Table 1 Summary of studies included in the review
Factor | Statistical significance | Clinical relevance | Observations |
|---|---|---|---|
Acquisition parameters | |||
Radiation dose exposure, tube current and tube potential | No consensus | Yes | Despite usually considered as non-significant, there are numerous contradictory study results, with some studies even showing inter-scan variability of volumetry measures in the realm of clinical relevance |
Signal-to-noise ratio (SNR) | No | Not an independent factor | |
Collimation | Yes | Untested | Generally considered as clinically not relevant, but untested |
High-resolution scan mode | Yes | Untested | Single study showing reduced volume overestimation of pulmonary nodules |
Field of view (scan FOV) | No | ||
Pitch | No | Not significant unless using high pitch mode (pitch factor = 3) in small nodules (< 5 mm) | |
Contrast enhancement | Yes | Untested | Overestimates the volume of the pulmonary nodule |
Reconstruction parameters | |||
Slice thickness | Yes | Yes | Thinner slice thickness improves accuracy, precision, and segmentation quality Should be thin enough to allow any nodule to be visible in ≥ 3 consecutive slices A thickness ≥ 2.5 mm is inadequate to detect 1 mm changes in nodule’s diameter |
Field of view (display FOV) | No | ||
Reconstruction interval | No consensus | Untested | Overlap (interval < thickness) improves accuracy and precision of volumetry in smaller nodules and thicker slices Likely not significant using 1 mm slice thickness |
Raw-data reconstruction algorithm | No | Yes (sub-solid nodules) | Iterative reconstruction (IR) algorithms outperform filtered back projection (FBP) for small part-solid nodules and at lower tube currents improving performance of volumetry tools The noise reduction provided by IR is not uniform and less significant at the nodules’ edges |
Kernel | Yes | Yes (sub-solid nodules) | Sharp kernel improves volumetry performance in thin 1 mm slices Smooth kernel outperforms sharp kernel in thicker ≥ 2.5 mm slices |
Post-processing | No | Image compression and vessel suppression considered as not significantly influencing volumetry of pulmonary nodules | |
CT scanner equipment | |||
Vendor | Yes | Untested | Only for small nodules not requiring follow-up |
Technology | No consensus | Untested | Multi-detector CT, flat-panel, dual energy spectral CT |
Software | |||
Software (package and version) and segmentation algorithm | Yes | No | The same software package and version should be consistently used through the follow-up of any pulmonary nodule |
Nodule | |||
Size | Yes | Yes | Performance of volumetry tools is degraded in smaller nodules and considered unreliable for growth estimation of nodules < 5 mm |
Density | Yes | Untested | Volumetry of non-solid nodules has worse accuracy and precision than for solid nodules |
Shape | Yes | Untested | Volumetry of nodules with irregular and spiculated shapes has lower accuracy and precision than volumetry of nodules with round, elongated, smooth or lobulated shapes |
Margin | Yes | Untested | Volumetry of nodules with poorly defined margins have higher variability |
Location | Yes | Untested | Attachment to surrounding structures (e.g., pleura, vessels, bronchial walls) degrades the performance of volumetry tools |
Patient | |||
Parenchymal changes | Yes | Untested | Only with increased attenuation of surrounding parenchyma (e.g., ILD) |
Breathing | Yes | Yes | Breathing artifacts are related to volume overestimation and increased measurement variability |
Cardiopulmonary haemodynamics | Yes | Yes | Complex cardiopulmonary interactions affecting the amount of blood inside or around a nodule, leading to increased volume measurement variability |
Observer | |||
Manual correction | Yes | Untested | Selectively correcting obvious segmentation errors improves the performance of volumetry tools |
Experience | No | ||
Training | Yes | Untested | Training with the volumetry tool is important in unexperienced observers |