MRI-Based Radiomics Enhances Diagnosis and Prognosis in Childhood Brain Tumors

Strategic Insights -

Childhood brain tumors pose significant diagnostic and treatment challenges despite being rare. Researchers have been exploring the use of magnetic resonance imaging (MRI) radiomics to improve diagnosis and prognosis. Radiomics involves extracting valuable information from standard imaging modalities like MRI, which is often preferred for pediatric brain tumor imaging due to its non-invasive nature and lack of radiation exposure. A systematic review of 559 MRI-based radiomics studies has identified key pre-processing methods, features, and classification algorithms used in pediatric brain tumor imaging. The study highlights the potential of radiomics to enhance our understanding of tumor biology, leading to improved classification, treatment planning, and management of childhood brain tumors.

Key Takeaways:

  • The study relied on 559 MRI-based radiomics studies from the PubMed and Engineering Village Compendex databases to conduct a systematic review.
  • Nineteen studies were included in the review, primarily focusing on ependymoma (EP) and medulloblastoma (MB) brain tumors.
  • Common pre-processing methods used in the studies included intensity normalization (11 studies) and bias correction (4 studies).
  • The most frequently used features were GLCM (12 studies) and GLRLM (7 studies), with LASSO (8 studies) and PCA (2 studies) being leading selection methods.
  • SVM was the most commonly used classification algorithm (9 studies), with an AUC range of 0.858-0.977.
  • The study recommended identifying and integrating the most informative radiomic features to enhance diagnostic and prognostic accuracy in childhood brain tumors.
  • The researchers highlighted the importance of addressing challenges such as limited datasets and varied imaging protocols.

Statistics:

  • 559 MRI-based radiomics studies were included in the systematic review.
  • Nineteen studies addressed ependymoma (EP) and medulloblastoma (MB) brain tumors.
  • 11 studies used intensity normalization pre-processing methods.
  • 4 studies used bias correction pre-processing methods.
  • 12 studies used GLCM features.
  • 7 studies used GLRLM features.
  • 8 studies used LASSO selection methods.
  • 2 studies used PCA selection methods.
  • 9 studies used SVM classification algorithms.

Sources:

  • (Magnetic resonance imaging (MRI) radiomics in paediatric neuro-oncology: A systematic review of clinical applications, feature interpretation, and biological insights in the characterisation and management of childhood brain tumours). DIGITAL HEALTH, 2025;11:20552076251336285.
  • (NewsRx LLC) - cited as "University of Birmingham Reports Findings in Brain Cancer [Magnetic resonance imaging (MRI) radiomics in paediatric neuro-oncology: A systematic review of clinical applications, feature interpretation, and biological insights in the ...]." Pediatrics Week. May 17, 2025; p 586.