Rapid Contour-based Segmentation for ¹⁸F-FDG PET Imaging of Lung Tumors by Using ITK-SNAP: Comparison to Expert-based...

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Accurate segmentation of lung tumors in fluorodeoxyglucose (FDG) positron emission tomography (PET) is crucial for treatment planning and response assessment.
• Complex tumor shapes present challenges for automated segmentation methods.
• Current standard methods, like thresholding, often lack precision.

Purpose of the Study:

• To evaluate the performance of ITK-SNAP software for segmenting complex-shaped lung tumors in FDG PET images.
• To compare ITK-SNAP's accuracy against an expert-based manual segmentation reference standard.
• To assess ITK-SNAP's reliability and compare it with established PET tumor delineation techniques.

Main Methods:

• Retrospective analysis of 76 FDG PET scans of thoracic lesions.
• Segmentation using ITK-SNAP's active contour algorithm by four raters, each performing segmentation twice.
• Establishment of an optimized reference standard using simultaneous truth and performance level estimation (STAPLE) algorithm with six expert raters.
• Assessment of accuracy using Dice coefficient and Hausdorff distance.
• Evaluation of interrater and intrarater reliability using intraclass correlation coefficients (ICCs).
• Comparison with thresholding methods (VOI41 and VOI50).

Main Results:

• ITK-SNAP demonstrated high accuracy with a Dice coefficient of 0.83 and a Hausdorff distance of 12.6 mm.
• Excellent interrater reliability (ICC = 0.94) and intrarater reliability (ICC > 0.97) were observed.
• Traditional thresholding methods (VOI41, VOI50) showed significantly lower accuracy (Dice coefficients of 0.48 and 0.34, respectively).
• ITK-SNAP outperformed VOI41 and VOI50 in both Dice coefficient and Hausdorff distance metrics.

Conclusions:

• ITK-SNAP software provides accurate and reliable segmentation for heterogeneous thoracic PET tumors using active contours.
• The software surpasses currently recommended threshold-based PET tumor delineation methods.
• ITK-SNAP offers a robust tool for precise tumor segmentation in clinical oncology.