Related Experiment Video
Updated: Jan 15, 2026
06:15
Protocol and Guidelines for Point-of-Care Lung Ultrasound in Diagnosing Neonatal Pulmonary Diseases Based on International Expert Consensus
Published on: March 6, 2019
51.9K
B-Line Detection and Localization in Lung Ultrasound Videos Using Spatiotemporal Attention.
Hamideh Kerdegari1, Nhat Tran Huy Phung1,2, Angela McBride2
1School of Biomedical Engineering & Imaging Sciences, King's College London, London SE1 7EU, UK.
Applied Sciences (Basel, Switzerland)
|October 8, 2025
Summary
This study introduces a spatiotemporal attention mechanism for lung ultrasound (LUS) videos to detect B-line artifacts in dengue patients. The AI model effectively identifies and localizes these lung abnormalities, improving diagnostic accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Pulmonology
Background:
- B-line artifacts in lung ultrasound (LUS) imaging are key indicators of lung abnormalities, particularly in dengue patients.
- Assessing these artifacts is crucial for diagnosis, especially in resource-limited settings.
Purpose of the Study:
- To develop and evaluate a spatiotemporal attention mechanism for automated B-line detection and localization in LUS videos.
- To improve the efficiency and accuracy of diagnosing lung abnormalities using LUS imaging.
Main Methods:
- A novel spatiotemporal attention model was designed, incorporating spatial attention for image focus and temporal attention for frame relevance.
- The model was trained in a weakly-supervised manner on LUS videos from severe dengue patients.
- Performance was evaluated based on B-line detection rates, spatial localization accuracy (IoU), and temporal localization accuracy (correlation coefficient).
Main Results:
- The model achieved an F1 score of 83.2% for classifying B-line positive videos.
- It demonstrated strong performance in localizing salient B-line regions spatially (IoU 69.7%) and temporally (correlation coefficient 0.67).
- The approach successfully identified videos with B-lines and pinpointed abnormal regions and frames.
Conclusions:
- The proposed spatiotemporal attention mechanism is effective for B-line detection and localization in LUS videos, even with weak supervision.
- This AI-driven approach shows significant potential for aiding in the diagnosis of lung abnormalities in dengue patients in resource-limited environments.
