Researchers have developed a novel semi-supervised segmentation method using generative adversarial networks (GANs) to improve the extraction of pulmonary embolism (PE) boundaries from CTPA scans. This approach leverages large datasets of unlabeled scans to enhance accuracy, addressing the challenges of manual interpretation in clinical settings. The method incorporates various loss functions, including dice loss, cross-entropy loss, feature matching loss, and self-training loss, to tackle imbalanced data and generate anatomically plausible segmentation maps. Experimental results demonstrate superior performance compared to fully supervised methods, achieving a Dice coefficient of 88.75% and a Hausdorff distance of 1.74. AI
IMPACT This research could lead to faster and more accurate diagnoses of pulmonary embolism, improving patient outcomes.
RANK_REASON The cluster is a research paper detailing a new method for medical image analysis. [lever_c_demoted from research: ic=1 ai=1.0]
- CTPA scans
- Dice loss
- Dimitris K Iakovidis
- feature matching loss
- generative adversarial network
- logarithmic loss
- pulmonary embolism
- self-training loss
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →