Dual-task deep learning model for prediction of medulloblastoma molecular subgroups with preoperative brain MRI.

e20655 Background: ADAURA demonstrated an overall survival (OS) benefit with adjuvant osimertinib in Epidermal Growth Factor Receptor mutant (EGFR-mut) non-small cell lung cancer (NSCLC). However, only 50% of patients underwent a preoperative brain MRI, and PET-CT rates were not reported. Concerns were raised that preoperative imaging rates were not consistent with real world practices, leading to patients being effectively “understaged.” There is limited data on real world rates of preoperative PET-CT and brain MRIs in patients with EGFR-mut NSCLC. Moreover, there is little reported data on the staging impact of preoperative PET-CT and brain MRI. We hypothesized that PET CTs and brain MRIs are performed in the majority of patients with EGFR-mut NSCLC who are evaluated for surgery and lead to detection of metastases missed by CT imaging alone. Methods: We retrospectively analyzed patients with EGFR-mut NSCLC who were evaluated for surgery at three academic New York hospitals between 2016-2021. EGFR-mut was defined as an EGFR exon 19 deletion or EGFR L858R mutation. Results: Between 2016 and 2021,109 patients with EGFR-mut NSCLC underwent preoperative evaluation. The median age was 70 (range: 43-89), and they were predominately female (74%). Forty-three percent of patients were Asian, and 58% of patients had a history of never smoking. Based on initial CT imaging, 71 patients (65%) were clinical stage 1, 16 patients (16%) were clinical stage 2, and 20 patients (19%) patients were clinical stage 3. Of the 109 total patients, 107 (98%) underwent a PET-CT and only 39 patients (36%) had a brain MRI. Of the 107 who had PET-CT, 16 patients (15%) were found to have Stage 4 disease. Of the 39 patients who had a brain MRI, 7 patients (18%) were found to have brain metastases. The use of both PET-CT and brain MRIs for patients evaluated for surgery upstaged 20 (18%) patients out of 109 patients to metastatic disease and resulted in 90 (83%) patients undergoing surgery. Conclusions: Our study demonstrates that in real world practice nearly all patients (97%) who were being evaluated for surgery underwent a PET-CT whereas only subset of patients (37%) had a brain MRI. The use of preoperative PET-CT and brain MRIs led to the detection of metastatic disease in 18% of total patients with EGFR-mut NSCLC. These findings underscore the critical role of comprehensive preoperative imaging. We strongly recommend that both a PET-CT and brain MRI are routinely preformed in all patients with EGFR-mut NSCLC prior to surgery to ensure accurate staging and optimize treatment planning. PET CT and brain MRI impact on staging patients with EGFR NSCLC evaluated for surgery. Patients evaluated for surgery 109 Patients who had a PET CT 107 (98%) Patients who had Brain MRI 39 (36%) Upstaged by PET-CT or Brain MRI to IV Disease 20 (18%)

Advancements in genomic profiling led to the discovery of four major molecular subgroups in medulloblastoma (MB), which have now been incorporated into the World Health Organization classification of central nervous system tumors. The current study aimed to determine the prognostic significance of the MB molecular subgroups among children in Malaysia. We assembled MB samples from children <18 years between January 2003 and June 2017 from four pediatric oncology centers in Malaysia. MB was sub-grouped using 850k DNA methylation testing at German Cancer Research Centre, Heidelberg, Germany. Fifty samples from patients diagnosed and treated as MB were identified. Two (4%) of the 50 patients' tumor DNA samples were insufficient for analysis. Of the remaining 48 patients, 41 (85%) samples were confirmed as MB, while for 7 (15%) patients, DNA methylation classification results were discrepant with the histopathological diagnosis of MB, with various other diagnoses. Of the 41 MB patients, 15 patients were stratified as standard-risk (SR), 16 patients as high-risk (HR), and ten as infants (age <3 years old). Molecular subgrouping of the whole cohort revealed four (14%) WNT, 11 (27%) SHH, 10 (24%) Group 3, and 16 (39%) Group 4. Treatment abandonment rates for older children and infants were 22.5% and 10%, respectively. After censoring treatment abandonment, for SR patients, the 5-year event-free survival (EFS) and overall survival (OS) were 43.1% ± 14.7% and 46.9 ± 15.6%, respectively, while in HR, 5-year EFS and OS were both 63.6% ± 14.5%. Infants had a 5-year EFS and OS of 55.6% ± 16.6% and 66.7% ± 15.7%, respectively. WNT tumors had the best 5y-OS, followed by Group 3, Group 4, and SHH in children ≥3 years old. In younger children, SHH MB patients showed favorable outcomes. The study highlights the importance of DNA methylation profiling for diagnostic accuracy. Most infants had SHH MB, and their EFS and OS were comparable to those reported in high-income countries. Due to the relatively small cohort and the high treatment abandonment rate, definite conclusions cannot be made regarding the prognostic significance of molecular subgroups of MB. Implementing this high-technology investigation would assist pathologists in improving the diagnosis and provide molecular subgrouping of MB, permitting subgroup-specific therapies.

