Related Experiment Video
Updated: Jul 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Deep learning image segmentation approaches for malignant bone lesions: a systematic review and meta-analysis
Joseph M Rich1, Lokesh N Bhardwaj1, Aman Shah2
1Keck School of Medicine, University of Southern California, Los Angeles, CA, United States.
Frontiers in Radiology
|August 24, 2023
Summary
Deep learning effectively segments malignant bone lesions on CT, MRI, and PET/CT scans. This review highlights U-Net variations and data augmentation as key strategies for improved accuracy in medical imaging analysis.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate quantification of malignant bone lesions is crucial but challenging for radiologists.
- Deep learning (DL) offers automated image segmentation solutions for medical imaging, including bone lesions.
Purpose of the Study:
- To review deep learning-based image segmentation methods for malignant bone lesions.
- To cover applications across Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and Positron-Emission Tomography/CT (PET/CT).
Main Methods:
- Systematic literature search in PubMed, Embase, Web of Science, and Scopus.
- Included 41 original articles published between February 2017 and March 2023.
- Followed Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
Main Results:
- Most studies utilized MRI, followed by CT and PET/CT.
- U-Net architecture and its modifications were commonly used.
- High performance was reported, with median Dice Similarity Coefficients (DSC) between 0.85-0.9 across modalities.
Conclusions:
- Deep learning demonstrates significant potential for segmenting malignant osseous lesions.
- Strategies like data augmentation and U-Net modifications enhance performance.
- Future work should focus on dataset homogeneity and clinical generalization.
