Simulated low-dose dark-field radiography for detection of COVID-19 pneumonia

• Published in: PloS one 2024-12, Vol.19 (12), p.e0316104
• Main Authors: Schick, Rafael C, Bast, Henriette, Frank, Manuela, Urban, Theresa, Koehler, Thomas, Gassert, Florian T, Sauter, Andreas P, Renger, Bernhard, Fingerle, Alexander A, Karrer, Alexandra, Makowski, Marcus R, Pfeiffer, Daniela, Pfeiffer, Franz
• Format: Article
• Language: English
• Subjects: Aged; Algorithms; Bacterial pneumonia; Betacoronavirus; Chest; Computed tomography; Consent; Coronavirus Infections - diagnosis; Coronavirus Infections - diagnostic imaging; Coronaviruses; COVID-19; COVID-19 - diagnostic imaging; Data systems; Diagnosis; Emphysema; Female; Humans; Image processing; Image reconstruction; Lung - diagnostic imaging; Lung diseases; Male; Medical imaging; Methods; Middle Aged; Pandemics; Performance evaluation; Pneumonia; Pneumonia, Viral - diagnostic imaging; Pneumothorax; Prospective Studies; Radiation; Radiation Dosage; Radiographs; Radiography; Radiography, Thoracic - methods; ROC Curve; SARS-CoV-2; Scanners; Severe acute respiratory syndrome coronavirus 2; Tomography, X-Ray Computed - methods; X-rays
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0316104

Dark-field radiography has been proven to be a promising tool for the assessment of various lung diseases. To evaluate the potential of dose reduction in dark-field chest radiography for the detection of the Coronavirus SARS-CoV-2 (COVID-19) pneumonia. Patients aged at least 18 years with a medically indicated chest computed tomography scan (CT scan) were screened for participation in a prospective study between October 2018 and December 2020. Patients were included if they had a CO-RADS (COVID-19 Reporting and Data System) score ≥ 4 (COVID-19 group) or if they had no pathologic lung changes (controls). A total of 89 participants with a median age of 60 years (interquartile range 48 to 68 yrs.) were included in this study. Dark-field and attenuation-based radiographs were simultaneously obtained by using a prototype system for dark-field radiography. By modifying the image reconstruction algorithm, low-dose radiographs were simulated based on real participant images. The simulated radiographs corresponded to 50%, 25%, and 13% of the full dose (41.9 μSv, median value). Four experienced radiologists served as blinded readers assessing both image modalities, displayed side by side in random order. The presence of COVID-19-associated lung changes was rated on a scale from 1 to 6. The readers' diagnostic performance was evaluated by analyzing the area under the receiver operating characteristic curves (AUC) using Obuchowski's method. Also, the dark-field images were analyzed quantitatively by comparing the dark-field coefficients within and between the COVID-19 and the control group. The readers' diagnostic performance in the image evaluation, as described by the AUC value (where a value of 1 corresponds to perfect diagnostic accuracy), did not differ significantly between the full dose images (AUC = 0.86) and the simulated images at 50% (AUC = 0.86) and 25% of the full dose(AUC = 0.84) (p>0.050), but was slightly lower at 13% dose (AUC = 0.82) (p = 0.038). For all four radiation dose levels, the median dark-field coefficients within groups were identical but different significantly by 15% between the controls and the COVID-19 pneumonia group (p