Background Radiogenomics of pediatric medulloblastoma (MB) offers an opportunity for MB risk stratification, which may aid therapeutic decision making, family counseling, and selection of patient groups suitable for targeted genetic analysis. Purpose To develop machine learning strategies that identify the four clinically significant MB molecular subgroups. Materials and Methods In this retrospective study, consecutive pediatric patients with newly diagnosed MB at MRI at 12 international pediatric sites between July 1997 and May 2020 were identified. There were 1800 features extracted from T2- and contrast-enhanced T1-weighted preoperative MRI scans. A two-stage sequential classifier was designed-one that first identifies non-wingless (WNT) and non-sonic hedgehog (SHH) MB and then differentiates therapeutically relevant WNT from SHH. Further, a classifier that distinguishes high-risk group 3 from group 4 MB was developed. An independent, binary subgroup analysis was conducted to uncover radiomics features unique to infantile versus childhood SHH subgroups. The best-performing models from six candidate classifiers were selected, and performance was measured on holdout test sets. CIs were obtained by bootstrapping the test sets for 2000 random samples. Model accuracy score was compared with the no-information rate using the Wald test. Results The study cohort comprised 263 patients (mean age ± SD at diagnosis, 87 months ± 60; 166 boys). A two-stage classifier outperformed a single-stage multiclass classifier. The combined, sequential classifier achieved a microaveraged F1 score of 88% and a binary F1 score of 95% specifically for WNT. A group 3 versus group 4 classifier achieved an area under the receiver operating characteristic curve of 98%. Of the Image Biomarker Standardization Initiative features, texture and first-order intensity features were most contributory across the molecular subgroups. Conclusion An MRI-based machine learning decision path allowed identification of the four clinically relevant molecular pediatric medulloblastoma subgroups. © RSNA, 2022 Online supplemental material is available for this article. See also the editorial by Chaudhary and Bapuraj in this issue.

Proteomic profiling of medulloblastoma reveals novel proteins differentially expressed within each molecular subgroup

Medulloblastoma (MB) is recognized as comprising four molecular subgroups with distinct transcriptional profiles, clinical features, and outcomes. Previous studies demonstrate that pediatric MBs present with subgroup-specific MRI manifestations. We hypothesized that combination of anatomical localization and conventional features based on MR imaging can predict these subgroups in adult MBs. MR Imaging manifestations of 125 adult patients with MB were analyzed retrospectively based on pre-operative MRI scans. MB molecular subgroups were evaluated by the expression profiling array and immunohistochemistry. A pediatric MB cohort of 60 patients were analyzed for comparison with data of adult patients. Multiple logistic regression analysis revealed that tumor location (P < 0.0001) and pattern of enhancement (P = 0.0048) were significantly correlated with molecular subgroups in adult MBs. Ninety-two percent of adult MBs were correctly predicted by using logistic regression model based on the anatomical localization patterns and pattern of enhancement. Exclusively intra-cerebellar growth, localization in the rostral cerebellum, and no brainstem contact were specific to adult SHH-MBs. Group 4-MBs in adult were characterized by minimal/no enhancement compared with other two subgroups. Infant SHH-MBs represented significant different localization patterns compared with SHH tumors in children and adults. We identified that molecular subgroups of adult MBs could be well predicted by tumor localization patterns and enhancement pattern. Our study also provided important evidence that MB subgroups in adult possibly derived from different cellular origins.

