Related Experiment Video
Updated: Aug 31, 2025
10:25
Brain Infarct Segmentation and Registration on MRI or CT for Lesion-symptom Mapping
Published on: September 25, 2019
48.3K
Convolutional Neural Network Techniques for Brain Tumor Classification (from 2015 to 2022): Review, Challenges, and
Yuting Xie1, Fulvio Zaccagna1,2, Leonardo Rundo3
1Department of Biomedical and Neuromotor Sciences, University of Bologna, 40126 Bologna, Italy.
Diagnostics (Basel, Switzerland)
|August 26, 2022
Summary
Convolutional neural networks show promise for brain tumor diagnosis from MR images. However, challenges in data, validation, and clinical usability hinder widespread application of these deep learning methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Neuroscience
Background:
- Convolutional Neural Networks (CNNs) are a key deep learning technique applied to brain tumor diagnosis using Magnetic Resonance (MR) images.
- Despite advancements, significant challenges impede the clinical integration of CNN-based brain tumor classification.
Purpose of the Study:
- To conduct a comprehensive review of studies utilizing CNN architectures for brain tumor classification on MR images.
- To identify effective strategies and critical impediments in the development and clinical application of this technology.
Main Methods:
- Systematic literature search to identify relevant studies.
- Data extraction on training data, classification tasks, network architectures, validation, and performance metrics.
- Evaluation of clinical relevance and identification of limitations and future research directions.
Main Results:
- Reviewed 83 studies, noting diverse classification problems and CNN strategies.
- Reported accuracies varied: 91.63-100% for tumor type differentiation and 60.0-99.46% for glioma grading.
- Inconsistencies in reporting hinder clear identification of superior methodologies; clinical usability remains a focus.
Conclusions:
- CNNs demonstrate strong potential for brain tumor classification from MR images.
- Methodological inconsistencies and a lack of focus on clinical usability present key challenges.
- Future research should address these limitations to advance clinical translation of CNN algorithms.
