Quantitative CT is quickly gaining relevance and importance in the study of lung disease. Barriers to widespread implementation of QCT in the past have included cumbersome software, inter-scanner variation, and lack of knowledge of the normal range of QCT measurements. With rapid advances in scanning technology and analytic software platforms, the evaluation of lung parenchyma and small airways is becoming an efficient technique for characterization and determining severity of obstructive lung diseases like COPD. The current study has provided normal values for comparing against extents of emphysema and gas trapping in disease cohorts. Normal airway parameters are also presented with analysis of multiple contributing variables. Previous smaller studies have evaluated non-smoking cohorts, but they lack the full volumetric QCT approach utilized here, as well as expiratory QCT evaluation, and airway measurements.
Prior research has produced varied results for %LAAI-950
in non-smokers. Gevenois et al. showed much higher scores for %LAAI-950
in a group of 42 non-smokers, around 8% (13
). This may be related, at least in part, to high inspiratory lung volumes (average TLCCT
about 110% of predicted values). In contrast, Irion et al. identified average %LAAI-950
scores less than 1% in a group of 30 normal subjects (14
). However, their cohort consisted of much younger subjects with an average age of 26 years. The cohort of 185 individuals evaluated by Marsh et al. is quite similar to the one presented here (15
). The proportion of males there was higher, 47.6% compared with 31.5%. However, the relationship of sex with %LAAI-950
is in accordance with the data presented here, showing that men have slightly higher scores than women. The median %LAAI-950
value obtained, 1.4%, was also comparable to the current study at 0.9%. That study, though, opted for evaluation of just three images per subject, as opposed to the volumetric approach utilized here (15
The lack of change in lung attenuation values with increasing age on univariate and multivariate analysis may at first glance appear to contradict previous studies which showed that lung attenuation decreased with age (13
). Seojima et al. performed a longitudinal study evaluating the same non-smokers at baseline and with 5-year follow-up. That study showed up to 1% increase per year of inspiratory %LAA and significant correlation with age. However, that study used a cutoff of −912HU for classifying LAA which would lead to higher estimates (16
). The difference between previous normal studies and our study may also be related to the fact that our study group did not include individuals less than 45 years old. Gevenois et al. had a study cohort age range of 23-71 years, with an average age of 42 years (13
). Marsh et al. had a study cohort age range of 25-75 years, with an average age of 54 years (15
). Our data suggest that CT attenuation remains relatively similar within the age range of 45 to 80.
We also examined the relationship between age and %LAAE-856. On univariate analysis there were no significant findings. Simple linear regression showed no direct correlation between age and %LAAE-856. The lack of correlation held up when we divided the cohort into two age groups. However, upon multivariate analysis, gas trapping measured by %LAAE-856 appeared to increase significantly with age (P<0.001). This remained true even when the model was adjusted for tracheal air (P=0.001). The apparent increase in gas trapping with age may represent a phenomenon of normal aging evident by the lower change in volume from TLC to FRC in the older subjects.
Influence of Scanner Model and Correcting for Tracheal Air
Because it has previously been shown that scanner make and model may result in significant variation in CT attenuation values of lung tissue (18
), we evaluated the role of scanner model and tracheal air attenuation as predictors of %LAAI-950
. The models also included TLCCT
, age, sex and BMI. While %LAAI-950
was significantly different between males and females on univariate analysis, sex was not a significant contributor to any of the multivariate models. The sex difference was most likely accounted for by the relatively higher TLCCT
Our initial multivariate model indicated a significant influence of scanner model for %LAAI-950 and %LAAE-856. When we placed average tracheal air attenuation in the model, scanner model was eliminated as an independent predictor of %LAAI-950, but remained a significant predictor for %LAAE-856. In both models the GE Lightspeed 16 scanner was the main contributor to the variation, but there were several scanners associated with variation in the model for %LAAE-856. For large multi-center studies differing scanner models will always be a substantial source of variation when measuring lung density. It seems that correcting for average tracheal air attenuation may alleviate some of these concerns for %LAAI-950. However, %LAAE-856 may be more sensitive to scanner variations and may not benefit from the tracheal air correction. This might be due to the lower CT dose used for the expiratory scans, with consequent increase in image noise, which may exaggerate inter-scanner differences. Further study should investigate the use of tracheal air attenuation as a possible correction that may be applied when evaluating lung density by CT, as well as possible correction by phantom scanning.