Objective: To determine the correlation between clinical characteristics and molecular subgroups of medulloblastoma (MB) in Thai pediatric patients at the Queen Sirikit National Institute of Child Health (QSNICH), Thailand. Materials and Methods: MB specimens operated between 2004 and 2018 were classified by Nanostring into four molecular subgroups, including Wingless signaling pathway (WNT), Sonic Hedgehog signaling pathway (SHH), Group 3, and Group 4. For the present cases, the clinical records were retrospectively analyzed. Results: Twenty-two MB cases with complete clinical records were analyzed. Group 4 was the most common molecular subgroup (31.82%), followed by WNT (27.27%), SHH (22.73%), and Group 3 (18.18%). The histologic subtypes included 18, three, and one cases of classic MB, MB with extensive nodularity (MBEN), and large cell MB, respectively. All SHH MBs were found in infants. All MBENs belonged to SHH subgroup, and the large cell MB was Group 3. All six WNT MB cases did not experience tumor recurrence. Five-year cause specific survival rates were 100% in WNT, 60% in SHH, 57.1% in Group 4, and 0% in Group 3. Five-year recurrence-free survival rates were 100% in WNT, 42.9% in Group 4, and 0% in SHH and Group 3. Conclusion: MB is a heterogeneous disease. Classification of MB, especially at the molecular subtype, is helpful for the management and prognostication. Keywords: Medulloblastoma; Molecular subgroup

Medulloblastoma (MB) is a molecularly heterogeneous brain malignancy with large differences in clinical presentation. According to genomic studies, there are at least four distinct molecular subgroups of MB: sonic hedgehog (SHH), wingless/INT (WNT), Group 3, and Group 4. The treatment and outcomes depend on appropriate classification. It is difficult for the classification algorithms to identify these subgroups from an imbalanced MB genomic data set, where the distribution of samples among the MB subgroups may not be equal. To overcome this problem, we used singular value decomposition (SVD) and group lasso techniques to find DNA methylation probe features that maximize the separation between the different imbalanced MB subgroups. We used multinomial regression as a classification method to classify the four different molecular subgroups of MB using the reduced DNA methylation data. Coordinate descent is used to solve our loss function associated with the group lasso, which promotes sparsity. By using SVD, we were able to reduce the 321,174 probe features to just 200 features. Less than 40 features were successfully selected after applying the group lasso, which we then used as predictors for our classification models. Our proposed method achieved an average overall accuracy of 99% based on fivefold cross-validation technique. Our approach produces improved classification performance compared with the state-of-the-art methods for classifying MB molecular subgroups.

MRS for non-invasive medulloblastoma subgrouping

Medulloblastoma (MB) is the most common malignant childhood brain tumor. Molecular subgrouping of MB has become a major determinant of management in high-income countries. Subgrouping is still very limited in low- and middle-income countries (LMICs), and its relevance to management with the incorporation of risk stratification (low risk, standard risk, high risk, and very high risk) has yet to be evaluated in this setting. We describe molecular findings from a tertiary care center in Pakistan and their implications for outcome. Children aged between 3 and 18years diagnosed with MB from April 2014 to December 2020 at Aga Khan University Hospital (AKUH) were included. Subgrouping was performed by NanoString through a collaboration with The Hospital for Sick Children, Toronto. Thirty-seven patients (30 males) were included in this study; median age was 9years. Twenty patients (54.1%) were high-risk, including 12 with metastatic disease. In 30 children, there was a clear molecular subgroup: 4 wingless (WNT) (10.8%), 6 sonic hedgehog (SHH) (16.2%), 3 Group 3 (8.1%), and 17 Group 4 (45.9%) MBs. Molecular subgrouping was inconclusive for three patients (8.1%) and not done in four patients (10.8%). All patients underwent surgery; 26 patients received radiation therapy at AKUH, and 9 were referred outside for radiotherapy; 24 patients received chemotherapy at AKUH (10 outside AKUH). Overall survival (OS) at 5years was 100%, 66.7%, 66.7%, and 88.2% for WNT, SHH, Group 3, and Group 4 patients, respectively (p=0.668). Low- and standard-risk patients had a 5-year OS of 100%, whereas very high-risk patients exhibited a significantly lower OS of 0% (p<0.001). WNT and Group 4 patients had excellent results despite one WNT patient having metastatic disease and eight Group 4 patients being high risk. Our study depicts that molecular subgrouping aids in accurately predicting survival, suggesting the potential benefit of tailored testing and treatment in the LMIC setting.

