-6 Bodard S, Denis L, Chabouh G, et al. Visualization of Renal Glomeruli in Human Native Kidneys With Sensing Ultrasound Localization Microscopy. Invest Radiol. 2024;59(8):561-8. https://doi.org/10.1097/RLI.0000000000001061 Additional Declarations No competing interests reported. Supplementary Files SupplementaryMaterials.docx Cite Share Download PDF Status: Under Review Version 1 posted Reviewers invited by journal 14 May, 2025 Editor invited by journal 23 Apr, 2025 Editor assigned by journal 23 Apr, 2025 Submission checks completed at journal 23 Apr, 2025 First submitted to journal 16 Apr, 2025 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-6464545","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":457771519,"identity":"09dac2d9-e66c-4f44-8183-0e9266a56390","order_by":0,"name":"Xing Hu","email":"","orcid":"","institution":"Department of Ultrasound, Huashan Hospital, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xing","middleName":"","lastName":"Hu","suffix":""},{"id":457771520,"identity":"8c1984e0-63b2-4ee4-b6d8-86c1d2becd7b","order_by":1,"name":"Gaobo Zhang","email":"","orcid":"","institution":"Department of Biomedical Engineerning, School of Information Sciencee, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Gaobo","middleName":"","lastName":"Zhang","suffix":""},{"id":457771521,"identity":"cd9ebc20-577d-4324-800c-7b945210390a","order_by":2,"name":"Xiandi Zhang","email":"","orcid":"","institution":"Department of Ultrasound, Huashan Hospital, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xiandi","middleName":"","lastName":"Zhang","suffix":""},{"id":457771522,"identity":"09be2079-faed-44d6-8c2c-561036a3900b","order_by":3,"name":"Yong Wang","email":"","orcid":"","institution":"Department of Ultrasound, Huashan Hospital, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Yong","middleName":"","lastName":"Wang","suffix":""},{"id":457771523,"identity":"c78e36df-1353-4dd1-a6bf-313c15044129","order_by":4,"name":"Xin Liu","email":"","orcid":"","institution":"Academy for Engineering and Technology, Fudan University","correspondingAuthor":false,"prefix":"","firstName":"Xin","middleName":"","lastName":"Liu","suffix":""},{"id":457771524,"identity":"dc376a67-cce7-4bb9-a7dc-708d98fbb510","order_by":5,"name":"Hong Ding","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9klEQVRIiWNgGAWjYDCCw0DM2AAk2BsfMPCAhRKI1cJz2IBILQdgWiSSidTCd5z38AvGHTZ58pGP2STe1Nxh4GfPMWD4uQO3FsnDfGkWjGfSig1vJ7NJzjn2jEGy540BY+8Z3FoMDvOYGTC2HU7cODv/mDQP22EGgxs5BsyMbcRomXmYTZrn32EGeyK0GD8AaZkvwcwmzdsGtEWCgBZJoC0MjG1piRt4kpkt5/Yd5pE486zgYC8eLXznzxh/YGyzSZzffpjxxptvh+X425M3PviJRwsQsEn/AbnwAIQHjpoDeDUwMDB/AJHyDQSUjYJRMApGwcgFAOPZUVovXYc8AAAAAElFTkSuQmCC","orcid":"","institution":"Department of Ultrasound, Huashan Hospital, Fudan University","correspondingAuthor":true,"prefix":"","firstName":"Hong","middleName":"","lastName":"Ding","suffix":""}],"badges":[],"createdAt":"2025-04-16 14:38:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6464545/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6464545/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83042583,"identity":"90b20312-d38f-41cf-968e-f31bb35b7570","added_by":"auto","created_at":"2025-05-19 10:56:23","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":385740,"visible":true,"origin":"","legend":"\u003cp\u003eResearch flowchart\u003c/p\u003e","description":"","filename":"Fig.1.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/84f91e49825ed1ddbf1008bb.jpg"},{"id":83040745,"identity":"aaed64b4-b938-467c-8393-882615ef761b","added_by":"auto","created_at":"2025-05-19 10:40:23","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":642040,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea \u003c/strong\u003eIn situ modeling and jugular vein catheterization of rat glioblastoma.\u003cstrong\u003eb \u003c/strong\u003eData acquisition and image reconstruction.\u003cstrong\u003ec \u003c/strong\u003eDelineation of infiltration area.\u003c/p\u003e","description":"","filename":"Fig.2.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/72737a67ac054ebfb2bcb08c.jpg"},{"id":83040754,"identity":"a7df31c7-8e3e-43e8-a6ab-435ffbb6180f","added_by":"auto","created_at":"2025-05-19 10:40:23","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":349735,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea\u003c/strong\u003eOverview of H\u0026amp;E staining; \u003cstrong\u003eb\u003c/strong\u003e The proportion of tumor cell area in normal brain regions is 0%; c The proportion of tumor cell area in the infiltration zone is 17.755% (indicated by the yellow arrow); \u003cstrong\u003ed\u003c/strong\u003e The proportion of tumor cell area in the core region is approximately 98.911% (indicated by the yellow arrow).\u003c/p\u003e","description":"","filename":"Fig.3.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/97aaa3082665e79ee8c73bde.jpg"},{"id":83040749,"identity":"72e09e51-1542-4a90-9941-9e369434fc05","added_by":"auto","created_at":"2025-05-19 10:40:23","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":243632,"visible":true,"origin":"","legend":"\u003cp\u003eVariable selection in the LASSO binary logistic regression model. \u003cstrong\u003ea\u003c/strong\u003e Show the coefficient path plot based on the logarithmic λ sequence, validating the optimal λ parameter in the LASSO model. Six features with non-zero coefficients were selected using the optimal λ value. \u003cstrong\u003eb\u003c/strong\u003e The relationship curve between binomial deviance and Log(λ) was plotted, with a dashed vertical line representing the one standard error criterion.\u003c/p\u003e","description":"","filename":"Fig.4.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/61367e077d004b2da1f28c31.jpg"},{"id":83040755,"identity":"f2a1e7fe-7055-4267-8dd7-ef3ee7c72353","added_by":"auto","created_at":"2025-05-19 10:40:23","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":294802,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea \u003c/strong\u003eNomogram predicts the risk of peritumoral invasion, incorporating three factors into the model.\u003cstrong\u003e \u003c/strong\u003eROC validation of the nomogram predicting GBM invasion risk. The y-axis represents the true positive rate, and the x-axis represents the true negative rate of risk prediction. \u003cstrong\u003eb\u003c/strong\u003e Training set data. \u003cstrong\u003ec\u003c/strong\u003e Validation set data. \u003cstrong\u003ed \u003c/strong\u003eThe calibration curve of GBM infiltration risk prediction is displayed: the y-axis represents the actual diagnosis of infiltration, and the x-axis represents the infiltration risk predicted by the model. The diagonal dashed line represents the perfect prediction performance of the ideal model, while the solid line corresponds to the calibration results of the training set (blue) and validation set (red). The closer the solid line is to the diagonal, the better the prediction performance of the model. \u003cstrong\u003ee\u003c/strong\u003e The decision curve analysis of GBM infiltration risk prediction is shown. The y-axis measures net returns, with the thick solid line representing the assumption that all patients have no infiltration, and the thin solid line representing the assumption that all patients have infiltration. (Blue) from the training set: When the threshold probability is between 5% and 94%, the model provides clinical net benefits. (Red) from the validation set: When the threshold probability is between 1% and 99%, the model provides clinical net benefits.\u003c/p\u003e","description":"","filename":"Fig.5.tif.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/2443c89befc577c5b2415cab.jpg"},{"id":83042857,"identity":"8e023924-0079-4435-973f-8c47bbe999f9","added_by":"auto","created_at":"2025-05-19 11:04:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2998828,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/3d19e629-d734-4a20-9417-3a5c5dfd773e.pdf"},{"id":83040751,"identity":"cda84525-2358-40d2-b2a1-819ca7db3203","added_by":"auto","created_at":"2025-05-19 10:40:23","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":32191,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryMaterials.docx","url":"https://assets-eu.researchsquare.com/files/rs-6464545/v1/9cc76312f4142c8e89d20d18.