Search

To develop and validate a deep learning model for predicting overall survival from whole-brain MRI without tumor segmentation in patients with diffuse gliomas.In this multicenter retrospective study, two deep learning models were built for survival prediction from MRI, including a DeepRisk model built from whole-brain MRI, and an original ResNet model built from expert-segmented tumor images. Both models were developed using a training dataset (n = 935) and an internal tuning dataset (n = 156) and tested on two external test datasets (n = 194 and 150) and a TCIA dataset (n = 121). C-index, integrated Brier score (IBS), prediction error curves, and calibration curves were used to assess the model performance.In total, 1556 patients were enrolled (age, 49.0 ± 13.1 years; 830 male). The DeepRisk score was an independent predictor and can stratify patients in each test dataset into three risk subgroups. The IBS and C-index for DeepRisk were 0.14 and 0.83 in external test dataset 1, 0.15 and 0.80 in external dataset 2, and 0.16 and 0.77 in TCIA dataset, respectively, which were comparable with those for original ResNet. The AUCs at 6, 12, 24, 26, and 48 months for DeepRisk ranged between 0.77 and 0.94. Combining DeepRisk score with clinicomolecular factors resulted in a nomogram with a better calibration and classification accuracy (net reclassification improvement 0.69, p0.001) than the clinical nomogram.DeepRisk that obviated the need of tumor segmentation can predict glioma survival from whole-brain MRI and offers incremental prognostic value.• DeepRisk can predict overall survival directly from whole-brain MRI without tumor segmentation. • DeepRisk achieves comparable accuracy in survival prediction with deep learning model built using expert-segmented tumor images. • DeepRisk has independent and incremental prognostic value over existing clinical parameters and IDH mutation status.

ObjectivesPhosphatase and tensin homolog (PTEN) mutation is an indicator of poor prognosis of low-grade and high-grade glioma. This study built a reliable model from multi-parametric magnetic resonance imaging (MRI) for predicting the PTEN mutation status in patients with glioma.MethodsIn this study, a total of 244 patients with glioma were retrospectively collected from our center (n = 77) and The Cancer Imaging Archive (n = 167). All patients were randomly divided into a training set (n = 170) and a validation set (n = 74). Three models were built from preoperative MRI for predicting PTEN status, including a radiomics model, a convolutional neural network (CNN) model, and an integrated model based on both radiomics and CNN features. The performance of each model was evaluated by accuracy and the area under the receiver operating characteristic curve (AUC).ResultsThe CNN model achieved an AUC of 0.84 and an accuracy of 0.81, which performed better than did the radiomics model, with an AUC of 0.83 and an accuracy of 0.66. Combining radiomics with CNN will further benefit the predictive performance (accuracy = 0.86, AUC = 0.91).ConclusionsThe combination of both the CNN and radiomics features achieved significantly higher performance in predicting the mutation status of PTEN in patients with glioma than did the radiomics or the CNN model alone.

Background and Aim: Guidelines recommend adjuvant treatment for patients with high-grade gliomas and low-grade gliomas with high risk of progression. In clinical practice, however, treatments may not conform to these suggested guidelines. In this study, we reviewed the treatments and outcomes in patients with gliomas at Sun Yat-Sen University Cancer Center (SYSUCC), China. Materials and Methods: Medical records and radiologic images of 1215 glioma patients who underwent surgery at the center from 2000 to 2017 were retrospectively reviewed, and their clinicopathological characteristics, treatment method, and overall survival (OS) were analyzed. The study was approved by the Ethics Committee of SYSUCC on February 20, 2019 (approval No. GZR2019-219). Results: A total of 1001 patients diagnosed with glioma (initially World Health Organization 2007 criteria, then 2016 criteria) were enrolled, including 90 patients with Grade I, 307 Grade II, 239 Grade III, and 365 Grade IV gliomas. A total of 331 of 604 patients with high-grade glioma (54.8%) and 159 of 397 with low-grade glioma (40.1%) received postsurgical radiotherapy, and 285 patients with high-grade tumors (47.1%) and 80 with low-grade tumors (20.2%) received adjuvant chemotherapy. The median OS was 17.5 months for Grade IV and 43.1 months for Grade III gliomas. The median OS of patients with low-grade glioma was not reached. The 5-year survival rates of patients with Grades I, II, III, and IV gliomas were 94.7%, 73.7%, 45.7%, and 18.6%, respectively. Multivariate analysis identified onset age, preoperative seizure, tumor location, pathological subtype, resection extent, and postsurgical treatment as independent predictors of OS in patients with high-grade gliomas. Patients with high-grade glioma who received postsurgical treatment had better survival than those without adjuvant therapy (Grade III: 52.6 vs. 20.3 months, P = 0.012; Grade IV: 22.6 vs. 12.1 months, P < 0.001). Among patients with diffuse low-grade gliomas, age, performance status, preoperative seizure, Ki-67 index, tumor subtype, and resection extent were associated with clinical outcomes. Conclusion: Glioma patients are not always treated according to guidelines. Although standard care may lead to favorable prognoses, individualized treatments may be more acceptable and result in better outcomes and should thus be considered in routine clinical practice.

