Background: Chronic obstructive pulmonary disease is caused by small-airway disease and emphysema. Although pulmonary function tests (PFT) measure airflow obstruction, they can't differentiate between airflow limitation and emphysema. Computed Tomography (CT) can be used to identify patients with emphysema. AI-based algorithms are convenient for pattern recognition on chest CT images and emphysema quantification.
Aim of the work: To evaluate an artificial intelligence-based prototype algorithm for quantification of emphysema on chest CT compared with PFT.
Patients and Methods: This cross-sectional study was carried at radiodiagnosis department Ain Shams university hospitals. A total of 35 patients who underwent both chest CT and PFT within 6 months were retrospectively included. The spirometry based Tiffeneau index (TI; which is the ratio of forced expiratory volume in the first second to forced vital capacity) was used to identify emphysema severity; a value of <0.7 was considered to imply airway obstruction. Lung volume analysis was calculated using local artificial intelligence-based 3D reconstruction software and emphysema was quantified using attenuation-based threshold of (-950 HU). Percentage of Low attenuation area (LAA %) was reflected by automated calculation of Goddard score. Emphysema quantification was compared to TI using the using Pearson's method.
Results: The mean TI for all patients was 0.77 ± 0.22. The mean percentages of emphysema (LAA%) 20.54% ± 21.8%. AI-based emphysema quantification showed good correlation with TI (p < 0.001). Conclusion: AI-based, automated emphysema quantification either with Goddard score or LAA % shows good correlation with TI, possibly contributing to an image-based diagnosis, COPD categorization, follow-up, and treatment strategies planning.