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Risk of Peritumoral Invasion in Rat Glioblastoma: Nomogram-based Ultrasound Localization Microscopy","fulltext":[{"header":"Introduction","content":"\u003cp\u003eGlioblastoma (GBM) is a highly malignant and aggressive primary brain tumor. Standard treatment generally involves maximal safe resection, followed by radiotherapy and chemotherapy [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. A major challenge is the diffuse invasion of surrounding tissue. Incomplete resection of invasive areas during surgery is a major cause of high recurrence and mortality rates due to residual tumor cells. Notably, 85% of recurrences occur at the surgical margins. Maximizing tumor removal while preserving neurological function is crucial [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. This presents significant challenges due to unclear imaging boundaries between invasive and normal tissues. Achieving a precise preoperative invasion risk assessment to guide intraoperative tumor resection, while preserving normal neurological function, is a key focus in glioma surgery [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eCurrent technologies are limited in accurately assessing the pathological features of noncontrast-enhanced tumor components, often resulting in missed small invasive foci during surgical resection or radiotherapy [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. Although MRI and CT provide good resolution and reduce brain shift effects, they require specific equipment standards and intraoperative conditions, which increase anesthesia time and safety risks. Furthermore, they do not provide real-time multiparametric imaging [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Invasive finger-like projections may not always be visible [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Fluorescence navigation uses selectively accumulating photosensitive agents to highlight tumor borders, improving resection rates and reducing residuals, while offering simplicity and repeatability in administration. However, agents such as 5-aminolevulinic acid rely on blood-brain barrier (BBB) disruption, limiting the discernibility of deeper structures. Indocyanine green bypasses the BBB but exhibits low signal intensity, and its spectral strength assessment is subjective. However, standardized quantitative measures for weakly fluorescent surrounding tissue are lacking, potentially leading to false negatives when distinguishing between low-grade tumors, inflammatory tissue, and edema [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. Emerging technologies, including Raman imaging, terahertz imaging, two-dimensional local correlation spectroscopy, ion mass spectrometry, and inflammation-specific imaging, demonstrate high sensitivity in detecting molecular markers or metabolites to delineate tumor boundaries. However, these techniques are operationally complex, involve small sample sizes, are time-consuming, and suffer from low signal-to-noise ratios [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. There is a need for highly accurate, rapid, noninvasive intraoperative tools to guide maximal tumor resection.\u003c/p\u003e \u003cp\u003eTumor cells frequently invade from enhanced regions to distant areas, with marginal zones exhibiting characteristics ranging from sparse tumor cell invasion to invasive islands. However, current imaging modalities typically reveal only dense tumor regions and fail to detect areas of low-density invasion [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Microvascular pathology is an early and common feature of GBM, characterized histologically by microvascular proliferation and necrosis [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Folkman (1971) hypothesized that tumor growth and spread depend on angiogenesis, suggesting that the formation of new blood vessels induced by tumors is essential for their invasion [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Cellular invasion often follows perivascular spaces, facilitating spread along blood vessels into surrounding tissue. The marginal zone is the primary site of tumor invasion. Developing invasive molecular models is challenging, primarily due to the brain\u0026rsquo;s intricate architecture and the diverse cellular composition in the tumor microenvironment, particularly considering the three-dimensional organization of tissue [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Therefore, detailed microanatomical studies, including thin-walled tissue and the vascular niche, are essential for understanding tumor biology [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Developing a reliable preoperative tool to visualize vascular anatomy and function, thus guiding surgical resection, is essential.\u003c/p\u003e \u003cp\u003e Ultrasound localization microscopy (ULM) has recently emerged as a novel microvascular imaging technology with significant potential for both structural and functional visualization. Unlike single-plane waves, ULM uses multi-angle coherent superposition to merge images from different angles, achieving uniform focusing at various depths and directions, greatly enhancing blood flow detection sensitivity [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. ULM quantifies microhemodynamics and perfusion across a broad dynamic range, effectively compensating for signal-to-noise ratio losses and resolution degradation caused by unfocused ultrasound beams. It addresses the trade-offs between spatial resolution and penetration depth in previous technologies [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Nomograms are practical tools for predicting event probabilities and guiding clinical decisions. Incorporating biomarkers enhances the precision and personalization of these predictive models [\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. However, few studies have reported using nomograms to predict the risk of peritumoral invasion in GBM. This study aims to create a nomogram using preoperative ULM data to help clinicians identify patients at high risk of GBM invasion. By enabling early detection and targeted interventions for residual lesions post-surgery, this approach aims to improve patient outcomes.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003e This study was conducted with approval from the Animal Ethics Committee of Fudan University (approval number: 202408008S) and was in accordance with the declaration of Helsinki. The research workflow is shown in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eRat GBM Modeling\u003c/h2\u003e \u003cp\u003eThirty-six male Sprague-Dawley rats (6 weeks old, weighing 200\u0026ndash;250 g) were used to create the GBM models. The rats were housed under controlled conditions, including regulated temperature, humidity, and a 12-hour light/dark cycle. The rats had free access to rodent food and sterilized tap water. The C6 glioma cell line, obtained from the National Institutes of Health, was cultured in Dulbecco\u0026rsquo;s Modified Eagle Medium with 10% fetal bovine serum and 5% penicillin-streptomycin.\u003c/p\u003e \u003cp\u003eA C6 cell suspension (5 \u0026times; 10\u003csup\u003e5\u003c/sup\u003e cells per 10 \u0026micro;L) was injected into each rat to establish the model. Anesthesia was induced by intraperitoneally administering a ketamine-xylazine mixture (60 mg/kg). The rats were secured in a stereotactic frame and stabilized with ear bars and a nose clamp. A 25-gauge flat-tip syringe (25G; Hamilton Company, Reno, NV, USA) was used to inject 20 \u0026micro;L of the cell suspension into the caudate nucleus. A 0.5 cm longitudinal scalp incision was made at the intersection of the rat's medial canthus and sagittal suture. The skin was retracted to expose the bregma and sagittal suture. The needle was displaced laterally by 3 mm from the intersection point, then advanced 1 mm toward the head to reach a depth of approximately 5.5 mm, at an injection rate of 4 \u0026micro;L/min. After a 2-minute retention, the needle was slowly withdrawn (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). The injection site was treated with 5% penicillin-streptomycin, and the bone defect was sealed with sterile bone wax. The scalp was sutured, and the rats were returned to their cages for recovery.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eMRI Pre-Scanning\u003c/h3\u003e\n\u003cp\u003eMRI scanning was performed one day before the super-resolution experiment. The rats were kept under light anesthesia with isoflurane (1-1.25% oxygen concentration) and placed on a heating pad for warmth. Imaging was performed via 3.0T MRI (uMR870, United Imaging Healthcare, 2022, China) in the coronal plane. The scanning sequences included T2-weighted imaging (T2WI), fluid-attenuated inversion recovery (FLAIR), and diffusion-weighted imaging (DWI). After a tail vein injection of 0.6 mmol/kg gadobutrol, contrast-enhanced scans were performed, with scanning parameters listed in Supplementary Table\u0026nbsp;1.\u003c/p\u003e\n\u003ch3\u003eSuper-Resolution Data Acquisition\u003c/h3\u003e\n\u003cp\u003eSurgical Site Preparation. The rats were anesthetized in a gas induction chamber with 4% isoflurane (R510-22-10, RWD) and maintained at a 1% concentration. After hair removal from the neck, the skin was disinfected with iodine, followed by 70% ethanol. A PE10 medical-grade polyethylene catheter (inner diameter: 0.28 mm, outer diameter: 0.61 mm) was used to cannulate the right jugular vein. Ophthalmic solution was applied to the rats' eyes to prevent dryness while they were positioned prone, and their heads were secured in a stereotactic frame. After incising the scalp, the periosteum was carefully removed using a scalpel. A craniotome (78001, RWD) was used to thin the skull until the vascular system became visible. A microscope (DOM-1001, RWD) was used to preserve larger vessels and the superior sagittal sinus, while micro-forceps retracted the skull to expose the dura mater. Hemostatic sponges were applied to control minor bleeding from superficial injuries, and sterile saline was used to prevent thermal damage and reduce bleeding and swelling.\u003c/p\u003e \u003cp\u003eMulti-Angle Plane Radiofrequency Data Acquisition. The L22-14vX LF transducer (15.625 MHz, MS200, Visualsonics Ltd., Toronto, ON, Canada) was securely attached to the stereotactic imaging frame using a high-precision linear stage (VT-80, Physik Instrumente, Auburn, MA, USA). The transducer was oriented to produce coronal anatomical brain sections. The tumor site was identified and aligned with the MRI coronal plane using the Mindray M10 Portable Ultrasound System (Mindray, Shenzhen, China). The Verasonics Vantage 256 system (Verasonics Ltd., Kirkland, WA, USA) and the L22-14vX-LF transducer were aligned for motion adjustments in 1 mm increments. A programmable microinjection pump (R462, RWD) infused diluted SonoVue (Bracco Imaging, Massy, France) at 80 \u0026micro;L/min. The imaging sequence included multiple 5-angle plane waves (-5\u0026deg;, -2.5\u0026deg;, 0\u0026deg;, 2.5\u0026deg;, 5\u0026deg;) [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Data collection was synchronized with a function generator, and all processing was performed in the MATLAB environment (The MathWorks, Natick, Massachusetts, USA, R2020a). The number of scanned sections per rat was determined by tumor size, as shown in Supplementary Table\u0026nbsp;2.