Emphasizing the limited clinical adoption of expiratory computed tomography (CT), challenges with breath holds in older patients with impaired lung function and the increased radiation exposure with multi-phase CT, researchers suggest that an emerging convolutional neural network (CNN) may enable single-phase CT detection and staging of chronic obstructive pulmonary disease (COPD).
For the retrospective study, recently published in Radiology: Cardiothoracic Imaging, researchers reviewed data from 8,893 participants (mean age of 59.6) to assess the capability of a CNN to predict spirometry measurements based on single-phase CT and subsequent comparison to multiphase CT for COPD staging.
The study authors found that the CNN model offered within-one stage accuracy rates of 83.5 percent for single-phase inspiratory CT and 84.1 percent for single-phase expiratory CT, rates that were comparable to the 86.3 percent within-one stage accuracy reported with inspiratory/expiratory CT.
“Our results suggest COPD diagnosis and staging using a single standard inspiratory image is feasible when using a CNN, which may increase accessibility to patients seeking treatment at institutions where an inspiratory-expiratory imaging protocol is unavailable. Moreover, this system can be applied to databases of inspiratory images (acquired for other clinical indications) for screening purposes, where at-risk patients flagged for COPD can be recommended for further evaluation and a formal definitive diagnosis,” wrote lead study author Amanda N. Lee, B.S., who is affiliated with the Computational Science Research Center and the Department of Mathematics and Statistics at San Diego State University in California, and colleagues.
Researchers also found that inclusion of clinical data in the CNN model led to 5.2 percent, 5.9 percent, and 4 percent increases in accuracy for predicting Global Initiative for Chronic Obstructive Lung Disease (GOLD) staging for single-phase inspiratory CT, single-phase expiratory CT, and inspiratory/expiratory CT respectively.
Three Key Takeaways
1. Feasibility of single-phase CT for COPD staging. The convolutional neural network (CNN) demonstrated comparable accuracy in COPD staging using single-phase inspiratory CT (83.5 percent) or expiratory CT (84.1 percent) compared to multiphase inspiratory/expiratory CT (86.3 percent), which could significantly reduce the need for additional imaging protocols.
2. Reduced radiation exposure. The ability to accurately stage COPD using a single inspiratory CT image minimizes the need for multiphase imaging, thereby reducing cumulative radiation exposure, especially critical for older patients with impaired lung function undergoing long-term evaluations.
3. Potential for broader accessibility. Single-phase inspiratory CT combined with CNN offers a practical alternative for facilities without access to inspiratory/expiratory imaging protocols, allowing screening and staging of COPD with images that may initially be acquired for other clinical purposes.
Noting that the inspiratory/expiratory CT protocol is not available at many facilities and the potential for cumulative radiation exposure in ongoing screening for older patients with impaired lung function, the study authors said the ability to rely on single-phase inspiratory CT could be a significant advance.
“One major benefit of our findings is the potential to accurately stage COPD without the need for a separate expiratory acquisition, thereby reducing radiation dose,” emphasized Lee and colleagues. “… It is broadly agreed that radiation exposure from CT cannot be ignored in the long-term evaluation of incurable diseases. Thus, it is a notable result that CNN-based staging using a single CT image is comparable to staging with both inspiratory and expiratory images.
(Editor’s note: For related content, see “FDA Clears Updated AI Software for Lung CT,” “Computed Tomography Study Finds Nearly 44 Percent of Interstitial Lung Abnormalities Are Not Reported” and “FDA Clears Adjunctive Lung Ventilation Software for CT.”)
In regard to study limitations, the authors conceded fluctuation and variability with spirometry assessments. They maintained that standardization of spirometric technique and CT imaging can affect the prognostic capability of CNNs for ascertaining COPD risk. Noting that over 68 percent of the cohort was comprised of non-Hispanic White participants, the researchers said further research is needed to determine whether the study findings are applicable to other racial and ethnic populations.
