Sign in Subscribe
Machine learning

Deep Learning Models for Accurate B-Line Detection and Localization in Lung Ultrasound Imaging

Strategic Insights

— 1 min read

Researchers from Makerere University in Kampala, Uganda, have developed two deep learning models, YOLOv5-PBB and YOLOv8-PBB, for accurate B-line detection and localization in lung ultrasound imaging. These models are designed to address the challenges of interpreting lung ultrasound images, which are often subject to observer variability and require significant expertise. The models were trained on a diverse dataset and showed high performance in precision, recall, and mean average precision, making them ideal candidates for mobile deployment in resource-limited settings.

Key Takeaways:

  • The researchers developed two deep learning models, YOLOv5-PBB and YOLOv8-PBB, for accurate B-line detection and localization in lung ultrasound imaging.
  • The models were trained on a diverse dataset from a publicly available repository and Ugandan health facilities.
  • Experimental results showed that YOLOv8-PBB achieved the highest precision (0.947), recall (0.926), and mean average precision (0.957).
  • YOLOv5-PBB, while slightly lower in performance, had advantages in model size (14 MB vs. 21 MB) and average inference time (33.1 ms vs. 47.7 ms), making it more suitable for real-time applications in low-resource settings.
  • The integration of these models into a mobile LUS screening tool provides a promising solution for B-line localization in resource-limited settings.
  • The researchers concluded that the YOLOv5-PBB and YOLOv8-PBB models offer high performance while addressing challenges related to inference speed and model size.

Statistics:

  • YOLOv8-PBB achieved precision of 0.947, recall of 0.926, and mean average precision of 0.957.
  • YOLOv5-PBB achieved precision of 0.931, recall of 0.918, and mean average precision of 0.936.
  • YOLOv8-PBB had a model size of 21 MB, while YOLOv5-PBB had a model size of 14 MB.
  • YOLOv8-PBB had an average inference time of 47.7 ms, while YOLOv5-PBB had an average inference time of 33.1 ms.

Sources:

  • Deep learning for accurate B-line detection and localization in lung ultrasound imaging. Frontiers in Artificial Intelligence, 2025;8:1560523.
  • Makerere University Reports Findings in Artificial Intelligence (Deep learning for accurate B-line detection and localization in lung ultrasound imaging). Robotics & Machine Learning. May 19, 2025; p 377.

Read more