\u003c/p\u003e \u003cp\u003eSingular value decomposition filtering was used to extract microbubble signals from the tissue background of each in-phase/quadrature dataset. A threshold was chosen at the inflection point of the singular value curve to optimize clutter filtering. The centroid of the microbubbles was determined using the radial symmetry algorithm in MATLAB. The Hungarian algorithm was used for frame-to-frame centroid pairing and trajectory estimation. A minimum trajectory length of 15 frames for microbubbles was set, followed by accumulating each acquisition to reconstruct the vascular system in ULM (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). Vessel reconstruction in superficial cortical areas was limited due to significant echo scattering from the ear bars and acoustic shadows from other skull regions [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eHistopathology\u003c/h3\u003e\n\u003cp\u003eAfter collecting the ULM data, the rats were humanely euthanized. Their brains were quickly extracted and preserved in formalin. To minimize analytical errors from sample position deviations, tumors were coronally sectioned into 1 mm thick slices using the Rat Brain Matrix (WPI-RBMA-600S), and guided by the Mindray M10 Portable Ultrasound System. This approach ensured that the ULM sections precisely matched the sample positions for histological staining analysis. Hematoxylin-eosin (H\u0026amp;E) staining was used to clearly visualize tumor cells. Quantitative analysis of H\u0026amp;E-stained sections was performed using digital image analysis software (Fiji/Image J v.1.54f), which included automated cell counting and density distribution analysis. This analysis quantified variations in tumor cell density, distinguishing between tumor regions (\u0026ge;\u0026thinsp;90% tumor cell occupancy), invasive regions (\u0026lt;\u0026thinsp;90% tumor cell occupancy), and normal brain regions (no tumor cells) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eImage Post-Processing\u003c/h3\u003e\n\u003cp\u003eFor each section, the T2, DWI, and T1CE sequences were co-registered with the FLAIR sequence. FSL bias field correction tools were used to correct magnetic field inhomogeneities, and the FLIRT linear registration tool enabled rigid image registration. In the T1CE sequence, the enhanced tumor region, including necrosis, was identified as the region of interest (ROI), while in the FLAIR sequence, high-signal and abnormal signals (such as tumors and edema) were designated as ROIs. Based on the image registration results, ROIs were manually delineated and mapped to the super-resolved images, encompassing both invasive regions and normal brain tissue (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). Image features were extracted using scale-invariant feature transform and wavelet transform algorithms in Fiji/ImageJ v.1.54f (Wayne Rasband, NIH, USA), with segmentation performed using the Trainable Weka Segmentation tool.\u003c/p\u003e \u003cp\u003eVascular segmentation was carried out in a supervised, semi-automated manner. Random forest classifiers were used to train models for image pre-collection, deep learning-based vascular segmentation, feature extraction, and scanning modulation, effectively representing the three-dimensional volume of interest in two-dimensional depictions. Vascular features, including area, skeleton, and midpoint, were extracted from binarized images using deep learning methods. The out-of-bag (OOB) error was defined as the error rate of samples not used in training decision trees, enabling the random forest to use OOB samples for unbiased internal estimation. To improve model generalization, the OOB error was maintained below 5%.\u003c/p\u003e \u003cp\u003eCurvature is defined as follows (1):\u003c/p\u003e \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"84\" height=\"66\"\u003e Lc represents the actual path length of a vessel segment, and L represents the linear distance between the segment endpoints.\u003c/p\u003e \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"193\" height=\"57\"\u003e\u003c/p\u003e \u003cp\u003eThe fractal dimension (FD) of vascular structures quantifies the complexity and branching patterns of vascular networks, reflecting the degree of branching and twisting. The box-counting method is applied for segmentation across multiple scales, followed by linear regression analysis to derive the FD. To accurately capture the fractal characteristics of the images, an appropriate two-dimensional Euclidean space (R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;\u0026gt;\u0026thinsp;0.99) was selected, ensuring a good fit between the fitted model and the actual data (2).\u003c/p\u003e \u003cp\u003e\u003cimg src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAScAAAA2CAYAAACFp1/LAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAADsMAAA7DAcdvqGQAAArXSURBVHhe7Z1JaBRPG4fLDz0ILrgc3NHEBU+uuamoiSsBPShxjZ4U9aDgCh7Eg+ISD4IrehAVFxQRIq4oop5cYvQiolFwiQrRg3rPN0+l3/nXdHoyPZnOTGd8Hyi6a+nq6p7u37xvdXVXl+YERlEUJWb8z1sqiqLEChUnRVFiiYqToiixRMVJUZRYouKkKEosUXFSFCWWqDgpihJLVJyUgvLt2zdTWlpqbty44aVEi9TfpUsXs2rVKi+1NYcOHbJlCP62UMesWbPsMh1hyihZwiBMRSkEjY2NzSUlJQwCbq6trfVSo+fPnz/N5eXlNrDup66urrlfv36B+bSLNtLWTFDP6NGj7VLJHbWclIIxcOBAc/XqVZMQBi+lY0gIjunWrZsXS+Xv379m69at5ufPn2bTpk2mR48eXk6LNXT48GHz5MkT29ZMTJgwwVy6dMmsXbtWLagIUHFSip4XL16YYcOGmY8fP1qhctm1a5cVFQRy8ODBXmoLO3bsMHPnzg0lTAJ1VVVV2W2V3FBxUmIFlkxFRUWy/4d10gT6gyRPAn1KbVkqP378MAsWLDBdu3Y1379/91Jb6ho0aJAZMGCAGT9+vBk1apSX02I1vXnzxsycOdNLaeHly5emf//+yX1iKRF3+6nYhm3VesoNFSclNiBCCxcuNPPmzaMvNGnlkEYeN/vGjRtNbW2tza+pqTHl5eXm1atXaa0btvny5Yu1aHr37u2ltojMuXPnzPLly82tW7fsPl2XDmsLVw/hcqGee/fumcmTJ5v3799ba6ukpCTF6mIbtqUOpf2oOCmx4eHDh9b1WrZsmY0jFgcPHjT19fU2D6sH60eEAAvl9+/frVw1F7aZMWOGXUcwvn79aoVu37591vVCvD5//tzKQnr79q231poHDx5YoUMgp06dap4+fWpFy09bdSiZUXFSYkPQzYwVghCQh9s1dOhQKw6A0PTq1cv07NnTxoPAemE7yowYMcLWQz/TypUrraBQF3W6Ll0mXr9+bQVy2rRpXorSEag4KbFhzJgx5sOHD4HuEHlYUjxR27Jli+1rwsXDNXPdMRdxBV3xunbtml1WVlbafFw6hMZfB/sLgvoaGxvNmjVrvJT0pKtDCYeKkxIbpk+fbvuQEB1EAC5cuGAtHvJI49E+4oBL1dDQ0OaTtLq6OpuP8BAQoe7du5vdu3fbfDqxcRkXL15s4y6TJk2yT/DcDnRAOOlEb2u/bMO21KHkQOJHVpSC4A7CZBAkgxcZBMlgSNII7sBIt7wbampqbL7gLyf5p06dCtxHUB1QXV3dKp20TANG2YZySm6oOCmdBkTl9OnTXqwFxKasrCzUCO5soU7qzmbEN9tUVFR0SHv+NSJz6xhF6x9/IoG8fLJ+/Xpz4MABL5YeyhGEkSNH5r2tLpwr+lziAOeF9gicTwYkFpKTJ0+aixcv2r4igQ5tBj1mM1AyLNTJPpcsWWKHHmQCt3PKlCnWLe2I9sQF7hG5b7gu5D7330s544lUZFBlQ0ODF2tufvz4cXPigvJiuUG9+/fv92LBkE8bMpWjTZRbt26dl6L4cS8PzuecOXO8WGEI645FTRhr6F+xmLhvSktL7Tr3ttxn3Jv8HnKvk+deP+0h8tlXUFA6KhmYFjUo8/Dhw822bdu8lGCyKQfHjh2zSyUVfsuILw+lk9PWNeG/7xg9/+jRo3bfXx36tM418wBTjzQOEHPQ7wq6UFbScXXY7vjx42b79u1JkxE3Q8qEcePA3acf2RewP+qU9rJMt62UIYhbyA9D+1iS7pq5Ygq75jH7deuVuARpl6S7+ZLn4p4/f8gWRmD37ds36dbJMXEc8hvs3LnTlmO/lJXzoBQP/N4JT8OLBcM7jALu8N27dwOvz1BgOUUJVbpBwCUg7rpRpIkZmG6d8rINSzEjMRtd81LWwS3nIqan4LaH7YmL68g6gXaIiUq7pKy0j7JSh5STJUHyWKdukHpYuvkEwT0e9iV5Uk7y3XMVNW57OE5pN7BOaGpqSh7viRMnbF5VVVXyuJTiIdO15l4fAtdBe6/PDrGccOsSdaeo7O3bt+3LkvJqApCGuvLPe+fOHS/V2HXSAZMwyCyk45F3m1BzXiEIA5/ncNvkrlOXgFlKXuKGTLYDaC+4VhAdslh0HIO0g3EwiR/EJH6slLY3NjZ6ay1WGvVJPudMwOpw90EbqIt0Kee299OnT97af0RpOaWDVz4Yz8NvAXwqBCZOnBjq3zKobcUYigWuuSFDhnixVLgP07lvQddnGDrUrUvXWEHcIYSMm88l08Utbg0k/rntslAgRByDhEz9bYgS5Wh/uidgrvi0B7Z32+SGuBDUtmIMxQ5uPkTdz1zQEeJYHK7FICBUNTU1XqxFlf2IFZSp09sFf5h9ivDhDxOXk9sesB6WLl3qxVramklYxSriwnUtRgFLhPPitguxEQsln/CybCbClFE6P1y3fOHBBWueTm+5D6UP1cXth8qKxA0SCW4/C8HvZ+KPSl7ixrNpCXFJpkmfD/WApBOkLrcviDpkXbZlH/7+Ij/uPilPHKQOgluHu04Z9zikrW6drLvngrhbN+1266B+kDhlwT0+guCWc/cbdKy5IHXTf1RfX9/cp0+f5H6Is04a61KWpb+sUjzwe/IbC+69IcHNB65Tud+zJfKhBIqiFC/SDRMGrCrejczUvZMOFSdFUUKD4KxevTpjnyhdGwmrKac+t4L2OSmKwKsfXMz+aZlygTr90zXx6guf/vV//tcPXy5wt0tH0D6KGfo9z5w5E9i35DJ79uychAlUnJQU+GeUQZRlZWWBDyOAzvqoOsK5sbnoMz1IyAZEjjrPnj2bfM+N/YwbN87cv3/fxtMhQhPm/TjKcI74pEuY9++KAc5rJlct16fNoOKkpMBnb/l87ZEjR+yQh3RPQ7EWNmzY4MX+G/XO0IibN29agSOEeRLKDR7lFFGIS9CUTqw3NDSkPAkOgm82ZfMtJr6oyXHqlFDRouKkpDB//nw7sHTv3r1JseAmR3jcwIwjly9ftnnAh90w4+mPWLFihXn+/Ln59euXfcycb9ozpZNLtuIECJROCRUtKk5KKzDJ3QF1mPEIjxuamprszUge0F+EaDGt9/nz522fA5ZTe7+zLX1DIob+PiLcNsmTQBtwrYKmdApLOpdOp4TKPypOSgq8vMu0R5lgeqSjR496MWP27NljRevZs2fW+kLgsJzc13/Cggi1d4ooJj2gL8w/pVNY0llNOiVU/lFxUlKgwzhMPxGiE1UfkZ9cpojKdTqmtlw63F2dEip/qDgpsaMjpogKQzqXTtApofKLipMSO3KZIiqX6ZjcCTj9IFx8VUKnhMofKk5K7MhliihcMtxN/5ROLlhAQbx79y7t5JoIpU4JlV9UnJSCg9gsWrTIdiYzFAGRuH79uhUjBAHriMkvScM6AiwryZPAk0LEY+zYsUmXz4X98LSNgZkMxOSTsjJwko52XMZ0ruGVK1cC57dzYZ/suy0BU7Ig8c+jKJ2KTFNEEbKd0ilX2Oe/MMFBPlHLSel0ZJoiipDNlE65gkXGeK9inxIq36g4KZ0O+RwwLpi4dLB582a7BB7xM+yA12+k36ojoO7q6mo7Ur6ystJLVaJAP5miKEosUctJUZRYouKkKEosUXFSFCWWqDgpihJLVJwURYklKk6KosQSFSdFUWKJipOiKDHEmP8DOPNpwRmuLv8AAAAASUVORK5CYII=\" width=\"295\" height=\"54\"\u003e\u003c/p\u003e \u003cp\u003eHere, r represents the grid scale, and M(r) denotes the minimum number of grids required to cover the fractal.\u003c/p\u003e \u003cp\u003eAssuming the mean vector of the vessel orientation is \u003cimg src=\"data:image/png;base64,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\" width=\"17\" height=\"23\"\u003e, the orientation variance (OV) of the vessels is calculated as follows (3):\u003c/p\u003e \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\" width=\"299\" height=\"51\"\u003e\u003c/p\u003e \u003cp\u003eLet N represent the total count of blood vessels, \u003cem\u003e\u0026theta;\u003csub\u003ei\u003c/sub\u003e\u003c/em\u003e denote the direction of the i-th data point, and \u003cimg src=\"data:image/png;base64,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\" width=\"17\" height=\"23\"\u003e be the average angle across all directions.