We aimed to use artificial intelligence to accurately identify molecular subgroups of medulloblastoma (MB), predict clinical outcomes, and incorporate deep learning-based imaging features into the risk stratification. The MRI features were extracted for molecular subgroups by a novel multi-parameter convolutional neural network (CNN) called Bi-ResNet-MB. Then, MR features were used to establish a prognosis model based on XGBoost. Finally, a novel risk stratification system to stratify the patients based on the M2R Score (Machine learning-based Medulloblastoma Risk Score) was proposed. A total of 139 MB patients (36 female, average age 7.27 ± 3.62 years) were treated at Beijing Tiantan Hospital. The Bi-ResNet-MB model excelled in molecular subgroup classification, achieving an average AUC of 0.946 (95% CI: 0.899-0.993). For prognostic prediction, our models achieved AUCs of 0.840 (95% CI: 0.792-0.888), 0.949 (95% CI: 0.899-0.999), and 0.960 (95% CI: 0.915-1.000) for OS, and 0.946 (95% CI: 0.905-0.987), 0.932 (95% CI: 0.875-0.989), and 0.964 (95% CI: 0.921-1.000) for PFS at 1, 3, and 5 years. In an independent validation dataset of 108 patients (33 female, average age 7.11 ± 2.92 years), the average AUC of molecular subgroup classification reached 0.894 (95% CI: 0.797-1.000). For PFS prediction at 1, 3, and 5 years, the AUCs were 0.832 (95% CI: 0.724-0.920), 0.875 (95% CI: 0.781-0.967), and 0.907 (95% CI: 0.760-1.000), respectively. Based on machine learning and MRI data, models for MB molecular subgroups and prognosis prediction and the novel risk stratification system may significantly benefit patients. Question Medulloblastoma exhibits significant heterogeneity, leading to considerable variations in patient prognosis and there is a lack of effective risk assessment strategies. Findings We have constructed a comprehensive machine learning system that excels in subgrouping diagnosis, prognosis assessment, and risk stratification for medulloblastoma patients preoperatively. Clinical relevance The utilization of non-invasive preoperative diagnosis and assessment is advantageous for clinicians in creating personalized treatment plans, particularly for high-risk patients. Additionally, it lays a foundation for the subsequent implementation of neoadjuvant therapy for medulloblastoma.

Medulloblastoma (MB) is one of the most common primary central nervous system tumors in children. Data is lacking of a large cohort of medulloblastoma patients in China. Also, our knowledge on the sensitivity of different molecular subgroups of MB to adjuvant radiation therapy (RT) or chemotherapy (CHT) is still limited. The authors performed a retrospective study of 173 medulloblastoma patients treated at two institutions from 2002 to 2011. Formalin-fixed paraffin embedded (FFPE) tissues were available in all the cases and sections were stained to classify histological and molecular subgroups. Univariate and multivariate analyses were used to investigate prognostic factors. Of 173 patients, there were 118 children and 55 adults, 112 males and 61 females. Estimated 5-year overall survival (OS) rates for all patients, children and adults were 52%, 48% and 63%, respectively. After multivariate analysis, postoperative primary radiation therapy (RT) and chemotherapy (CHT) were revealed as favorable prognostic factors influencing OS and EFS. Postoperative primary chemotherapy (CHT) was found significantly improving the survival of children (p<0.001) while it was not a significant prognostic factor for adult patients. Moreover, patients in WNT subtype had better OS (p = 0.028) than others (SHH and Non-SHH/WNT subtypes) given postoperative adjuvant therapies. Postoperative primary RT was found to be a strong prognostic factor influencing the survival in all histological and molecular subgroups (p<0.001). Postoperative primary CHT was found significantly to influence the survival of classic medulloblastoma (CMB) (OS p<0.001, EFS p<0.001), SHH subgroup (OS p = 0.020, EFS p = 0.049) and WNT subgroup (OS p = 0.003, EFS p = 0.016) but not in desmoplastic/nodular medulloblastoma (DMB) (OS p = 0.361, EFS p = 0.834) and Non-SHH/WNT subgroup (OS p = 0.127, EFS p = 0.055). Our study showed postoperative primary CHT significantly influence the survival of CMB, SHH subgroup and WNT subgroup but not in DMB and Non-SHH/WNT subgroup of MB.

