Brain tumors are the most common solid tumors in children and young adults, but the scarcity of large histopathology datasets has limited the application of computational pathology in this group. This study implements two weakly supervised multiple-instance learning (MIL) approaches on patch features obtained from state-of-the-art histology-specific foundation models to classify pediatric brain tumors in hematoxylin and eosin whole slide images (WSIs) from a multi-center Swedish cohort. WSIs from 540 subjects (age 8.5 ± 4.9 years) diagnosed with brain tumors were gathered from the six Swedish university hospitals. Instance (patch)-level features were obtained from WSIs using three pre-trained feature extractors: ResNet50, UNI, and CONCH. Instances were aggregated using attention-based MIL (ABMIL) or clustering-constrained attention MIL (CLAM) for patient-level classification. Models were evaluated on three classification tasks based on the hierarchical classification of pediatric brain tumors: tumor category, family, and type. Model generalization was assessed by training on data from two of the centers and testing on data from four other centers. Model interpretability was evaluated through attention mapping. The highest classification performance was achieved using UNI features and ABMIL aggregation, with Matthew's correlation coefficient of 0.76 ± 0.04, 0.63 ± 0.04, and 0.60 ± 0.05 for tumor category, family, and type classification, respectively. When evaluating generalization, models utilizing UNI and CONCH features outperformed those using ResNet50. However, the drop in performance from the in-site to out-of-site testing was similar across feature extractors. These results show the potential of state-of-the-art computational pathology methods in diagnosing pediatric brain tumors at different hierarchical levels with fair generalizability on a multi-center national dataset.

The quantification of the Ki-67 labeling index (LI) is critical for assessing tumor proliferation and prognosis in tumors, yet manual scoring remains a common practice. This study presents an automated framework for Ki-67 scoring in whole slide images (WSIs) developed for research settings, using an Apache Groovy code script for QuPath and complemented by a Python post-processing script that provides cell density maps and summary tables. Tissue segmentation is performed, then cell segmentation is conducted using StarDist, a deep learning model, followed by adaptive thresholding to classify Ki-67 positive and negative nuclei. The pipeline was applied to a cohort of 632 pediatric brain tumor cases with 734 Ki-67 WSIs from the Children's Brain Tumor Network. Medulloblastoma showed the highest Ki-67 LI (median: 19.84), followed by atypical teratoid rhabdoid tumor (median: 19.36). Moderate values were observed in brainstem glioma-diffuse intrinsic pontine glioma (median: 11.50), high-grade glioma (grades 3, 4) (median: 9.50), and ependymoma (median: 5.88). Lower indices were found in meningioma (median: 1.84), while the lowest were seen in low-grade glioma (grades 1, 2) (median: 0.85), dysembryoplastic neuroepithelial tumor (median: 0.63), and ganglioglioma (median: 0.50). The results aligned with the consensus of the oncology, demonstrating a significant correlation in Ki-67 LI across most of the tumor families/types.

BACKGROUND AND PURPOSE: CNS embryonal tumor with pleomorphic adenoma gene-like 1 (PLAGL1)/pleomorphic adenoma gene-like 2 (PLAGL2) amplification (ET, PLAGL) is a newly identified, highly malignant pediatric tumor. Systematic MRI descriptions of ET, PLAGL are currently lacking. MATERIALS AND METHODS: MRI data from 19 treatment-na?ve patients with confirmed ET, PLAGL were analyzed. Evaluation focused on anatomic involvement, tumor localization, MRI signal characteristics, DWI behavior, and the presence of necrosis and hemorrhage. Descriptive statistics (median, interquartile range, percentage) were assessed. RESULTS: Ten patients had PLAGL1 and nine had PLAGL2 amplifications. The solid components of the tumors were often multinodular with heterogeneous enhancement (mild to intermediate in 47% and intermediate to strong in 47% of cases). Nonsolid components included cysts in 47% and necrosis in 84% of the cases. The tumors showed heterogeneous T2WI hyper- and isointensity (74%), relatively little diffusion restriction (ADC values less than contralateral normal-appearing WM in 36% of cases with available DWI), and tendencies toward hemorrhage/calcification (42%). No reliable distinction was found between PLAGL1- and PLAGL2-amplified tumors or compared with other embryonal CNS tumors. CONCLUSIONS: The study contributes to understanding the imaging characteristics of ET, PLAGL. It underscores the need for collaboration in studying rare pediatric tumors and advocates the use of harmonized imaging protocols for better characterization.