\u003c/p\u003e \u003cp\u003eBlood flow reflects the functional perfusion status based on Poiseuille\u0026rsquo;s flow principles (4):\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eQ\u0026thinsp;=\u0026thinsp;A\u0026times;V (4)\u003c/h2\u003e \u003cp\u003eWhere Q represents blood flow, A denotes the vessel\u0026rsquo;s cross-sectional area, and V indicates blood flow velocity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eThe samples were randomly divided into two groups: one for training the predictive model and the other for evaluation. Continuous variables were analyzed using a t-test. LASSO regression, conducted with the 'glmnet' package in R, was used to select relevant microvascular parameters. A multifactorial regression model was then developed using the Akaike Information Criterion (AIC) for model selection, with significant parameters chosen as input variables. A personalized nomogram based on ULM scores was created from the training set to predict the risk of tumor invasion by aggregating scores and aligning them with a predefined scale.\u003c/p\u003e \u003cp\u003eThe model\u0026rsquo;s calibration was tested using the Hosmer-Lemeshow method, and its clinical significance was further evaluated using decision curve analysis. Classification performance was evaluated using metrics from the receiver operating characteristic (ROC) curve, including accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC). The optimal critical values for each indicator were determined using the Youden index (Youden index\u0026thinsp;=\u0026thinsp;sensitivity\u0026thinsp;+\u0026thinsp;specificity \u0026minus;\u0026thinsp;1) and validated in the validation group. The model\u0026rsquo;s validation was conducted using the validation set, with statistical significance set at P\u0026thinsp;=\u0026thinsp;0.05.\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eULM Baseline Features\u003c/h2\u003e \u003cp\u003eThis study analyzed 208 sections from rat GBM models. The training dataset consisted of 146 sections: 72 from invasion areas and 74 from normal brain areas. The validation dataset included 62 sections: 32 from invasion areas and 30 from normal brain areas. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the baseline characteristics of ULM for both datasets.\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\u003eBaseline characteristics of the ULM cohort\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"3\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eTraining set (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c8\" namest=\"c6\"\u003e \u003cp\u003eValidation set (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eInvasion\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNormal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003eInvasion\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNormal\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNo. of examinations\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e---\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiameter (\u0026micro;m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e47.12\u003c/p\u003e \u003cp\u003e(43.47, 50.76)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e29.26\u003c/p\u003e \u003cp\u003e(26.48, 32.03)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e47.97\u003c/p\u003e \u003cp\u003e(42.83, 53.10)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e29.55\u003c/p\u003e \u003cp\u003e(25.53, 33.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVascularity (%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.25\u003c/p\u003e \u003cp\u003e(22.79, 25.71)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14.97\u003c/p\u003e \u003cp\u003e(14.04, 15.90)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e24.67\u003c/p\u003e \u003cp\u003e(22.82, 26.52)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e11.67\u003c/p\u003e \u003cp\u003e(9.92, 13.41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCurvature\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.1837\u003c/p\u003e \u003cp\u003e(1.1753, 1.1921)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1.1531\u003c/p\u003e \u003cp\u003e(1.1479, 1.1583)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e1.1804\u003c/p\u003e \u003cp\u003e(1.1711, 1.1897)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1.1169\u003c/p\u003e \u003cp\u003e(1.1331, 1.1606)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBranch points (/mm\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e74.03\u003c/p\u003e \u003cp\u003e(64.80, 83.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e42.31\u003c/p\u003e \u003cp\u003e(36.97, 47.65)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e72.56\u003c/p\u003e \u003cp\u003e(59.84, 85.27)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e27.17\u003c/p\u003e \u003cp\u003e(20.45, 33.90)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBranches (/mm\u0026sup2;)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e189.23\u003c/p\u003e \u003cp\u003e(167.97, 210.50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e112.48\u003c/p\u003e \u003cp\u003e(99.94, 125.03)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e187.11\u003c/p\u003e \u003cp\u003e(157.73, 216.49)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e72.17\u003c/p\u003e \u003cp\u003e(56.90\u0026minus;87.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e15.4132\u003c/p\u003e \u003cp\u003e(15.2398\u0026ndash;15.5864)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e14.2518\u003c/p\u003e \u003cp\u003e(13.9825\u0026ndash;14.5211)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e15.4877\u003c/p\u003e \u003cp\u003e(15.2535, 15.7219)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e13.5946\u003c/p\u003e \u003cp\u003e(13.0999, 14.0894)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVelocity\u003c/p\u003e \u003cp\u003e(mm/s)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e29.92\u003c/p\u003e \u003cp\u003e(28.60, 31.24)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27.06\u003c/p\u003e \u003cp\u003e(26.27, 29.85)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e30.72\u003c/p\u003e \u003cp\u003e(28.78, 32.66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e28.29\u003c/p\u003e \u003cp\u003e(26.85, 29.72)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e0.060\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e24.57\u003c/p\u003e \u003cp\u003e(22.47, 26.67)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e16.91\u003c/p\u003e \u003cp\u003e(15.75, 18.07)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e25.80\u003c/p\u003e \u003cp\u003e(23.29, 27.70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e16.02\u003c/p\u003e \u003cp\u003e(13.74, 18.30)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlood flow (\u0026micro;L)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e51.50\u003c/p\u003e \u003cp\u003e(43.19, 59.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33.77\u003c/p\u003e \u003cp\u003e(23.38, 44.15)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c6\" namest=\"c5\"\u003e \u003cp\u003e57.63\u003c/p\u003e \u003cp\u003e(44.36, 70.91)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e40.05\u003c/p\u003e \u003cp\u003e(31.55, 48.55)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colspan=\"2\" nameend=\"c9\" namest=\"c8\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u003c/b\u003e\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eCI, confidence interval; FD, fractal dimension; OV, orientation variance\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eParameter filtering\u003c/p\u003e \u003cp\u003eIn the training set, GBM invasion (presence or absence) was used as the dependent variable to determine the optimal λ value. Six features with non-zero coefficients were selected: diameter, vascularity, curvature, FD, velocity, and OV (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). This approach enables variable selection in datasets with multiple features, retaining the most explanatory variables and creating a more interpretable model. The final model optimally balances generalizability and complexity.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eAnalysis of Peritumoral Invasion Risk Factors\u003c/h2\u003e \u003cp\u003eUnivariate analysis showed a positive correlation between peritumoral invasion and several variables, including diameter (odds ratios [OR], 1.081; 95% confidence interval [CI], 1.006 to 1.161; P\u0026thinsp;=\u0026thinsp;0.003), vascularity (OR, 1.289; 95% CI, 1.109 to 1.500; P\u0026thinsp;=\u0026thinsp;0.001), curvature (OR, 1.210; 95% CI, 0.949 to 1.543; P\u0026thinsp;=\u0026thinsp;0.125), FD (OR, 1.723; 95% CI, 0.773 to 3.838; P\u0026thinsp;=\u0026thinsp;0.183), velocity (OR, 1.041; 95% CI, 0.912 to 1.188; P\u0026thinsp;=\u0026thinsp;0.556), and OV (OR, 1.185; 95% CI, 1.043 to 1.347; P\u0026thinsp;=\u0026thinsp;0.009). The corresponding forest plot visualizes the 95% CI for these ORs (Supplementary Fig.