Related Experiment Video
Updated: Sep 28, 2025
06:48
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
Published on: January 7, 2019
9.0K
Magnetic resonance image-based brain tumour segmentation methods: A systematic review.
Jayendra M Bhalodiya1,1, Sarah N Lim Choi Keung1, Theodoros N Arvanitis1
1Institute of Digital Healthcare, Warwick Manufacturing Group, The University of Warwick, UK.
Digital Health
|March 28, 2022
Summary
Automated brain tumour segmentation using magnetic resonance imaging (MRI) shows U-Net deep learning is highly accurate. Increased open-access data is crucial for improving AI in medical imaging, especially for diffusion- and perfusion-weighted MRI.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computational Biology
Background:
- Brain tumour analysis relies on accurate image segmentation from magnetic resonance imaging (MRI).
- Open-access MRI datasets are vital for developing and validating segmentation methods.
- Artificial intelligence (AI) in medical imaging necessitates larger data repositories for advanced method development.
Purpose of the Study:
- To review automated brain tumour segmentation techniques for medical imaging specialists and clinicians.
- To compare automated segmentation methods against manual segmentation for tumour component identification.
Main Methods:
- Systematic review of 572 brain tumour segmentation studies (2015-2020).
- Analysis of segmentation techniques across various MRI sequences (T1, T2, FLAIR, DWI, PWI) and AI approaches (deep learning, physics-based, semi-automatic).
- Synthesis of methods based on MRI sequences, study population, technical approach, and performance metrics (e.g., Dice score).
Main Results:
- T1, T2, gadolinium-enhanced T1, and FLAIR MRI sequences are most frequently used in segmentation algorithms.
- Limited utilization of diffusion-weighted imaging (DWI) and perfusion-weighted imaging (PWI) was observed.
- U-Net deep learning demonstrated high accuracy (Dice score 0.9) for MRI-based brain tumour segmentation and was the most cited technique.
Conclusions:
- U-Net is a highly effective deep learning technology for MRI-based brain tumour segmentation.
- Encouraging contributions of open-access datasets is essential for improving AI algorithm training, testing, and validation.
- Further development is needed for segmentation techniques utilizing DWI and PWI, particularly regarding data availability.