Image-based prediction of molecular subgroups of Medulloblastoma (MB) has the potential to optimize and personalize therapy. The objective of the study is to distinguish between broad molecular subgroups of MB using MR-Texture analysis. Thirty-eight MB patients treated between 2007 and 2020 were retrospectively analyzed. Texture analysis was performed on contrast enhanced T1(T1C) and T2 weighted (T2W) MR images. Manual segmentation was performed on all slices and radiomic features were extracted which included first order, second order (GLCM - Grey level co-occurrence matrix) and shape features. Feature enrichment was done using LASSO (Least Absolute Shrinkage and Selection Operator) regression and thereafter Support Vector Machine (SVM) and a 10-fold cross-validation strategy was used for model development. The area under Receiver Operator Characteristic (ROC) curve was used to evaluate the model. A total of 174 and 170 images were obtained for analysis from the Axial T1C and T2W image datasets. One hundred and sixty-four MR based texture features were extracted. The best model was arrived at by using a combination of 30 GLCM and six shape features on T1C MR sequence. A 10-fold cross-validation demonstrated an AUC of 0.93, 0.9, 0.93, and 0.93 in predicting WNT, SHH, Group 3, and Group 4 MB subgroups, respectively. Radiomic analysis of MR images in MB can predict molecular subgroups with acceptable degree of accuracy. The strategy needs further validation in an external dataset for its potential use in ab initio management paradigms of MBs. Medulloblastoma can be classified into four distinct molecular subgroups using radiomic feature classifier from non-invasive Multiparametric Magnetic resonance imaging. This can have future ramifications in the extent of surgical resection of Medulloblastoma which can ultimately result in reduction of morbidity.

BackgroundCytotoxic neural stem cells (NSCs) have emerged as a promising treatment for Medulloblastoma (MB), the most common malignant primary pediatric brain tumor. The lack of accurate pre-clinical models incorporating surgical resection and tumor recurrence limits advancement in post-surgical MB treatments. Using cell lines from two of the 5 distinct MB molecular sub-groups, in this study, we developed an image-guided mouse model of MB surgical resection and investigate intra-cavity NSC therapy for post-operative MB.MethodsUsing D283 and Daoy human MB cells engineered to express multi-modality optical reporters, we created the first image-guided resection model of orthotopic MB. Brain-derived NSCs and novel induced NSCs (iNSCs) generated from pediatric skin were engineered to express the pro-drug/enzyme therapy thymidine kinase/ganciclovir, seeded into the post-operative cavity, and used to investigate intra-cavity therapy for post-surgical MB.ResultsWe found that surgery reduced MB volumes by 92%, and the rate of post-operative MB regrowth increased 3-fold compared to pre-resection growth. Real-time imaging showed NSCs rapidly homed to MB, migrating 1.6-fold faster and 2-fold farther in the presence of tumors, and co-localized with MB present in the contra-lateral hemisphere. Seeding of cytotoxic NSCs into the post-operative surgical cavity decreased MB volumes 15-fold and extended median survival 133%. As an initial step towards novel autologous therapy in human MB patients, we found skin-derived iNSCs homed to MB cells, while intra-cavity iNSC therapy suppressed post-surgical tumor growth and prolonged survival of MB-bearing mice by 123%.ConclusionsWe report a novel image-guided model of MB resection/recurrence and provide new evidence of cytotoxic NSCs/iNSCs delivered into the surgical cavity effectively target residual MB foci.