Purpose Accurately predicting outcome (i.e., overall survival (OS) time) for higher grade glioma (HGG) has great clinical value and would provide optimized guidelines for treatment planning. Radiomics focuses on revealing underlying pathophysiological information in biomedical images for disease analysis and demonstrates promising prognostic clinical performance. In this paper, we propose a novel sparse representation-based radiomics framework to predict if HGG patients would have long or short OS time. Methods First, taking advantages of the scale invariant feature transform (SIFT) feature in image characterizing, we developed a sparse representation-based method to convert a local SIFT descriptor into a global tumor feature. Next, because preserving sample structure is beneficial for feature selection, we proposed a locality preserving projection and sparse representation-combined feature selection method to select more discriminative features for tumor classification. Finally, we employed a multifeature collaborative sparse representation classification to combine the information of multimodal images to classify OS time. Results Three experiments were performed on the two datasets provided by different institutions. Specifically, the proposed model was trained and independently tested on dataset 1 (135 subjects), on dataset 2 (86 subjects), and on the combination of dataset 1 and dataset 2, respectively. Experimental results demonstrated that the proposed method achieved encouraging prediction performance, exhibiting a testing accuracy of 93.33% on dataset 1 (one modality), 92.31% on dataset 2 (two modalities), and 87.93% on the combined dataset (one modality). Conclusions The sparse representation theory provides reasonable solutions to feature extraction, feature selection, and classification for radiomics. This study provides a promising tool to enhance the prediction performance of HGG patient's outcome.

Purpose Amide proton transfer (APT) imaging has been increasingly applied in tumor characterization that complements diffusion and dynamic contrast-enhanced MRI. However, quantification of in vivo APT effect is challenging because of concomitant semisolid magnetization transfer (MT) and nuclear overhauser enhancement effects. A direct saturation corrected (DISC) chemical exchange saturation transfer (CEST) analysis has been recently proposed that simplifies the determination of in vivo CEST effects. Our present study aimed to extend the DISC analysis to pulsed radiofrequency CEST MRI and evaluate it at 3T. Methods Nine adult male Sprague-Dawley rats implanted C6 gliomas underwent multiparametric MRI of T1 , T2 , CEST, and T1 -weighted gadolinium-enhanced imaging 1 day before and 3 days after chemoradiotherapy. The routine MT asymmetry, 3-point method, and the extended DISC analysis were compared in tumor characterization with histology as a reference. Regional variations were assessed by 1-way analysis of variance. Results T1 , T2 , and MT asymmetry and the DISC CEST effects showed significant alterations in tumor/necrosis with respect to the contralateral reference (P 0.05), whereas the 3-point method detected no regional alteration at both time points (P > 0.05). Conclusion Our study translated DISC CEST MRI to 3T, evaluated it in glioma rat models, and confirmed its advantages in resolving tumor heterogeneity over the routine asymmetry and 3-point analyses.

Background: To develop and validate a deep learning signature (DLS) from diffusion tensor imaging (DTI) for predicting overall survival in patients with infiltrative gliomas, and to investigate the biological pathways underlying the developed DLS. Methods: The DLS was developed based on a deep learning cohort (n = 688). The key pathways underlying the DLS were identified on a radiogenomics cohort with paired DTI and RNA-seq data (n=78), where the prognostic value of the pathway genes was validated in public databases (TCGA, n = 663; CGGA, n = 657). Findings: The DLS was associated with survival (log-rank P < 0.001) and was an independent predictor (P < 0.001). Incorporating the DLS into existing risk system resulted in a deep learning nomogram predicting survival better than either the DLS or the clinicomolecular nomogram alone, with a better calibration and classification accuracy (net reclassification improvement 0.646, P < 0.001). Five kinds of pathways (synaptic transmission, calcium signaling, glutamate secretion, axon guidance, and glioma pathways) were significantly correlated with the DLS. Average expression value of pathway genes showed prognostic significance in our radiogenomics cohort and TCGA/CGGA cohorts (log-rank P < 0.05). Interpretation: DTI-derived DLS can improve glioma stratification by identifying risk groups with dysregulated biological pathways that contributed to survival outcomes. Therapies inhibiting neuron-to-brain tumor synaptic communication may be more effective in high-risk glioma defined by DTI-derived DLS. Funding: A full list of funding bodies that contributed to this study can be found in the Acknowledgements section.

