User Interaction in Semi-Automatic Segmentation of Organs at Risk: a Case Study in Radiotherapy

Area of Science:

• Medical Imaging
• Radiotherapy Planning
• Human-Computer Interaction

Background:

• Accurate organ segmentation is crucial for radiotherapy planning.
• Manual segmentation is time-consuming and variable.
• Automated methods often require post-processing corrections.

Purpose of the Study:

• To evaluate two semi-automatic segmentation methods ('strokes' and 'contour') based on user interaction.
• To analyze the impact of human-computer interaction on segmentation quality and process.
• To provide insights for improving semi-automatic segmentation design.

Main Methods:

• Two physicians performed segmentation on 42 cases across five organs at risk.
• Evaluated subjective and objective measures of interaction process and segmentation quality.
• Correlated various process and result measures for both 'strokes' and 'contour' methods.

Main Results:

• Identified 36 quantifiable and 10 non-quantifiable correlations per interaction type.
• Found strong or moderate correlations in 20 measures for 'contour' and 22 for 'strokes'.
• Demonstrated significant relationships between interaction design and segmentation outcomes.

Conclusions:

• Semi-automatic segmentation requires less cognitively demanding user interactions.
• Interface design must offer flexibility to accommodate physician workflows and preferences.
• Correlated measures offer valuable insights for optimizing user interaction in segmentation tools.