Background

Central nervous system (CNS) tumors are the second most common childhood malignancy and a leading cause of cancer-related mortality. This national population-based study describes childhood primary CNS tumors diagnosed in Sweden over 40 years, incorporating the latest WHO classification and long-term survival data.

Methods

All primary CNS tumors in children (0-14 years), diagnosed between 1984 and 2021 and registered in Swedish Cancer Registries, were reviewed. Data on tumor location, histology, and the reclassified supratentorial CNS-PNETs were incorporated. Tumors were categorized according to the International Classification of Childhood Cancer, third edition. Incidence and survival rates were analyzed.

Results

Overall, 2954 children (<15 years) were diagnosed with a CNS tumor in Sweden 1984-2021. The average incidence rate was 4.6/100 000 children per year, and it remained stable during the study period. Astrocytomas constituted 49%, embryonal tumors 17%, and ependymomas 7% of cases. Five-year overall survival improved from 75% to 82% over the study period. However, several tumor types showed a continued decline in survival at 30 years of follow-up.

Conclusions

This 40-year population-based study provides comprehensive data on childhood CNS tumors in Sweden. The incidence remained stable during the 4 decades. The distribution of tumor diagnoses is in line with other countries. Survival has improved over time, but for some diagnoses, late mortality is seen many years after primary diagnosis, highlighting the importance of long-term follow-up.

Purpose: To implement and evaluate deep learning-based methods for the classification of pediatric brain tumors in MR data.

Materials and methods: A subset of the “Children’s Brain Tumor Network” dataset was retrospectively used (n=178 subjects, female=72, male=102, NA=4, age-range [0.01, 36.49] years) with tumor types being low-grade astrocytoma (n=84), ependymoma (n=32), and medulloblastoma (n=62). T1w post-contrast (n=94 subjects), T2w (n=160 subjects), and ADC (n=66 subjects) MR sequences were used separately. Two deep-learning models were trained on transversal slices showing tumor. Joint fusion was implemented to combine image and age data, and two pre-training paradigms were utilized. Model explainability was investigated using gradient-weighted class activation mapping (Grad-CAM), and the learned feature space was visualized using principal component analysis (PCA).

Results: The highest tumor-type classification performance was achieved when using a vision transformer model pre-trained on ImageNet and fine-tuned on ADC images with age fusion (MCC: 0.77 ± 0.14 Accuracy: 0.87 ± 0.08), followed by models trained on T2w (MCC: 0.58 ± 0.11, Accuracy: 0.73 ± 0.08) and T1w post-contrast (MCC: 0.41 ± 0.11, Accuracy: 0.62 ± 0.08) data. Age fusion marginally improved the model’s performance. Both model architectures performed similarly across the experiments, with no differences between the pre-training strategies. Grad-CAMs showed that the models’ attention focused on the brain region. PCA of the feature space showed greater separation of the tumor-type clusters when using contrastive pre-training.

Conclusion: Classification of pediatric brain tumors on MR-images could be accomplished using deep learning, with the top-performing model being trained on ADC data, which is used by radiologists for the clinical classification of these tumors.

