Related Experiment Video
Updated: Jun 7, 2026
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
Advancing Brain Tumor Diagnosis Using Deep Learning: A Systematic and Critical Review on Methodological Approaches to
Simona Aresta1, Cinzia Palmirotta2, Muhammad Asim1
1Ailice Labs, Department of Science, Technology and Society, University School for Advanced Studies IUSS Pavia, 27100 Pavia, Italy.
Brain Sciences
|May 27, 2026
Summary
Deep learning (DL) models show promise for brain tumor segmentation and classification using magnetic resonance imaging (MRI). Further validation and explainable AI are crucial for clinical use.
Area of Science:
- Neuro-oncology
- Medical Imaging
- Artificial Intelligence
Background:
- Brain tumors, particularly gliomas, are aggressive cancers.
- Magnetic resonance imaging (MRI) is a key non-invasive diagnostic tool.
- Deep learning (DL) and artificial intelligence (AI) offer advanced analysis methods.
Purpose of the Study:
- To systematically review DL and AI applications in brain tumor segmentation and classification.
- To evaluate the performance and limitations of these methods.
- To identify future research directions for clinical applicability.
Main Methods:
- Systematic review of PubMed and Scopus (2022-March 2025).
- Inclusion criteria and data extraction by independent authors.
- Quality and risk of bias assessment using QUADAS checklist.
Main Results:
- 31 studies included; 8 covered both segmentation and classification.
- Segmentation models achieved high Dice Similarity Coefficients (DSC), especially with advanced architectures (>90%).
- Classification models reported high accuracy (91.3%-99.4%), sensitivity, and specificity, but generalizability concerns remain.
Conclusions:
- DL models show significant potential for glioma segmentation and classification.
- Standardized validation, multi-center data, and explainable AI are essential.
- Improving transparency and robustness is key for clinical integration.