Objectives Deep learning (DL) based on MRI of medulloblastoma enables risk stratification, potentially aiding in therapeutic decisions. This study to develop DL models that identify four medulloblastoma molecular subgroups and prognostic related genetic signatures. Materials and Methods In this retrospective study, consecutive patients with newly diagnosed MB at MRI (T1-, T2- and contrast-enhanced T1-weighted) at two medical institutes between January 2015 and June 2023 were identified. Two-stage sequential DL models were designed—MB-CNN that first identifies wingless (WNT), sonic hedgehog (SHH), Group 3, and Group 4. Further, prognostic related genetic signatures DL models (MB-CNN_TP53/MYC/Chr11) were developed to predict TP53 mutation, MYC amplification, and chromosome 11 loss status. A hybrid model combining MB-CNN and conventional data (clinical information and MRI features) was compared to a logistic regression model constructed only with conventional data. Four-classification tasks were evaluated with confusion matrices (accuracy) and two-classification tasks with ROC curves (area under the curve, AUC). Results The datasets comprised 449 patients (mean age ± SD at diagnosis, 13.55 years ± 2.33, 249 males). MB-CNN accurately classified MB subgroups in the external test dataset (accuracy was in the range of 76.29–78.71%). MB-CNN_TP53/MYC/Chr11 models effectively predicted signatures (AUC of TP53 in SHH: 0.91, MYC amplification in Group 3: 0.87, chromosome 11 loss in Group 4: 0.89). The accuracy of hybrid model outperformed the logistic regression model (82.20% vs. 59.14%, P = .009) and showed comparable performance to MB-CNN (82.20% vs. 77.50%, P = .105). Conclusion MRI-based DL models allowed identification of the molecular medulloblastoma subgroups and prognostic related genetic signatures.

Inflammation and immunoreaction markers were correlated with the survival of patients in many tumors. However, there were no reports investigating the relationships between preoperative hematological markers and the prognosis of medulloblastoma (MB) patients based on the molecular subgroups (WNT, SHH, Group 3, and Group 4). A total 144 MB patients were enrolled in the study. The differences of preoperative hematological markers among molecular subgroups of MB were compared by One-way ANOVA method. Kaplan-Meier method was used to calculate the curves of progression free survival (PFS) and overall survival (OS). The comparison of survival rates in different groups were conducted by the Log-rank test. Multivariate analysis was used to evaluate independent prognostic factors. Increased preoperative NLR (neutrophil-to-lymphocyte ratio, PFS, P = 0.004, OS, P < 0.001) and PLR (platelet-to-lymphocyte ratio, PFS, P = 0.028, OS, P = 0.003) predicted poor prognosis in patients with MB, while preoperative MLR (monocyte-to-lymphocyte ratio), MPV (mean platelet volume), PDW (platelet distribution width), and AGR (albumin-to-globulin ratio) were revealed no predictive value on the prognosis of patients with MB. Furthermore, high preoperative NLR and PLR predicted unfavorable prognosis in childhood MB patients. However, preoperative NLR and PLR were not associated with the prognosis in adult MB patients. Multivariate analysis demonstrated preoperative NLR (PFS, P = 0.029, OS, P = 0.005) and PLR (PFS, P = 0.023, OS, P = 0.005) were the independent prognostic factors in MB patients. Emphatically, the levels of preoperative NLR and PLR in Group 3 MB were significantly higher than those in WNT MB. High preoperative NLR was associated with unfavorable OS in Group 3 (P = 0.032) and Group 4 (P = 0.027) tumors. Similarly, increased preoperative PLR predicted poor PFS (P = 0.012) and OS (P = 0.009) in Group 4 tumors. Preoperative NLR and PLR were the potential prognostic markers for MB patients. Preoperative NLR and PLR were significantly associated with the survival of Group 3 and Group 4 tumors.