Background Constitutional mismatch repair deficiency (CMMRD) syndrome is a rare and aggressive cancer predisposition syndrome. Because a scarcity of data on this condition contributes to management challenges and poor outcomes, we aimed to describe the clinical spectrum, cancer biology, and impact of genetics on patient survival in CMMRD. Methods In this cohort study, we collected cross-sectional and longitudinal data on all patients with CMMRD, with no age limits, registered with the International Replication Repair Deficiency Consortium (IRRDC) across more than 50 countries. Clinical data were extracted from the IRRDC database, medical records, and physician-completed case record forms. The primary objective was to describe the clinical features, cancer spectrum, and biology of the condition. Secondary objectives included estimations of cancer incidence and of the impact of the specific mismatch-repair gene and genotype on cancer onset and survival, including after cancer surveillance and immunotherapy interventions. Findings We analysed data from 201 patients (103 males, 98 females) enrolled between June 5, 2007 and Sept 9, 2022. Median age at diagnosis of CMMRD or a related cancer was 8.9 years (IQR 5.9-12.6), and median follow-up from diagnosis was 7.2 years (3.6-14.8). Endogamy among minorities and closed communities contributed to high homozygosity within countries with low consanguinity. Frequent dermatological manifestations (117 [93%] of 126 patients with complete data) led to a clinical overlap with neurofibromatosis type 1 (35 [28%] of 126). 339 cancers were reported in 194 (97%) of 201 patients. The cumulative cancer incidence by age 18 years was 90% (95% CI 80-99). Median time between cancer diagnoses for patients with more than one cancer was 1.9 years (IQR 0.8-3.9). Neoplasms developed in 15 organs and included early-onset adult cancers. CNS tumours were the most frequent (173 [51%] cancers), followed by gastrointestinal (75 [22%]), haematological (61 [18%]), and other cancer types (30 [9%]). Patients with CNS tumours had the poorest overall survival rates (39% [95% CI 30-52] at 10 years from diagnosis; log-rank p&lt;0.0001 across four cancer types), followed by those with haematological cancers (67% [55-82]), gastrointestinal cancers (89% [81-97]), and other solid tumours (96% [88-100]). All cancers showed high mutation and microsatellite indel burdens, and pathognomonic mutational signatures. MLH1 or MSH2 variants caused earlier cancer onset than PMS2 or MSH6 variants, and inferior survival (overall survival at age 15 years 63% [95% CI 55-73] for PMS2, 49% [35-68] for MSH6, 19% [6-66] for MLH1, and 0% for MSH2; p&lt;0.0001). Frameshift or truncating variants within the same gene caused earlier cancers and inferior outcomes compared with missense variants (p&lt;0.0001). The greater deleterious effects of MLH1 and MSH2 variants as compared with PMS2 and MSH6 variants persisted despite overall improvements in survival after surveillance or immune checkpoint inhibitor interventions. fInterpretation The very high cancer burden and unique genomic landscape of CMMRD highlight the benefit of comprehensive assays in timely diagnosis and precision approaches toward surveillance and immunotherapy. These data will guide the clinical management of children and patients who survive into adulthood with CMMRD. Copyright (c) 2024 Elsevier Ltd. All rights reserved, including those for text and data mining, AI training, and similar technologies.

Visual difficulties are common after brain tumors, despite a lack of visual complaints at diagnosis. These include difficulties with eye movements, visual coordination, vergence, accommodation, and photophobia, in addition to more obvious problems such as visual field defects.This case report presents the results of a thorough neuro-visual evaluation in a boy with sequelae after a brain tumor including intermittent double vision that was not explained by routine visual examination. Subjective complaints included poor reading perseverance, intermittent blurred and double vision, headache around the eyes when performing near activities, less efficient eye movement behavior in reading tasks, and increased sensitivity to visual motion. The patient participated in a multidisciplinary visual rehabilitation program that included reading glasses with prism compensation and tinted glasses, as well as training with the aim of improving eye teaming, near vision functions, and perseverance in eye movements.The patient responded quickly to the vision therapy program, with positive changes after just four weeks. Repeated neuro-visual evaluations over eight months showed remarkable improvements that were stable over time. This encouraging case report supports the notion that neuro-visual evaluation and rehabilitation should be included in the follow-up of patients after brain tumors.

BackgroundMolecular analyses have shown that tumours diagnosed as supratentorial primitive neuro-ectodermal tumours of the central nervous system (CNS-PNETs) in the past represent a heterogenous group of rare childhood tumours including high-grade gliomas (HGG), ependymomas, atypical teratoid/rhabdoid tumours (AT/RT), CNS neuroblastoma with forkhead box R2 (FOXR2) activation and embryonal tumour with multi-layered rosettes (ETMR). All these tumour types are rare and long-term clinical follow-up data are sparse. We retrospectively re-evaluated all children (0-18 years old) diagnosed with a CNS-PNET in Sweden during 1984-2015 and collected clinical data.MethodsIn total, 88 supratentorial CNS-PNETs were identified in the Swedish Childhood Cancer Registry and from these formalin-fixed paraffin-embedded tumour material was available for 71 patients. These tumours were histopathologically re-evaluated and, in addition, analysed using genome-wide DNA methylation profiling and classified by the MNP brain tumour classifier.ResultsThe most frequent tumour types, after histopathological re-evaluation, were HGG (35%) followed by AT/RT (11%), CNS NB-FOXR2 (10%) and ETMR (8%). DNA methylation profiling could further divide the tumours into specific subtypes and with a high accuracy classify these rare embryonal tumours. The 5 and 10-year overall survival (OS) for the whole CNS-PNET cohort was 45% +/- 12% and 42% +/- 12%, respectively. However, the different groups of tumour types identified after re-evaluation displayed very variable survival patterns, with a poor outcome for HGG and ETMR patients with 5-year OS 20% +/- 16% and 33% +/- 35%, respectively. On the contrary, high PFS and OS was observed for patients with CNS NB-FOXR2 (5-year 100% for both). Survival rates remained stable even after 15-years of follow-up.ConclusionsOur findings demonstrate, in a national based setting, the molecular heterogeneity of these tumours and show that DNA methylation profiling of these tumours provides an indispensable tool in distinguishing these rare tumours. Long-term follow-up data confirms previous findings with a favourable outcome for CNS NB-FOXR2 tumours and poor chances of survival for ETMR and HGG.

