Robustness of ¹⁸F-FDG PET Radiomic Features in Lung Cancer: Impact of Advanced Reconstruction Algorithm.

Area of Science:

• Nuclear Medicine
• Radiomics
• Oncology

Background:

• ¹⁸F-FDG PET radiomics shows promise for quantifying lung cancer intratumor heterogeneity.
• Robustness of radiomic features (RFs) is affected by image reconstruction, segmentation, and discretization.
• The specific impact of the advanced block sequential regularized expectation maximization (BSREM) algorithm on RFs is not well understood.

Purpose of the Study:

• To investigate variations in ¹⁸F-FDG PET radiomic features using the BSREM reconstruction algorithm.
• To compare the robustness of RFs under different reconstruction, segmentation, and discretization methods.

Main Methods:

• Retrospective analysis of ¹⁸F-FDG PET data from 120 lung cancer patients.
• Reconstruction using advanced BSREM and conventional ordered-subset expectation maximization (OSEM).
• Application of three segmentation methods (manual, 40% threshold, Nestle) and two discretization methods (absolute, relative).
• Assessment of RF stability and robustness using coefficient of variation and intraclass correlation coefficient.

Main Results:

• High instability (CV > 20%) observed in 19% (reconstruction), 33% (segmentation), and 36% (discretization) of RFs.
• Significant robustness (ICC > 0.90) found in 60% (reconstruction), 19% (segmentation), and 35% (discretization) of RFs.
• BSREM demonstrated significantly greater robustness compared to OSEM across varying segmentation and discretization conditions (P < 0.05).

Conclusions:

• The BSREM reconstruction method enhances the stability and robustness of ¹⁸F-FDG PET radiomic features.
• Advanced reconstruction techniques like BSREM can provide more consistent radiomic analysis.
• Improved consistency may enhance the diagnostic and prognostic value of PET-based radiomics in lung cancer.