Supervised interpretation of echocardiograms with a psychological model of expert supervision

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Accurate segmentation of echocardiograms is crucial for cardiac diagnosis.
• Existing segmentation algorithms lack the precision and contextual understanding of human experts.

Purpose of the Study:

• To develop a collaborative scheme for supervised binary segmentation of echocardiograms.
• To integrate human expertise with algorithmic precision for improved image analysis.

Main Methods:

• A user-friendly system allows experts to interactively correct computer-generated segmentations.
• Image properties and human visual attention maps refine boundary detection.
• The system extracts heart chamber contours from 2D echocardiogram sequences.

Main Results:

• The developed system successfully extracts heart chamber contours.
• The collaborative approach enhances segmentation accuracy by combining human and algorithmic strengths.

Conclusions:

• This human-AI collaborative scheme offers a reliable method for echocardiogram segmentation.
• Future work includes incorporating richer expert inputs and cardiologist knowledge for enhanced classification and reasoning.