Lobar segmentation offers novel insights into the study of lungs in health and disease. The current study shows distinct regional differences in the normal functioning of healthy adult lungs. The upper lobes show markedly less volume change between inspiration and expiration. This leads to significantly higher %LAAE-856 in the upper lobes when compared with the lower lobes. These differences might be accounted for by the relatively more posterior position of the lower lobes, and also by the effectiveness of diaphragmatic movement in promoting expiratory evacuation of air from the lower lobes. These normal lobar variations should be taken into account when evaluating lung attenuation parameters in disease cohorts. They may be important for calibrating future QCT findings to a range of normal in individual lobes.
seems to be most strongly associated with many airway parameters in non-smokers. This is probably because TLCCT
functions as a composite variable accounting for all other body size variables. In cigarette smokers, women have been shown to have a higher WA% than men even after adjusting for other modifiers (29
), leading to speculation that sex affects the airway response to cigarette smoke. However, since we have shown this relationship exists even in non-smokers, sex, along with age may need to be considered as a covariate in studies of airway diameters. In our study, WA% showed slight decrease with increasing age, while ILA showed a slight increase with age. Airway wall parameters are greatly influenced by CT scanner resolution which needs to be accounted for when evaluating these measures in future cohorts. Our multivariate model showed no strong associations with Pi10, so this may be a good independent measure for evaluating small airway disease.
Some limitations of the current research will need to be studied for their effects on the data presented. As in most multicenter studies, spirometrically gated CT scanning was not feasible, and indeed we were obliged to eliminate 8 subjects who failed to attain an adequate inspiration. With a fixed mAs, different noise levels likely occurred related to patient size. The relatively small number of subjects, though larger than prior cohorts, limits statistical precision. Variations due to image analysis software cannot be determined, but are likely to be small. Further large scale studies in non-smokers would be difficult to achieve, but they may be necessary to fully understand QCT of healthy lungs as it relates to disease cohorts. The predominance of females may have led to some differences in the measured QCT parameters. The relatively narrow age range of the subjects has already been discussed. Also, the low level of recruitment of African-Americans indicates that no conclusions can be drawn regarding racial differences in QCT parameters.
CT scanner variation is a major factor when quantifying parameters of the lungs and airways. The utilization of numerous different types of scanners introduces more error into the measurements. Similarly, the lower mAs used in expiratory scanning in our subjects likely increased the noise and variability in those scans. Scanner manufacturers utilize different algorithms when compiling the raw data from the scans. The assorted reconstruction kernels lead to varying measures of tissue attenuation values, which in turn affect all of the QCT parameters. Further detailed study will need to be performed in order to establish a standard method of correcting for scanner related variation, perhaps based on phantom or tracheal air measurements. Also, the accuracy of the airway wall thickness, and similar variables, is limited by voxel resolution, particularly at the subsegmental level. Because of variations in slice thickness and interval available on different scanner models, we were unable to prescribe uniform voxel dimensions in this study, although submillimeter thickness and interval was achieved in all cases. We have shown that this variability in voxel dimensions affects the measures of airway wall parameters.
Although the attenuation characteristics of normal lung differ by age and gender, these associations do not persist on multivariate analysis. Potential sources of variation in measurement of attenuation-based quantitative CT parameters include depth of inspiration, CT resolution and scanner model. Adjustment for tracheal air attenuation may help alleviate some of the scanner-related variation present in QCT analysis. Since QCT parameters also vary by lobe, understanding of normal lobar values may improve regional QCT assessment of diffuse lung diseases. Sources of variation in QCT airway measurements include age, sex, BMI, depth of inspiration, and voxel resolution. Future studies will need to address these variations in order to gain the most precision from QCT techniques.