\u0026nbsp;1). Multivariate analysis using AIC identified diameter (OR, 1.109; 95% CI, 1.039 to 1.184; P\u0026thinsp;=\u0026thinsp;0.002), vascularity (OR, 1.348; 95% CI, 1.180 to 1.541; P\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and OV (OR, 1.167; 95% CI, 1.033 to 1.318; P\u0026thinsp;=\u0026thinsp;0.013) as independent predictors of peritumoral invasion (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eUnivariate and multivariate logistic regression analysis of risk factors associated with invasion\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eCharacteristic\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c5\" namest=\"c2\"\u003e \u003cp\u003eUnivariable Analysis\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colspan=\"4\" nameend=\"c9\" namest=\"c6\"\u003e \u003cp\u003eMultivariable Model\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eβ-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eP-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eOR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eβ-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eP-value\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eOR (95% CI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiameter\u003c/p\u003e \u003cp\u003e(\u0026micro;m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.078\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.037\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.003\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.081\u003c/p\u003e \u003cp\u003e(1.006\u0026ndash;1.161)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.104\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.002\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.109\u003c/p\u003e \u003cp\u003e(1.039\u0026ndash;1.184)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVascularity\u003c/p\u003e \u003cp\u003e(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.254\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.077\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.289\u003c/p\u003e \u003cp\u003e(1.109\u0026minus;1.500)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.299\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.068\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.348\u003c/p\u003e \u003cp\u003e(1.180\u0026ndash;1.541)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCurvature\u003c/p\u003e \u003cp\u003e(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.191\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.124\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.125\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.210\u003c/p\u003e \u003cp\u003e(0.949\u0026ndash;1.543)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.544\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.409\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.183\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.723\u003c/p\u003e \u003cp\u003e(0.773\u0026ndash;3.838)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVelocity\u003c/p\u003e \u003cp\u003e(mm/s)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.040\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.068\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.556\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.041\u003c/p\u003e \u003cp\u003e(0.912\u0026ndash;1.188)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.170\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.065\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.009\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e1.185\u003c/p\u003e \u003cp\u003e(1.043\u0026ndash;1.347)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.154\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.062\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.013\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e1.167\u003c/p\u003e \u003cp\u003e(1.033\u0026ndash;1.318)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eFD, fractal dimension; SE, standard error; OR, odds ratio; OV, orientation variance\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eConstruction of the Prediction Model\u003c/h2\u003e \u003cp\u003eRisk factors from the training set, identified through multivariable logistic regression with AIC, were used to create a nomogram (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). The results indicated that vascularity had the greatest impact on peritumoral invasion, followed by OV and diameter. Each parameter was assigned a score, which was then summed to calculate a cumulative risk score. A straight line was used to visually estimate the peritumoral invasion probability for each slice, with predictions closely aligning with pathological findings (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eEvaluation and Validation of the Model\u003c/h2\u003e \u003cp\u003eThe training set achieved an AUC of 0.964 (95% CI, 0.933\u0026ndash;0.994), while the validation set achieved an AUC of 0.995 (95% CI, 0.984-1.000), as shown in the ROC curve (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb-c). The calibration curve shows that the model\u0026rsquo;s GBM peritumoral invasion risk predictions closely align with actual values in both the training and validation sets, as confirmed by the Hosmer-Lemeshow test (P \u0026gt; 0.05) (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ed). Decision curve analysis shows that the training set\u0026rsquo;s net benefit is higher across the 1\u0026ndash;99% threshold probability range, while the validation set shows a higher net benefit from 5\u0026ndash;94% (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ee). The model\u0026rsquo;s diagnostic performance is detailed in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eAssessment of model diagnostic performance\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGroup\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eParameter\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eAccuracy\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSensitivity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eSpecificity\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eNPV\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eAUC\u003c/p\u003e \u003cp\u003e(95% CI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003eOptimal cutoff value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eTraining\u003c/p\u003e \u003cp\u003eset\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDiameter\u003c/p\u003e \u003cp\u003e(\u0026micro;m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.897\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.923\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.877\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.857\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.934\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003cp\u003e(0.900\u0026minus;0.981)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e34.881\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVascularity\u003c/p\u003e \u003cp\u003e(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.845\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.852\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.788\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.743\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.882\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.899\u003c/p\u003e \u003cp\u003e(0.850\u0026ndash;0.948)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e20.703\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.747\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.746\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.729\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.671\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.816\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.806\u003c/p\u003e \u003cp\u003e(0.736\u0026ndash;0.875)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e20.520\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNomogram\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.884\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.843\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.908\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.864\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.964\u003c/p\u003e \u003cp\u003e(0.933, 0.994)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e--\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e \u003cp\u003eValidation\u003c/p\u003e \u003cp\u003eset\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDiameter\u003c/p\u003e \u003cp\u003e(\u0026micro;m)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.887\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.865\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.920\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.821\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.868\u003c/p\u003e \u003cp\u003e(0.774\u0026ndash;0.961)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e36.024\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVascularity\u003c/p\u003e \u003cp\u003e(%)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.935\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.917\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.962\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.971\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.963\u003c/p\u003e \u003cp\u003e(0.919\u0026minus;1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e16.523\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOV\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.790\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.889\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.714\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.706\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.893\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.870\u003c/p\u003e \u003cp\u003e(0.783\u0026ndash;0.956)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e22.702\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNomogram\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e0.936\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e0.929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e0.941\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e0.929\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e0.995\u003c/p\u003e \u003cp\u003e(0.984, 1.000)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e--\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003eAUC, area under the curve; NPV, negative predictive value; OV, orientation variance; PPV, positive predictive value\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eSurgery remains the primary treatment for GBM; however, residual lesions frequently occur as postoperative complications. Retrospective studies indicate that patients with residual lesions after surgery have a 5-year mortality rate 10%-20% higher than those with negative margins [\u003cspan additionalcitationids=\"CR25\" citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. Therefore, identifying patients at higher risk for residual lesions following surgery and implementing early intervention measures is critical. GBM\u0026rsquo;s invasive growth is associated with microvascular heterogeneity, and microvascular pathology is closely linked to the tumor\u0026rsquo;s invasive biological behavior. Newly formed microvessels exhibit abnormal morphological and functional characteristics, including increased branching, curvature, and enhanced perfusion [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. Previous studies suggest that changes in microvascular structure and function are early events in tumor progression, underscoring the need for early monitoring [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e]. This emphasizes the potential to improve the visualization of early invasion and enhance surgical resection precision. ULM provides in vivo, real-time, high-resolution imaging of microvessels, enabling the precise characterization of microvascular features in invasive regions [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. This capability is crucial for improving the accuracy of early predictions regarding invasion. Our study employs a microbubble tracking algorithm to quantitatively analyze microvascular structure and function, enabling clearer differentiation of microvessels smaller than 50 \u0026micro;m compared to traditional imaging techniques. This approach overcomes the limitations of conventional imaging in capturing sufficient hemodynamic information, enabling more accurate tumor invasion risk assessments and supporting precise surgical resection.\u003c/p\u003e \u003cp\u003eMost research on postoperative residual lesions depends on risk factor analyses. To visualize and quantify the risk of peritumoral invasion, we developed a risk assessment model for gliomas. Nomograms are useful tools for predicting peritumoral invasion and identifying high-risk patients, guiding treatment decisions. Consequently, the model, based on ULM\u0026rsquo;s multivariable logistic regression, enables accurate predictions of event probabilities and offers personalized risk assessments for potential invasion [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e]. Our results closely align with pathological findings, confirming the model\u0026rsquo;s utility in guiding precise management of residual lesions while minimizing damage to normal brain tissue.\u003c/p\u003e \u003cp\u003eIn this study, 208 samples were randomly split into training and validation sets. We identified vascularity, OV, and diameter as independent risk factors for peritumoral invasion using both univariate and multivariate logistic regression analyses. We developed a personalized nomogram incorporating multidimensional imaging scores based on these risk factors. Vascularity is crucial in peritumoral invasion, as GBM induces angiogenesis through the secretion of vascular endothelial growth factor (VEGF), promoting invasive growth. Additionally, the high permeability of newly formed vessels facilitates the spread of tumor cells into surrounding tissues, increasing the risk of invasion and metastasis [\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e]. Furthermore, the abnormal distribution of blood vessels is a defining characteristic of tumor invasion. Irregular vascular morphology reflects structural instability, creating a turbulent blood flow microenvironment that promotes tumor growth and migration, which encourages tumor cell invasion in these areas [\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Uneven oxygen supply and microenvironmental changes due to abnormal blood flow directionality may induce adaptive mutations in tumor cells, further enhancing their invasive and metastatic potential [\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e]. Finally, changes in vascular dilation provide a basis for tumor invasion, as larger vessel diameters increase the likelihood of tumor cells invading along vascular walls. Continuous VEGF stimulation promotes vascular dilation and enhances vascular permeability and plasticity, thereby creating migration pathways for tumor cell invasion [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e, \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHowever, the limited sample size may have led to some potential significances going undetected due to insufficient statistical power. LASSO regression simplifies the model by selecting features, thus helping prevent overfitting. However, it may overlook less influential factors that could be valuable in some biological contexts [\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e]. For example, branch number and branch point count may reflect vascular complexity or irregularity. Since OV partially captures this information, multicollinearity may reduce their role as independent predictors, thus obscuring their biological significance [\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e]. The microvascular network in the invasive region exhibits high biological heterogeneity, which may reduce the impact of parameters that vary significantly among individuals (e.g., FD and curvature) in multivariate analyses, potentially causing them to fail to reach significance [\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e]. The heterogeneity and complexity of the microenvironment pose challenges to model construction in statistical analyses. However, novel analytical methods, such as nonlinear regression or Bayesian models, may better capture the subtle effects of these variables in future studies.\u003c/p\u003e \u003cp\u003eOverall, this model assists clinicians in personalizing and accurately identifying peritumoral invasion risks, guiding GBM surgical resection with real-time visualization while minimizing damage. It has practical implications for developing personalized treatment plans, reducing postoperative residual risks, and improving survival outcomes [\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e]. For patients identified as high risk for invasion through the nomogram, maximally safe resection combined with postoperative radiotherapy, chemotherapy, tumor-treating fields, or molecularly targeted therapies may be considered to extend survival. Conversely, for low-risk patients, limited surgical resection and reduced radiation dosages could be used to manage the disease while minimizing side effects [\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e]. With technological advancements and broader dissemination, ULM is poised to become a routine tool in diagnosing and treating GBM, ultimately improving the quality of life for patients. Although this study focuses on GBM, the methodology has broader applicability and could be extended to other brain tumors and invasive systemic malignancies.\u003c/p\u003e \u003cp\u003eThis study also identifies a few limitations. First, despite ULM\u0026rsquo;s ability to discern fine structures, its complexity and high operational requirements necessitate specialized training and experienced personnel, limiting its widespread application, particularly in resource-limited healthcare settings. Second, although the rat model simulates human GBM, physiological differences remain, requiring further clinical trials to validate and refine its diagnostic thresholds. Third, while ULM provides micro-level advantages, its limited imaging range makes it difficult to cover the entire brain in a single scan, necessitating multiple imaging sessions that increase time costs and present challenges for data processing. Finally, this study lacks additional imaging and clinical data beyond ULM data. Future studies should integrate multimodal parameters for a more comprehensive assessment to enhance the model\u0026rsquo;s robustness [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e].\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn summary, this study presents a prospective evaluation of microcirculation quantification indices in a rat glioma model using ULM and establishes a nomogram based on three risk factors to predict preoperative peritumoral invasion risk, validated as a more accurate tool for assessing surgical resection. In the future, ULM is expected to become a key tool in diagnosing and treating GBM, with broad applications in intraoperative navigation and radiotherapy planning.