Related Experiment Video
Updated: Nov 8, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.1K
Anatomy-aided deep learning for medical image segmentation: a review.
Lu Liu1,2, Jelmer M Wolterink1, Christoph Brune1
1Applied Analysis, Department of Applied Mathematics, Faculty of Electrical Engineering, Mathematics and Computer Science, University of Twente, Drienerlolaan 5, 7522 NB, Enschede, The Netherlands.
Physics in Medicine and Biology
|April 27, 2021
Summary
Anatomy-aided deep learning (DL) improves medical image segmentation by incorporating anatomical knowledge, addressing limitations of standard DL methods. This review categorizes approaches and discusses future directions for anatomy-informed AI in medicine.
Area of Science:
- Artificial Intelligence
- Medical Imaging
- Computer-Aided Diagnosis
Background:
- Deep learning (DL) is prevalent in medical image segmentation but faces challenges.
- Standard DL models often struggle with complex anatomical variations and subtle pathologies.
- Manual segmentation relies heavily on anatomical context, a cue often underutilized in DL.
Purpose of the Study:
- To systematically review anatomy-aided deep learning approaches for medical image segmentation.
- To categorize anatomical information types and their representation methods in DL.
- To identify and discuss challenges and future research directions in this field.
Main Methods:
- Comprehensive literature search of over 70 papers on anatomy-aided DL.
- Categorization of anatomical information used (e.g., shape, spatial relationships, topology).
- Analysis of methods for integrating anatomical information into DL models.
Main Results:
- Identified diverse categories of anatomical information and their integration strategies.
- Highlighted challenges such as data scarcity and anatomical variability.
- Provided a structured overview of current anatomy-aided DL methodologies.
Conclusions:
- Anatomy-aided DL offers significant potential to enhance medical image segmentation accuracy and robustness.
- Further research is needed to address current limitations and fully leverage anatomical priors.
- Future work should focus on developing more sophisticated methods for anatomical information integration.
