Abstract Background Dynamic chest radiography (DCR) is a recently developed functional x‐ray imaging technique that detects pulmonary ventilation impairment as a decrease in changes in lung density during respiration. However, the diagnostic performance of DCR is uncertain owing to an insufficient number of clinical cases. One solution is virtual imaging trials (VITs), which is an emerging alternative method for efficiently evaluating medical imaging technology via computer simulation techniques. Purpose This study aimed to estimate the typical threshold thickness of residual normal tissue below which the presence of emphysema may be detected by DCR via VITs using virtual patients with different physiques and a user‐defined ground truth. Methods Twenty extended cardiac‐torso (XCAT) phantoms that exhibited changes in lung density during respiration were generated to simulate virtual patients. To simulate a locally collapsed lung, an air sphere was inserted into each lung regions in the phantom. The XCAT phantom was virtually projected using an x‐ray simulator. The respiratory changes in pixel value (ΔPV) were measured on the projected air spheres (simulated lesions) to calculate the percentage of decrease (ΔPV%) relative to ΔPV exp‐ins in the absence of an air sphere. The relationship between the amount of residual normal tissue and ΔPV% was fitted to a cubic approximation curve (hereafter, performance curve), and the threshold at which the ΔPV% began to decrease (normal‐tissue thre ) was determined. The goodness of fit for each performance curve was evaluated according to the coefficient of determination ( R 2 ) and the 95% confidence interval derived from the standard errors between the measured and theoretical values corresponding to each performance curve. The ΔPV% was also visualized as a color scaling to validate the results of the VITs in both virtual and clinical patients. Results For each lung region in all body sizes, the ΔPV% decreased as the amount of residual normal tissue decreased and could be defined as a function of the amount of residual normal tissue in front of and behind the simulated lesions with high R 2 values. Meanwhile, the difference between the measured and theoretical values corresponding to each performance curve was only partially included in the 95% confidence interval. The normal‐tissue thre values were 146.0, 179.5, and 170.9 mm for the upper, middle, and lower lungs, respectively, which were demonstrated in virtual patients and one real patient, where the value of the residual normal tissue was less than that of normal‐tissue thre ; any reduction in the residual normal tissue was reflected as a reduced ΔPV and depicted as a reduced color intensity. Conclusions The performance of DCR‐based pulmonary impairment assessment depends on the amount of residual normal tissue in front of and behind the lesion rather than on the lesion size. The performance curve can be defined as a function of the amount of residual normal tissue in each lung region with a specific threshold of normal tissue remaining where lesions become detectable, shown as a decrease in ΔPV. The results of VITs are expected to accelerate future clinical trials for DCR‐based pulmonary function assessment.

Estimation of threshold thickness of residual normal tissue in lung dysfunction detectable by dynamic chest radiography: A virtual imaging trial