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eAIC\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eAkaike information criterion\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eAUC\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eArea under the curve\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eBBB\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eBlood\u003cdel cite=\"mailto:Editor\" datetime=\"2025-04-14T18:13\"\u003e-\u003c/del\u003ebrain barrier\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eCI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eConfidence interval\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eDWI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eDiffusion-weighted imaging\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eFLAIR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eFluid-attenuated inversion recovery\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eGBM\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eGlioblastoma\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eH\u0026amp;E\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eHematoxylin-eosin\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eLASSO\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eLeast absolute shrinkage and selection operator\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eNPV\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eNegative predictive value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eOOB\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eOut-of-bag\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eOR\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eOdds ratios\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003ePPV\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003ePositive predictive value\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eROC\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eReceiver operating characteristic\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eROI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eRegion of interest\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eT2WI\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eT2-weighted imaging\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eULM\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eUltrasound localization microscopy\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 134px;\"\u003eVEGF\u003cbr\u003e\u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 441px;\"\u003eVascular endothelial growth factor\u003cbr\u003e\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conducted with approval from the Animal Ethics Committee of Fudan University (approval number: 202408008S) and was in accordance with the declaration of Helsinki.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analyzed during the current study are not publicly available but available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003cbr\u003e\u003c/strong\u003eThis study was supported by National Natural Science Foundation of China (grant number 82272017).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026rsquo; Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contribute to the study\u0026rsquo;s conception and design. Conception and\u0026nbsp;\u003c/p\u003e\n\u003cp\u003edesign of the study: Hu X, Zhang GB, Liu X, Ding H. Data acquisition: Hu X, Zhang GB, Zhang XD. Data analysis or interpretation: Hu X, Zhang GB, Yong Wang.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWriting of the first draft of the manuscript: Hu X, Zhang GB, Zhang XD. Critical revision of the manuscript: Yong Wang, Liu X, Ding H. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor details\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003csup\u003e1\u003c/sup\u003e\u003c/strong\u003e\u003cstrong\u003eDepartment of Ultrasound, Huashan Hospital, Fudan University, No. 12 Middle Urumqi Road, Shanghai 200040, China. \u003csup\u003e2\u003c/sup\u003e\u003c/strong\u003e\u003cstrong\u003eDepartment of Biomedical Engineering, School of Information Science and Technology, Fudan University, No. 2005 Songhu Road, Shanghai 200438, China. \u003csup\u003e3\u003c/sup\u003eAcademy for Engineering and Technology, Fudan University, No. 2005 Songhu Road, Shanghai 200438, China.\u003c/strong\u003e\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eCheng M, Hu C, Yao Z, et al. Harnessing Reconstructed Macrophage Modulation of Infiltration-Excluded Immune Microenvironments To Delineate Glioma Infiltrative Region. ACS Appl Mater Interfaces. 2023. https://doi.org/10.1021/acsami.2c16586\u003c/li\u003e\n\u003cli\u003eNie S, Zhu Y, Yang J, et al. Clinicopathologic analysis of microscopic tumor extension in glioma for external beam radiotherapy planning. BMC Med. 2021;19(1):269. https://doi.org/10.1186/s12916-021-02143-w\u003c/li\u003e\n\u003cli\u003eKarschnia P, Smits M, Reifenberger G, et al. A framework for standardised tissue sampling and processing during resection of diffuse intracranial glioma: joint recommendations from four RANO groups. Lancet Oncol. 2023;24(11):438-50. https://doi.org/10.1016/S1470-2045(23)00453-9\u003c/li\u003e\n\u003cli\u003eLee E, Lee EA, Kong E, et al. An agonistic anti-Tie2 antibody suppresses the normal-to-tumor vascular transition in the glioblastoma invasion zone. Exp Mol Med. 2023;55(2):470-84. https://doi.org/10.1038/s12276-023-00939-9\u003c/li\u003e\n\u003cli\u003eSegura-Collar B, Garranzo-Asensio M, Herranz B, et al. Tumor-Derived Pericytes Driven by EGFR Mutations Govern the Vascular and Immune Microenvironment of Gliomas. Cancer Res. 2021;81(8):2142-56. https://doi.org/10.1158/0008-5472\u003c/li\u003e\n\u003cli\u003eChai R, Fang S, Pang B, et al. Molecular pathology and clinical implications of diffuse glioma. Chin Med J (Engl). 2022;135(24):2914-25. https://doi.org/10.1097/CM9.0000000000002446\u003c/li\u003e\n\u003cli\u003eBliesener Y, Lebel RM, Acharya J, Frayne R, Nayak KS. Pseudo Test-Retest Evaluation of Millimeter-Resolution Whole-Brain Dynamic Contrast-enhanced MRI in Patients with High-Grade Glioma. Radiology. 2021;300(2):410-20. https://doi.org/10.1148/radiol.2021203628\u003c/li\u003e\n\u003cli\u003eGerritsen JKW, Zwarthoed RH, Kilgallon JL, et al. Impact of maximal extent of resection on postoperative deficits, patient functioning, and survival within clinically important glioblastoma subgroups. Neuro Oncol. 2023;25(5):958-72. https://doi.org/10.1093/neuonc/noac255\u003c/li\u003e\n\u003cli\u003eDella Pepa GM, Ius T, La Rocca G, et al. 5-Aminolevulinic Acid and Contrast-Enhanced Ultrasound: The Combination of the Two Techniques to Optimize the Extent of Resection in Glioblastoma Surgery. Neurosurgery. 2020;86(6):529-40. https://doi.org/10.1093/neuros/nyaa037\u003c/li\u003e\n\u003cli\u003eWang Z, Zhang M, Chi S, Zhu M, Wang C, Liu Z. Brain Tumor Cell Membrane-Coated Lanthanide-Doped Nanoparticles for NIR-IIb Luminescence Imaging and Surgical Navigation of Glioma. Adv Healthc Mater. 2022;11(16):e2200521. https://doi.org/10.1002/adhm.202200521\u003c/li\u003e\n\u003cli\u003eCroci D, Santalla M\u0026eacute;ndez R, Temme S, et al. Multispectral fluorine-19 MRI enables longitudinal and noninvasive monitoring of tumor-associated macrophages. Sci Transl Med. 2022;14(667):eabo2952. https://doi.org/10.1126/scitranslmed.abo2952\u003c/li\u003e\n\u003cli\u003eDo MH, Shi W, Ji L, et al. Reprogramming tumor-associated macrophages to outcompete endovascular endothelial progenitor cells and suppress tumor neoangiogenesis. Immunity. 2023;56(11):2555-69.e5. https://doi.org/10.1016/j.immuni.2023.10.010\u003c/li\u003e\n\u003cli\u003eFolkman J. Tumor angiogenesis: therapeutic implications. N Engl J Med. 1971;285(21):1182-6. https://doi.org/10.1056/NEJM197111182852108\u003c/li\u003e\n\u003cli\u003eMathur R, Wang Q, Schupp PG, et al. Glioblastoma evolution and heterogeneity from a 3D whole-tumor perspective. Cell. 2024;187(2):446-63.e16. https://doi.org/10.1016/j.cell.2023.12.013\u003c/li\u003e\n\u003cli\u003eWang X, Xu T, Niu C. Vascularization ability of glioma stem cells in different three-dimensional microenvironments. Regen Biomater. 2023;11:rbad094. https://doi.org/10.1093/rb/rbad094\u003c/li\u003e\n\u003cli\u003eHingot V, Brodin C, Lebrun F, et al. Early Ultrafast Ultrasound Imaging of Cerebral Perfusion correlates with Ischemic Stroke outcomes and responses to treatment in Mice. Theranostics. 2020;10(17):7480-91. https://doi.org/10.7150/thno.44233\u003c/li\u003e\n\u003cli\u003eErrico C, Pierre J, Pezet S, et al. Ultrafast ultrasound localization microscopy for deep super-resolution vascular imaging. Nature. 2015;527(7579):499-502. https://doi.org/10.1038/nature16066\u003c/li\u003e\n\u003cli\u003eDemen\u0026eacute; C, Robin J, Dizeux A, et al. Transcranial ultrafast ultrasound localization microscopy of brain vasculature in patients. Nat Biomed Eng. 2021;5(3):219-28. https://doi.org/10.1038/s41551-021-00697-x\u003c/li\u003e\n\u003cli\u003eWang R, Dai W, Gong J, et al. Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients. J Hematol Oncol. 2022;15(1):11. https://doi.org/10.1186/s13045-022-01225-3\u003c/li\u003e\n\u003cli\u003eNiu X, Yang Y, Zhou X, et al. A prognostic nomogram for patients with newly diagnosed adult thalamic glioma in a surgical cohort. Neuro Oncol. 2021;23(2):337-8. https://doi.org/10.1093/neuonc/noaa268\u003c/li\u003e\n\u003cli\u003eLinazi G, Maimaiti A, Abulaiti Z, et al. Prognostic value of anoikis-related genes revealed using multi-omics analysis and machine learning based on lower-grade glioma features and tumor immune microenvironment. Heliyon. 2024;10(17):e36989. https://doi.org/10.1016/j.heliyon.2024.e36989\u003c/li\u003e\n\u003cli\u003eLowerison MR, Vaithiyalingam Chandra Sekaran N, Dong Z, et al. Super-Resolution Ultrasound Reveals Cerebrovascular Impairment in a Mouse Model of Alzheimer\u0026rsquo;s Disease. J Neurosci. 