Hybrid Deep Learning-Based Enhanced Occlusion Segmentation in PICU Patient Monitoring
2024

Improving Occlusion Segmentation in Pediatric Intensive Care Units

Sample size: 175 publication 10 minutes Evidence: moderate

Author Information

Author(s): Munoz Mario Francisco, Huy Hoang Vu, Le Thanh-Dung, Jouvet Philippe, Noumeir Rita

Primary Institution: École de Technologie Supérieure

Hypothesis

Can a hybrid deep learning approach effectively segment occlusions in remote patient monitoring for PICU patients?

Conclusion

The proposed hybrid approach significantly enhances the segmentation of occlusions in remote patient monitoring within PICU settings.

Supporting Evidence

  • The proposed method achieved an accuracy of 92.5%, recall of 93.8%, and precision of 90.3%.
  • An average performance gain of 2.75% was observed compared to the baseline CNN-based framework.
  • The study utilized a dataset of 175 high-resolution images from PICU patients.

Takeaway

This study created a smart computer program that helps doctors see through things that might block their view of sick kids in the hospital.

Methodology

A hybrid approach combining Google DeepLabV3+ and Segment Anything Model (SAM) was used for occlusion segmentation, trained on a small dataset from PICU.

Potential Biases

Potential overconfidence in SAM's predictions could lead to missed detections of small objects.

Limitations

The dataset cannot be publicly released due to ethical implications, and the method may struggle with very small or thin occlusions.

Participant Demographics

Pediatric patients in a PICU setting.

Statistical Information

P-Value

p<0.05

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1109/OJEMB.2024.3503499

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication