Related Experiment Video
Updated: Jun 1, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.6K
Metastasis lesion segmentation from bone scintigrams using encoder-decoder architecture model with multi-attention
Ailing Xie1,2, Qiang Lin1,2,3, Yang He2,3
1School of Mathematics and Computer Science, Northwest Minzu University, Lanzhou, China.
Quantitative Imaging in Medicine and Surgery
|January 22, 2025
Summary
This study introduces a deep learning model for segmenting bone metastasis (BM) on scintigraphy scans. The model improves diagnostic accuracy by automatically identifying lesions, outperforming existing methods.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Bone scintigraphy faces challenges in diagnosing bone metastasis (BM) due to resolution limits and lesion variability.
- Deep learning offers automated solutions for identifying and delineating BM lesions, improving diagnostic accuracy.
Purpose of the Study:
- To develop a deep learning-based approach for automatic segmentation of bone scintigrams.
- To enhance the accuracy of diagnosing bone metastasis (BM) through automated lesion segmentation.
Main Methods:
- An encoder-decoder deep learning model was developed for segmentation.
- Multi-attention learning (Non-local Attention, Vision Transformer) and multi-scale learning strategies were employed.
- The model was designed to enhance skeletal contrast and highlight hotspots for accurate lesion detection.
Main Results:
- The proposed model achieved a Dice Similarity Coefficient (DSC) of 0.6720 on SPECT bone scintigrams.
- Compared to existing models, the proposed method showed improvements of 5.6% in DSC, 2.03% in Precision, and 7.9% in Recall.
Conclusions:
- The developed segmentation model is a promising tool for automatic extraction of metastasis lesions from SPECT bone scintigrams.
- This approach supports the advancement of deep learning for automated characterization of bone metastasis (BM).