Importance Medulloblastoma recurrence in patients who have previously received irradiation has a dismal prognosis and lacks a standard salvage regimen.Objective To evaluate the response rate of pediatric patients with medulloblastoma recurrence using an antiangiogenic metronomic combinatorial approach (Medulloblastoma European Multitarget Metronomic Anti-Angiogenic Trial [MEMMAT]).Design, Setting, and Participants This phase 2, investigator-initiated, multicenter nonrandomized controlled trial assessed 40 patients with relapsed or refractory medulloblastoma without a ventriculoperitoneal shunt who were younger than 20 years at original diagnosis. Patients were enrolled between April 1, 2014, and March 31, 2021.Interventions Treatment consisted of daily oral thalidomide, fenofibrate, celecoxib, and alternating 21-day cycles of low-dose (metronomic) oral etoposide and cyclophosphamide, supplemented by intravenous bevacizumab and intraventricular therapy consisting of alternating etoposide and cytarabine.Main Outcomes and Measures The primary end point was response after 6 months of antiangiogenic metronomic therapy. Secondary end points included progression-free survival (PFS), overall survival (OS), and quality of life. Adverse events were monitored to assess safety.Results Of the 40 patients (median [range] age at treatment start, 10 [4-17] years; 25 [62.5%] male) prospectively enrolled, 23 (57.5%) achieved disease control after 6 months of treatment, with a response detected in 18 patients (45.0%). Median OS was 25.5 months (range, 10.9-40.0 months), and median PFS was 8.5 months (range, 1.7-15.4 months). Mean (SD) PFS at both 3 and 5 years was 24.6% (7.9%), while mean (SD) OS at 3 and 5 years was 43.6% (8.5%) and 22.6% (8.8%), respectively. No significant differences in PFS or OS were evident based on molecular subgroup analysis or the number of prior recurrences. In patients demonstrating a response, mean (SD) overall 5-year PFS was 49.7% (14.3%), and for patients who remained progression free for the first 12 months of treatment, mean (SD) 5-year PFS was 66.7% (16.1%). Treatment was generally well tolerated. Grade 3 to 4 treatment-related adverse events included myelosuppression, infections, seizures, and headaches. One heavily pretreated patient with a third recurrence died of secondary acute myeloid leukemia.Conclusions and Relevance This feasible and well-tolerated MEMMAT combination regimen demonstrated promising activity in patients with previously irradiated recurrent medulloblastoma. Given these results, this predominantly oral, well-tolerated, and outpatient treatment warrants further evaluation.

The quantification of the Ki-67 labeling index (LI) is critical for assessing tumor proliferation and prognosis in tumors, yet manual scoring remains a common practice. This study presents an automated workflow for Ki-67 scoring in whole slide images (WSIs) using an Apache Groovy code script for QuPath, complemented by a Python-based post-processing script, providing cell density maps and summary tables. The tissue and cell segmentation are performed using StarDist, a deep learning model, and adaptive thresholding to classify Ki-67 positive and negative nuclei. The pipeline was applied to a cohort of 632 pediatric brain tumor cases with 734 Ki-67-stained WSIs from the Children’s Brain Tumor Network. Medulloblastoma showed the highest Ki-67 LI(median: 19.84), followed by atypical teratoid rhabdoid tumor (median: 19.36). Moderate values were observed in brainstem glioma-diffuse intrinsic pontine glioma (median: 11.50), high-grade glioma (grades 3 & 4) (median: 9.50), and ependymoma (median: 5.88). Lower indices were foundin meningioma (median: 1.84), while the lowest were seen in low-grade glioma (grades 1 & 2)(median: 0.85), dysembryoplastic neuroepithelial tumor (median: 0.63), and ganglioglioma (median:0.50). The results aligned with the consensus of the oncology, demonstrating a significant correlationin Ki-67 LI across most of the tumor families/types, with high malignancy tumors showing thehighest proliferation indices and lower malignancy tumors exhibiting lower Ki-67 LI. The automated approach facilitates the assessment of large amounts of Ki-67 WSIs in research settings.