PRISM Lite: A lightweight model for interactive 3D placenta segmentation in ultrasound

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare

Background:

• Placenta volume from 3D ultrasound (3DUS) is crucial for monitoring fetal growth and predicting pregnancy outcomes.
• Manual segmentation is accurate but time-consuming and subjective.
• Existing automated deep learning methods lack consistent high-quality results and clinical feasibility due to computational demands.

Purpose of the Study:

• To develop a lightweight, interactive segmentation model for real-time placenta analysis from 3DUS images.
• To enable efficient and accurate placenta segmentation suitable for clinical settings, including low-resource environments.

Main Methods:

• Proposed a novel lightweight interactive segmentation model for placenta.
• Utilized a fully automated model for initial segmentation, incorporating a human-in-the-loop approach for iterative refinement.
• Evaluated performance using Dice score and normalized surface Dice metrics.

Main Results:

• Achieved superior segmentation performance compared to state-of-the-art models.
• Demonstrated significantly fewer parameters and faster inference times.
• Showed robustness to imperfect initial segmentation masks.

Conclusions:

• The proposed lightweight interactive model provides an efficient and accurate solution for real-time placenta segmentation from 3DUS.
• It is well-suited for clinical deployment, overcoming computational limitations of previous methods.