Choroid Plexus Tumors, Craniopharyngioma, Ependymoma, Glioma, Medulloblastoma, Pediatric Brain Tumors
Conditions
Keywords
Artificial Intelligence, Molecular Subtyping, Diagnostic Performance, Deep Learning, Pediatric Neuro-Oncology
Brief summary
This multicenter observational cohort study aims to develop and validate an artificial intelligence (AI)-assisted diagnostic system for preoperative molecular subtyping of pediatric brain tumors using routine magnetic resonance imaging (MRI). The study will include seven major pediatric brain tumor categories: glioma, medulloblastoma, ependymoma, atypical teratoid/rhabdoid tumor (AT/RT), intracranial germ cell tumors, craniopharyngioma, and choroid plexus tumors. The study includes a retrospective cohort for model development and internal/external validation, and a prospective cohort for further validation. Retrospective data will be collected from pediatric patients who underwent first surgical treatment between January 1, 2020 and December 31, 2025. Prospective enrollment will begin on July 15, 2026, with an anticipated sample size of 150 participants. The AI system will analyze preoperative MRI sequences, including T1-weighted, contrast-enhanced T1-weighted, T2-weighted, and FLAIR images, to predict key molecular markers and integrated diagnostic categories. The primary objective is to evaluate the diagnostic performance of the AI system for prespecified molecular prediction tasks using postoperative histopathology and molecular testing as the reference standard. Secondary objectives include assessing agreement with integrated diagnosis, comparing performance against blinded radiologists, and exploring prognostic associations of AI-predicted subgroups.
Detailed description
Pediatric brain tumors are the most common solid tumors in children and represent a highly heterogeneous group of diseases with marked variation in histology, molecular alterations, anatomic location, treatment response, and prognosis. Several molecular features, including H3K27M mutation, BRAF V600E mutation, ZFTA fusion, SMARCB1 loss, CTNNB1 mutation, and TP53 alteration, are clinically important for diagnostic classification, risk stratification, prognosis assessment, and treatment planning. However, most molecular characterization currently depends on postoperative tissue-based testing, and noninvasive preoperative prediction remains limited. This study is designed to evaluate an AI-assisted MRI-based diagnostic system for pediatric brain tumors in a real-world multicenter observational setting. The study will include seven target tumor categories: glioma, medulloblastoma, ependymoma, atypical teratoid/rhabdoid tumor, intracranial germ cell tumors, craniopharyngioma, and choroid plexus tumors. The retrospective component will collect multimodal data, including clinical variables, preoperative MRI, pathology reports, molecular testing results, treatment information, and follow-up data, from eligible pediatric patients treated from January 1, 2020 through December 31, 2025. The prospective component will consecutively enroll eligible patients from July 15, 2026 onward for additional validation of model performance. Preoperative MRI data will be preprocessed using standardized procedures, including bias field correction, skull stripping, isotropic resampling, and intensity normalization. The AI model will be developed to support classification of tumor type and prediction of key molecular subtypes/markers from presurgical MRI. Model performance will be evaluated using postoperative pathology and molecular testing as the reference standard. The primary endpoint is the area under the receiver operating characteristic curve (AUC) for prespecified molecular prediction tasks. Secondary analyses will evaluate agreement between AI output and integrated final diagnosis, comparative performance against blinded radiologists on independent test sets, multiclass tumor classification performance, biomarker-specific sensitivity and specificity, and progression-free survival stratified by AI-predicted subgroup. This study is observational and is not intended for medical device registration. Biospecimen banking is not a registration objective of this study.
Interventions
None listed
Sponsors
Huashan Hospital
Xinhua Hospital, Shanghai Jiao Tong University School of Medicine
Quzhou Hospital Affiliated to Wenzhou Medical University
Ruitong Intelligent Medical (Shanghai) Technology Co., Ltd.