Twist is a transcription factor involved in the process of epithelial to mesenchymal transition (EMT) of carcinoma cells, and the promotion of invasion of gliomas through the mesenchymal adjusting process. However, its clinical significance in human glioma has not yet to be understood. To delineate the clinical-pathological significance and prognostic value of Twist, the expression of Twist was evaluated by Immunohistochemistry for 187 glioma samples. We found that Twist demonstrated frequent nuclear expression in the glioma samples and its expression levels were associated with tumor grade (P0.001). Furthermore, high Twist expression was correlated with a poor outcome in patients with glioma (P=0.001), particularly with high grade glioma (P=0.026). Interestingly, Twist expression showed positive correlation with microvascular density (MVD) (r=0.145, P=0.048) as well as vasculogenic mimicry (VM) (r=0.273, P0.001) in the tumors. These results suggest that Twist could be a predictor for poor prognosis in glioma patients. Additionally, Twist expression was associated with two major microcirculation patterns: endothelial-dependent vessels and VM in glioma, indicating that Twist could be a potential molecular target for anti-glioma therapy.

Background Differentiating glioma recurrence from treatment-related changes can be challenging on conventional imaging. We evaluated the efficacy of quantitative parameters measured by dual-energy spectral computed tomographic (CT) for this differentiation. Methods Twenty-eight patients were examined by dual-energy spectral CT. The effective and normalized atomic number (Zeff and Zeff-N, respectively); spectral Hounsfield unit curve (λHU) slope; and iodine and normalized iodine concentration (IC and ICN, respectively) in the post-treatment enhanced areas were calculated. Pathological results or clinicoradiologic follow-up of ≥2 months were used for final diagnosis. Nonparametric and t-tests were used to compare quantitative parameters between glioma recurrence and treatment-related changes. Sensitivity, specificity, positive and negative predictive values (PPV and NPV, respectively), and accuracy were calculated using receiver operating characteristic (ROC) curves. Predictive probabilities were used to generate ROC curves to determine the diagnostic value. Results Examination of pre-contrast λHU, Zeff, Zeff-N, IC, ICN, and venous phase ICN showed no significant differences in quantitative parameters (P > 0.05). Venous phase λHU, Zeff, Zeff-N, and IC in glioma recurrence were higher than in treatment-related changes (P 3, achieving 66.7, 91.7, 83.3, and 91.7% sensitivity; 100.0, 77.8, 88.9, and 77.8% specificity; 100.0, 73.3, 83.3, and 73.3% PPV; 81.8, 93.3, 88.9, and 93.3% NPV; and 86.7, 83.3, 86.7, and 83.3% accuracy, respectively. The respective areas under the curve (AUCs) were 0.912, 0.912, 0.931, and 0.910 in glioma recurrence and treatment-related changes. Conclusions Glioma recurrence could be potentially differentiated from treatment-related changes based on quantitative values measured by dual-energy spectral CT imaging.

Twist is a transcription factor involved in the process of epithelial to mesenchymal transition (EMT) of carcinoma cells, and the promotion of invasion of gliomas through the mesenchymal adjusting process. However, its clinical significance in human glioma has not yet to be understood. To delineate the clinical-pathological significance and prognostic value of Twist, the expression of Twist was evaluated by Immunohistochemistry for 187 glioma samples. We found that Twist demonstrated frequent nuclear expression in the glioma samples and its expression levels were associated with tumor grade (P

BACKGROUND: The present study explored the predictive value of systemic inflammatory indexes in diagnosing grade III gliomas of oligodendroglial origin. METHODS: A retrospective study of 154 patients with grade III gliomas was conducted. Systemic inflammatory indexes, including neutrophil-to-lymphocyte ratio (NLR), albumin-to-gamma-glutamyl transferase ratio (AGR), platelet-to-lymphocyte ratio, lymphocyte-to-monocyte ratio, prognostic nutritional index, and fibrinogen-to-albumin ratio, were reviewed. The resulting predictive model was externally validated using a demographic-matched cohort of 49 grade III glioma patients. RESULTS: In the training set, gliomas of oligodendroglial origin tended to have a lower NLR (P=0.018) and a higher AGR (P=0.036) than those with tumors of astrocytic origin. Moreover, both NLR and AGR had predictive value for oligodendroglial tumors, when compared with astrocytic tumors. The best diagnostic value was obtained using NLR + AGR (AUC =64.9%, 95% CI: 55.5–74.3%, P=0.005). In the validation set, NLR + AGR satisfactorily predicted the presence of oligodendroglial tumors (AUC =66.5%, 95% CI: 50.6–82.4%, P

e14019 Background: High-grade glioma (HGG) is the most common malignant brain tumor and lacks effective treatment regimen. Anlotinib is a multikinase inhibitor blocking angiogenesis and tumor cell proliferation simultaneously. This study was performed to evaluate the efficacy and safety of anlotinib alone or in combination with temozolomide (TMZ) in the treatment of recurrent HGG. Methods: This is a single-center, retrospective study. Eligible patients (pts) were diagnosed with pathologically confirmed high grades (WHO III/IV) glioma and had recurrent or progressive disease on or after prior treatment. Other key eligibility criteria included Karnofsky Performance Status (KPS) ≥ 40, aged 16 ̃75 years and having at least one measurable lesion (RANO criteria). Pts were administrated with anlotinib once daily for 14 days every 3 weeks till disease progression, intolerable toxicities or death. The initial dose was 12mg for younger pts ( < 40 years old) with KPS ≥ 60 and 10 mg for others. Combination treatment was allowed if previous TMZ was effective and tolerable. TMZ was administered on dose-dense schedule (150mg/m2, QD, d1-d7 and d15-d21 every 28 days) or metronomic schedule (25-50mg/m2 QD). The primary endpoint was progression-free survival at 6 months (PFS6m) accessed according to RANO criteria. The second endpoints included overall survival (OS), objective response rate (ORR) and disease control rate (DCR). Results: Between August 2019 and June 2020, 23 pts with HGG (15 grade IV; 8 grade III; 12 males, 11 females) were enrolled. The median age and median KPS was 42 years and 60. 16 pts have multifocal or disseminated disease. 18 pts received ≥2 lines previous treatment. At the data cutoff date on September 2020, the median duration of treatment was 9 weeks (range: 3-33). The PFS6m was 39.1% and the median PFS was 4.2 months (95% CI: 2.8, 5.6). The median OS was not reached (95% CI: NE, NE) and the OS at 12 months (OS12m) was 54.8%. 8 pts observed tumor response and 9 pts had stable disease. The ORR and DCR were 34.8% and 73.9% respectively. The results of survival analysis for subgroups were summarized in table below. Grade 1 or 2 treatment-related adverse events (TRAEs) occurred in 65.2% pts. No ≥ grade 3 TRAE was found. All hematological TRAEs occurred in patients received combination regimen. No TRAE-induced treatment termination occurred. The lower incidence of TRAE may partly attributed to that most pts (18/23) received lower initial dose (10mg) of anlotinib and the relatively shorter treatment duration. Conclusions: This study showed treatment with anlotinib alone or in combination with TMZ had promising efficacy and favorable tolerability in patients with recurrent HGG.[Table: see text]

Glioblastoma is a highly invasive and vascularized primary CNS tumor. Although anti-angiogenesis therapy was effective in some tumors but not in GBM. Vasculogenic mimicry (VM), a vessel-like structure formed by highly invasive tumor cells, has been considered responsible for the failure of GBM patients for anti-angiogenesis therapy. Tenascin-c (TNC), an extracellular protein involves in tumor angiogenesis, is overexpressed in GBM. However, how TNC contributes to VM formation remains unclear. We first knocked down the expression of TNC in two glioma cell lines (U251 and A172), which had TNC overexpression. Then the formations of VM were observed by three-dimensional culture. Transwell and wound-healing assay were applied to investigate the role of TNC in cell invasion and migration abilities. The expression of MMP2, MMP9 and the phosphorylation status of AKT were determined by western-blot, RT-qPCR and gelatin zymography. Furthermore, AKT inhibitor, MK-2206 was used to block the AKT/MMP2/MMP9 signaling pathway to investigate its role in TNC regulation at different concentration (0µm, 5µm, 10µm and 20µm) for 24 hours. The knockdown efficiency was detected by western blot and RT-qPCR in U251 and A172 after siRNA transfection for 48 hours. Three-dimensional culture showed that VM-like structure formation decreased in knockdown groups compared to controls dramatically (19.67 ± 6.65 and 46.34 ± 9.10 vs. 125.67 ± 8.34 p

Magnetic resonance amide proton transfer (APT) imaging has proved its potential for detecting tumors and evaluating treatment response by depicting chemical exchange saturation transfer effect between the endogenous protein/peptide amide proton and bulk water. However, conventional asymmetry analysis for APT effect measurement is susceptible to concomitant contributions, particular semisolid magnetic transfer (MT) and nuclear overhauser effect. In this study, dominant saturation transfer (ST) effects, including direct water saturation and MT, were estimated from a sum of two Lorentzian functions. APT effect was then quantified by subtracting the Z-spectrum with the two ST effects. Feasibility of the method in the assessment of glioma therapeutic response was investigated in a rodent model at 3 Tesla.

BACKGROUND AND OBJECTIVE: We have found, in previous study, there were four types of microcirculation pattern in human glioblastoma: endothelium dependent vessels, tumor cell dependent vessels, extracellular matrix (ECM) dependent vessels, and mosaic vessels. This study is, by utilizing live cell imaging technique, to visualize the differentiation of glioma stem cells (GSCs) into endothelial cells. METHODS: Live cell imaging was used to record the transforming process of GSCs. The immunofluorescence staining was also performed to evaluate the differentiation phenomenon during GSCs three dimensional culture at time points of 0 hour, 2 hours and 6 hours. RESULTS: Tubular structures were observed in GSC-1 three dimensional culture, which was derived from a GBM patient. Some GSC-1 cells were found endothelial maker CD34 expression, and lost stem phenotype following three dimensional culture. The number of CD34+ cells were increased, and reach to highest up to 6-hour three dimensional culture (P < 0.05). Two main phenomenon were recorded by live cell imaging: (1) the number of CD34+ cells was gradually increasing following GSCs three dimensional culture; (2) The process of a single GSC-1 cell from CD34- to CD34+ . CONCLUSIONS: The result of live cell imaging has confirmed the results from conventional immunofluorescence staining that glioma stem cell may differentiate into endothelial cells, suggesting that GSCs could contribute to vasculargenesis in GBM.

Background: Vasculogenic mimicry (VM) describes the functional plasticity of aggressive tumor cells to form newly recognized microcirculation, lined by tumor cells rather than endothelial cells, in solid tumors. The presence of VM in glioma is significantly associated with high tumor grade and poor prognosis. However, whether VM is a regular feature or only a temporary phenomenon during glioma growth is unknown. This study was designed to observe VM during the growth of subcutaneous and orthotopic xenograft glioma in Balb/c nude mice. Methods: The human glioblastoma cell line (U87) was used as xenografts in Balb/c nude mice models. The xenografts were obtained at different stages of tumor growth, and evaluated for VM and endothelium-dependent vessels by dual staining for endothelial marker CD34 and periodic acid-Schiff (PAS). Results: It was found that the PAS-positive patterns which were identified as VM were persistent during tumor growth of both subcutaneous and orthotropic xenografts. Further analysis showed that the microvessel density (MVD) of endothelium-dependent vessels was positively correlated with the tumor size of subcutaneous xenograft (r = 0.406, P = 0.009), while no significant correlation was found between VM density (VMD) and the tumor size (r = 0.107, P = 0.512). Furthermore, VMD was negatively correlated with MVD (r = −0.404, P = 0.010). Conclusion: The study results revealed that both VM and endothelium-dependent vessels coexist persistently during glioblastoma xenograft growth, indicating that VM may complement microcirculation in gliomas.

BACKGROUND: Vasculogenic mimicry(VM) has been reported in many malignant human tumors, including gliomas. In the present study, we have found four types of microcirculation vessels in human glioblastomas. Method and material: Thirty-seven human glioblastoma samples were examined for VM by CD34 and PAS dual staining,as well as immunofluorescence dual staining for CD34 and GFAP. The GFAP positive (GFAP+) cells were considered as glioma cells. Even though CD34 positive (CD34+) was endothelial cells, but both CD34 + / GFAP+ should be also considered as glioma cells. Result: CD34 and PAS dual staining revealed 4 types of microcirculation vessels in human glioblastoma samples: endothelium dependent vessels, tumor cell dependent vessels, extracellular matrix (ECM) dependent vessels, and mosaic vessels. While immunofluorescence study confirmed VM should be vessels lined with GFAP+ cells (could be CD34- or co-express CD34). Twelve samples were found VM by CD34 and PAS dual staining (32%), while 20 samples were revealed VM by immunofluorescence dual staining (57%). CONCLUSION: Our results suggest that glioblastoma cells have potential to express endothelial cell marker CD34. And immunofluorescence dual staining is a more sensitive as well accurate technique revealing VM vessels in glioblastoma.