Related Experiment Video
Updated: Oct 4, 2025
10:06
High-resolution Functional Magnetic Resonance Imaging Methods for Human Midbrain
Published on: May 10, 2012
13.1K
Suggestive annotation of brain MR images with gradient-guided sampling
Chengliang Dai1, Shuo Wang1, Yuanhan Mo1
1Data Science Institute, Imperial College London, United Kingdom.
Medical Image Analysis
|February 8, 2022
Summary
This study introduces an efficient framework for annotating brain MR images, significantly reducing manual effort. The method achieves comparable performance in segmentation tasks using a fraction of the data.
Area of Science:
- Medical Imaging
- Machine Learning
- Artificial Intelligence
Background:
- Supervised machine learning for medical image analysis requires large, manually annotated datasets.
- Curating annotated medical image datasets is time-consuming and resource-intensive.
Purpose of the Study:
- To propose an efficient annotation framework for brain MR images.
- To reduce the manual annotation cost and improve data efficiency in medical imaging.
Main Methods:
- Developed a framework to suggest informative brain MR images for expert annotation.
- Evaluated the framework on brain tumor segmentation (BraTS 2019) and whole brain segmentation (MALC) tasks.
Main Results:
- Brain tumor segmentation achieved comparable performance using only 7% of suggestively annotated samples.
- Whole brain segmentation achieved comparable performance using 42% of suggestively annotated samples.
Conclusions:
- The proposed framework significantly saves manual annotation costs.
- The framework enhances data efficiency in medical imaging applications.
