The influence of initial outlines on manual segmentation

Area of Science:

• Medical Imaging Analysis
• Radiology
• Computational Pathology

Background:

• Manual segmentation of medical images is time-consuming.
• Initial outlines can potentially reduce segmentation time.
• The impact of initial outlines on segmentation precision is not fully understood.

Purpose of the Study:

• To evaluate the influence of initial outlines on manual segmentation precision.
• To assess both intraobserver and interobserver variability in segmentation tasks.
• To quantify the effect of initial outlines on the accuracy of tumor boundary delineation.

Main Methods:

• Three observers independently segmented pleural mesothelioma tumors in CT scans.
• Observers re-segmented tumors using their own or others' initial outlines.
• Jaccard similarity coefficient (J) and linear mixed effects models were used to assess precision.
• The percentage of minor alterations to initial outlines was recorded.

Main Results:

• Interobserver precision significantly increased when using initial outlines (average J=0.796 vs. 0.371).
• Initial outlines strongly influenced intraobserver precision (average J difference=0.476).
• Observers made minor alterations to 74.5% of initial outlines.

Conclusions:

• Intraobserver and interobserver precision in medical image segmentation are highly dependent on initial outlines.
• The tendency for minor alterations suggests a strong anchoring effect of initial outlines.
• Findings impact study design, tumor assessment, AI validation, and radiotherapy planning.