2024;44(9):e1251232024. https://doi.org/10.1523/JNEUROSCI.1251-23.2024\u003c/li\u003e\n\u003cli\u003eNy\u0026uacute;l-T\u0026oacute;th \u0026Aacute;, Negri S, Sanford M, et al. Novel intravital approaches to quantify deep vascular structure and perfusion in the aging mouse brain using ultrasound localization microscopy (ULM). J Cereb Blood Flow Metab. 2024;44(11):1378-96. https://doi.org/10.1177/0271678X241260526\u003c/li\u003e\n\u003cli\u003eSun J, Cheng W, Guo S, et al. A ratiometric SERS strategy for the prediction of cancer cell proportion and guidance of glioma surgical resection. Biosens Bioelectron. 2024;261:116475. https://doi.org/10.1016/j.bios.2024.116475\u003c/li\u003e\n\u003cli\u003eHe A, Xu L, Yang X, Gu Z, Cai Y, Zhou H. Risk factors for surgical compliance and impact on the survival of patients with glioma: a population-based propensity score-matched study. J Cancer Res Clin Oncol. 2023;149(16):14797-815. https://doi.org/10.1007/s00432-023-05261-5\u003c/li\u003e\n\u003cli\u003eDel Bene M, Perin A, Casali C, et al. Advanced Ultrasound Imaging in Glioma Surgery: Beyond Gray-Scale B-mode. Front Oncol. 2018;8:576. https://doi.org/10.3389/fonc.2018.00576\u003c/li\u003e\n\u003cli\u003eMei X, Chen YS, Zhang QP, et al. Association between glioblastoma cell-derived vessels and poor prognosis of the patients. Cancer Commun (Lond). 2020;40(5):211-21. https://doi.org/10.1002/cac2.12026\u003c/li\u003e\n\u003cli\u003eYin J, Dong F, An J, et al. Pattern recognition of microcirculation with super-resolution ultrasound imaging provides markers for early tumor response to anti-angiogenic therapy. Theranostics. 2024;14(3):1312-24. https://doi.org/10.7150/thno.89306\u003c/li\u003e\n\u003cli\u003eJiang H, Zuo M, Li W, Zhuo S, Wu P, An C. Multimodal imaging-based prediction of recurrence for unresectable HCC after downstage and resection-cohort study. Int J Surg. 2024;110(9):5672-84. https://doi.org/10.1097/JS9.0000000000001752\u003c/li\u003e\n\u003cli\u003eZhu C, Zou C, Guan G, et al. Development and validation of an interferon signature predicting prognosis and treatment response for glioblastoma. Oncoimmunology. 2019;8(9):e1621677. https://doi.org/10.1080/2162402X.2019.1621677\u003c/li\u003e\n\u003cli\u003eEllingson BM, Hagiwara A, Morris CJ, et al. Depth of Radiographic Response and Time to Tumor Regrowth Predicts Overall Survival Following Anti-VEGF Therapy in Recurrent Glioblastoma. Clin Cancer Res. 2023;29(20):4186-95. https://doi.org/10.1158/1078-0432.CCR-23-1235\u003c/li\u003e\n\u003cli\u003eSpinelli C, Adnani L, Meehan B, et al. Mesenchymal glioma stem cells trigger vasectasia-distinct neovascularization process stimulated by extracellular vesicles carrying EGFR. Nat Commun. 2024;15(1):2865. https://doi.org/10.1038/s41467-024-46597-x\u003c/li\u003e\n\u003cli\u003eZhu X, Huang Q, DiSpirito A, et al. Real-time whole-brain imaging of hemodynamics and oxygenation at micro-vessel resolution with ultrafast wide-field photoacoustic microscopy. Light Sci Appl. 2022;11(1):138. https://doi.org/10.1038/s41377-022-00836-2\u003c/li\u003e\n\u003cli\u003eRosberg R, Smolag KI, Sj\u0026ouml;lund J, et al. Hypoxia-induced complement component 3 promotes aggressive tumor growth in the glioblastoma microenvironment. JCI Insight. 2024;9(19):e179854. https://doi.org/10.1172/jci.insight.179854\u003c/li\u003e\n\u003cli\u003eStadlbauer A, Zimmermann M, Kitzw\u0026ouml;gerer M, et al. MR Imaging-derived Oxygen Metabolism and Neovascularization Characterization for Grading and IDH Gene Mutation Detection of Gliomas. Radiology. 2017;283(3):799-809. https://doi.org/10.1148/radiol.2016161422\u003c/li\u003e\n\u003cli\u003eMcCall JR, DeRuiter R, Ross M, et al. Longitudinal 3-D Visualization of Microvascular Disruption and Perfusion Changes in Mice During the Evolution of Glioblastoma Using Super-Resolution Ultrasound. IEEE Trans Ultrason Ferroelectr Freq Control. 2023;70(11):1401-16. https://doi.org/10.1109/TUFFC.2023.3320034\u003c/li\u003e\n\u003cli\u003eFrost HR, Amos CI. Gene set selection via LASSO penalized regression (SLPR). Nucleic Acids Res. 2017;45(12):e114. https://doi.org/10.1093/nar/gkx291\u003c/li\u003e\n\u003cli\u003eHormuth DA, Wong AH, Yankeelov TE. Mathematical modeling of tumor vascularization: understanding the interplay of angiogenesis and vascular remodeling. Front Physiol. 2021;12:725-36.\u003c/li\u003e\n\u003cli\u003eLiu G, Yang J, Wang J, et al. Extended axial imaging range: Widefield swept-source optical coherence tomography angiography. J Biophotonics. 2017;10(11):1464-72. https://doi.org/10.1002/jbio.201600325\u003c/li\u003e\n\u003cli\u003eSchupper AJ, Hadjipanayis CG. Novel approaches to targeting gliomas at the leading/cutting edge. J Neurosurg. 2023;139(3):760-8. https://doi.org/10.3171/2023.1.JNS221798\u003c/li\u003e\n\u003cli\u003eHuang R, Wu H, Lu X, Sun X. Clinical characteristics and prognostic factors of solitary and multiple adult gliomas: a retrospective study based on propensity score matching. Eur Rev Med Pharmacol Sci. 2023;27(21):10481-98. https://doi.org/10.26355/eurrev_202311_34325\u003c/li\u003e\n\u003cli\u003eStaub-Bartelt F, Suresh Babu MP, Szel\u0026eacute;nyi A, Rapp M, Sabel M. Establishment of Different Intraoperative Monitoring and Mapping Techniques and Their Impact on Survival, Extent of Resection, and Clinical Outcome in Patients with High-Grade Gliomas-A Series of 631 Patients in 14 Years. Cancers (Basel). 2024;16(5):926. https://doi.org/10.3390/cancers16050926\u003c/li\u003e\n\u003cli\u003eYan J, Huang B, Tonko J, et al. Transthoracic ultrasound localization microscopy of myocardial vasculature in patients. Nat Biomed Eng. 2024;8(6):689-700. https://doi.org/10.1038/s41551-024-01206-6\u003c/li\u003e\n\u003cli\u003eBodard S, Denis L, Chabouh G, et al. Visualization of Renal Glomeruli in Human Native Kidneys With Sensing Ultrasound Localization Microscopy. Invest Radiol. 2024;59(8):561-8. https://doi.org/10.1097/RLI.0000000000001061\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Ultrasound localization microscopy, Glioblastoma, Visualization, Invasion, Microvascular","lastPublishedDoi":"10.21203/rs.3.rs-6464545/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6464545/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e \u003cb\u003eBackground\u003c/b\u003e This study aims to develop and assess a nomogram based on multiparametric ultrasound localization microscopy to evaluate the risk of peritumoral invasion.\u003c/p\u003e \u003cp\u003e \u003cb\u003eMethods\u003c/b\u003e Thirty-six in situ rat glioblastoma models were created. After craniotomy, ultrasound localization microscopy was used to quantify microvascular morphology and hemodynamics, which were combined with multimodal magnetic resonance imaging to manually delineate the invasive and normal brain regions. The least absolute shrinkage and selection operator regression algorithm was applied to select ultrasound localization microscopy parameters, followed by multivariable logistic regression to identify significant variables. A nomogram to predict peritumoral invasion risk was constructed using R software, and its diagnostic performance was evaluated.\u003c/p\u003e \u003cp\u003e \u003cb\u003eResults\u003c/b\u003e Vascularity (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), orientation variance (p\u0026thinsp;=\u0026thinsp;0.013), and diameter (p\u0026thinsp;=\u0026thinsp;0.002) were identified as independent predictors of peritumoral invasion. The prediction model demonstrated strong discriminatory power, with an area under the curve of 0.964 (0.933\u0026ndash;0.994) for the training set and 0.995 (0.984\u0026ndash;1.000) for the validation set. The goodness-of-fit Hosmer-Lemeshow test statistics were 5.135 (p\u0026thinsp;=\u0026thinsp;0.702) and 3.163 (p\u0026thinsp;=\u0026thinsp;0.237), indicating that the predicted invasion risk closely matched the actual risk. Decision curve analysis revealed that when the invasion incidence ranged from 1\u0026ndash;99% in the training set and from 5\u0026ndash;94% in the validation set, the nomogram provided clinical benefit, demonstrating good generalizability.\u003c/p\u003e \u003cp\u003e \u003cb\u003eConclusions\u003c/b\u003e We developed and validated a nomogram to predict peritumoral invasion in glioblastoma, enabling clinicians to perform preoperative risk assessments and implement personalized surgical strategies to improve resection rates.\u003c/p\u003e","manuscriptTitle":"Risk of Peritumoral Invasion in Rat Glioblastoma: Nomogram-based Ultrasound Localization Microscopy","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-19 10:40:18","doi":"10.21203/rs.3.rs-6464545/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewersInvited","content":"","date":"2025-05-14T18:10:52+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-04-23T17:45:12+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-04-23T04:12:20+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-04-23T04:10:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Imaging","date":"2025-04-16T14:24:52+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-imaging","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bmim","sideBox":"Learn more about [BMC Medical Imaging](http://bmcmedimaging.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bmim/default.aspx","title":"BMC Medical Imaging","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5595d0ff-6fd8-48ff-94ac-0c223e06322e","owner":[],"postedDate":"May 19th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-05-19T10:40:19+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-19 10:40:18","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6464545","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6464545","identity":"rs-6464545","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}