Purpose
The most common method for emphysema quantification on computed tomography (CT) is segmenting the lungs and calculating the percentage of voxels with a density below a certain threshold.[Müller 1987, Lowe 2019] High image noise will therefore impact the emphysema quantification. Noise may cause a voxel in a low density region to be above the threshold, causing underestimation, but it can simultaneously cause voxels in a high density region to be below the threshold, causing overestimation of emphysema severity.
Noise can be defined as the standard...
Methods and materials
We used ultra-low dose CT scans from a cohort of 50 COPD patients. Scanning was performed with third generation dual-source CT (SOMATOM Force, Siemens Healthineers) at 70ref-mAs, and 100kVp with Sn-filter (median CTDIvol 0.38mGy). Images were reconstructed with slice thickness/increment 1.0/0.7mm, medium-smooth kernel, and filtered backprojection.
The trachea was segmented in a 61x61x61 voxel region, centred 1.5cm above the carina. To prevent an overestimation due to the partial volume effect, a morphological erosion was used to remove a 2mm layer. Every segmentation was visually checked,...
Results
The true image noise based on the full trachea segmentation was 26.7-38.3HU. For ROI the range was 21.5-44.2HU, for VOI 26.0-40.2HU. Distance between limits of agreement decreased from 11.1 to 4.1HU when switching from ROI to VOI. Systematic bias increased from 0.3 to 0.6HU. Switching from ROI to VOI did not require more clinical analysis time.See Figure2 for the results of the Bland-Altman analysis.
Conclusion
Measuring image noise with a VOI (a sphere) instead of an ROI (a circle) yields a reduction in variability of 64% for ultra-low dose CT.
Personal information and conflict of interest
H. J. Wisselink; Groningen/NL - nothing to disclose
G. J. Pelgrim; Groningen/NL - nothing to disclose
M. Rook; Groningen/NL - nothing to disclose
M. P. van den Berge; Groningen/NL - nothing to disclose
R. Vliegenthart; Groningen/NL - nothing to disclose
G. H. de Bock; Groningen/NL - nothing to disclose
References
Müller 1988 – Müller, Staples, Miller et al., "“Density mask”: an objective method to quantitate emphysema using computed tomography", Chest, DOI: 10.1378/chest.94.4.782
Lowe 2019 – Lowe, Regan, Anzueto et al., "COPDGene 2019: Redefining the Diagnosis of Chronic Obstructive Pulmonary Disease.", Chronic Obstructive Pulmonary Diseases: Journal of the COPD Foundation, DOI: 10.15326/jcopdf.6.5.2019.0149
Messerli 2017 – Messerli, Ottilinger, Warschkow et al., "Emphysema quantification and lung volumetry in chest X-ray equivalent ultralow dose CT-Intra-individual comparison with standard dose CT", European Journal of Radiology, DOI: 10.1016/j.ejrad.2017.03.003