Study design
Observational model
COHORT
Time perspective
OTHER
Eligibility
Sex/Gender
ALL
Age
0 Years to 17 Years
Healthy volunteers
No
Inclusion criteria
* Age younger than 18 years at the time of index surgery. * Evaluated at a participating study center and scheduled for first surgical treatment of a suspected target pediatric brain tumor. * Preoperative brain MRI available before surgery, including at minimum T1-weighted, contrast-enhanced T1-weighted, T2-weighted, and FLAIR sequences in DICOM format; MRI preferably performed within 7 days before surgery and before biopsy or tumor-directed therapy. * Postoperative histopathology confirming one of the following target tumor categories: glioma, medulloblastoma, ependymoma, atypical teratoid/rhabdoid tumor, intracranial germ cell tumors, craniopharyngioma, or choroid plexus tumors. * For the prospective cohort, written informed consent provided by a parent or legal guardian, with child assent obtained when appropriate according to age, understanding, and local ethics requirements.
Exclusion criteria
* Postoperative pathology confirming a non-target tumor type. * Recurrent tumor, repeat surgery, or prior tumor-directed surgery before the index surgery. * Preoperative MRI of inadequate quality for analysis, including severe motion artifact, severe susceptibility/metal artifact, or incomplete field of view. * Prior biopsy, radiotherapy, chemotherapy, or other tumor-directed treatment before the index preoperative MRI that is judged to substantially affect imaging interpretation. * Concurrent malignant disease other than the target brain tumor. * Inability to comply with follow-up requirements in the prospective cohort, in the investigator's judgment, because of severe comorbidity or other practical limitations.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Area Under the Receiver Operating Characteristic Curve (AUC) for MRI-Based Prediction of Prespecified Molecular Markers | Assessed at final model evaluation using all eligible retrospective cases collected from January 1, 2020 through December 31, 2025 and all eligible prospectively enrolled cases with available reference-standard data collected from July 15, 2026 through D | Diagnostic discrimination of the AI-assisted system for binary prediction of prespecified key molecular markers or molecular subtypes from preoperative MRI, using postoperative histopathology and molecular testing as the reference standard. AUC values and 95% confidence intervals will be calculated for each prespecified molecular prediction task. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Agreement Between AI-Based Diagnosis and Integrated Final Diagnosis | Assessed at final diagnostic adjudication for each eligible participant from study start on July 15, 2026 through study completion on December 30, 2029, including retrospective cases with complete reference-standard data. | Agreement between AI-predicted diagnostic output based on preoperative MRI and the integrated final diagnosis established from imaging, postoperative histopathology, and molecular testing. Agreement will be quantified using Cohen's kappa coefficient. |
| Comparative Diagnostic Performance of the AI System Versus Blinded Radiologists | Assessed at blinded reader evaluation after completion of dataset curation and test set locking, anticipated by December 30, 2029. | Comparison of sensitivity, specificity, accuracy, and AUC between the AI system and radiologists blinded to pathology and molecular results on a predefined independent test dataset. |
| Macro-Average AUC for Seven-Class Tumor Classification | Assessed at final model evaluation using eligible cases with complete imaging and reference-standard diagnostic data through December 30, 2029. | Macro-average area under the receiver operating characteristic curve for classification of the seven target pediatric brain tumor categories. |
| Weighted F1 Score for Seven-Class Tumor Classification | Assessed at final model evaluation using eligible cases with complete imaging and reference-standard diagnostic data through December 30, 2029. | Weighted F1 score of the AI system for multiclass classification across glioma, medulloblastoma, ependymoma, atypical teratoid/rhabdoid tumor, intracranial germ cell tumors, craniopharyngioma, and choroid plexus tumors. |
| Sensitivity and Specificity for Prediction of Key Molecular Biomarkers | Assessed at final model evaluation using cases with complete biomarker reference-standard results through December 30, 2029. | Sensitivity and specificity of the AI system for prediction of tumor-specific biomarkers, including but not limited to H3K27M mutation, BRAF V600E mutation, ZFTA fusion, SMARCB1 loss, CTNNB1 mutation, and TP53 alteration, depending on tumor type and data availability. |
Contacts
CONTACTLei Jin
Outcome results
None listed