RPS 1804 - Technological advancements in chest imaging: MRI, photon counting CT and more

Author Block: D. Kütting, J. A. Luetkens, D. Thomas, A. M. C. Boehner, T. Dell, A. Faron; Bonn/DE
Purpose: Interstitial lung abnormalities (ILA) impact survival and quality of life, yet predictive imaging markers for progression are lacking. MRI holds promise in enhancing ILA phenotyping, potentially enabling tailored follow-up strategies while reducing repetitive CT imaging and associated radiation exposure.
Methods or Background: Assessment of detectability and estimation of disease progression of ILA in a single center, lung cancer screening cohort (224 participants) receiving same day CT/MRI. Radiologists independently evaluated chest images for presence of ILA using standardized criteria. Follow-up exams were reviewed for disease progression. MRI sequences included T2-TSE MVXD, T2-STIR, and diffusion-weighted imaging (DWI). Statistical analyses evaluated the agreement between CT and MRI findings and MRI's diagnostic performance for ILA detection and progression prediction.
Results or Findings: Among the 224 participants (mean age 58.5 ± 5.7 years; 45% female), 26 exhibited ILA on baseline CT, with 65% categorized as subpleural fibrotic. Baseline CT findings served as the reference standard. MRI detected ILA in 30 cases, 20 of which were congruent with CT findings, yielding a sensitivity of 76.9% and a specificity of 94.9% (McNemar's test, p=0.3173). MRI detected ILA in 19/26 cases using T2-TSE MVXD, 20/26 using T2-STIR (7/20 with hyperintense signal), and 6/26 using DWI (3/6 with hyperintense signal). Seven subjects showed progressive disease on follow-up, with 6 of the subjects initially having a subpleural fibrotic pattern. Hyperintense signals in STIR and DWI sequences predicted progression, with hazard ratios of 6.79 and 5.43, respectively. The combination of hyperintense signals in STIR and DWI had a positive predictive value of 100%.
Conclusion: MRI reliably detected ILA and predicted disease progression, particularly in the fibrotic subtype. MRI offered valuable insights for ILA monitoring and phenotyping, potentially improving patient management and reducing radiation exposure.
Limitations: Limited amount of patients
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: University Hospital Bonn

Author Block: Z. M. Xie¹, X. Gao¹, J. Gu¹, Z. Zhang¹, H. Yu², L. Zhu²; ¹shanghai/CN, ²Shanghai/CN
Purpose: To evaluate the clinical value of phase-resolved functional lung imaging(PREFUL)in diagnosing and regional specificity assessment of ventilation and perfusion status of chronic obstructive pulmonary disease (COPD) patients of different severity.
Methods or Background: 100 healthy volunteers, 40 patients with COPD(18 as GOLD1, 10 as GOLD2, 7 as GOLD3, and 5 as GOLD4 )underwent MRI using 3D PREFUL under free breathing at 3.0 T(Free-breathing 1H MRI acquisition, no contrast agent administration). The PREFUL postprocessing method was used for the extraction of dynamic perfusion and ventilation parameters. Mean ventilation and perfusion maps, ventilation flow-volume loops(FVL) correlation, ventilation defect percentage(VDP), perfusion defect percentage (QDP), map of ventilation/perfusion defects (V/Q defects), and matched defect percentage on both perfusion and ventilation maps (VQM) were calculated.
Results or Findings: Compare to the homogenous ventilation and perfusion maps of healthy volunteers, COPD patients showed significant heterogeneity. The mean ventilation and perfusion percentage in COPD patients were significantly lower than the healthy volunteers (P<0.01), while the FVL is statistically higher in COPD patients (P<0.01). The ventilation map showed regional differences in visual agreement with emphysema on CT and all 3D PREFUL-derived ventilation parameters correlated with FEV1 and FEV1/FVC in the patients with COPD(all P<0.05). Besides, our data showed a trend of correlation between different GOLD grades and VDP (P<0.05) in the COPD patients, but no significant difference in QDP among groups. Compared to patients with low-grade COPD (GOLD 1-2), severe COPD (GOLD 3-4) had higher VQM which indicated a better consistency of regional defect in ventilation and perfusion maps(P<0.01).
Conclusion: MRI PREFUL plays a promising role in evaluating the severity of COPD and visually predicting regional ventilation and perfusion defect in 3D lung imaging.
Limitations: The study included a relatively small sample size of COPD patients
Funding for this study: This work was supported in part by the National Natural Science Foundation of China (8207070786), Young Scientists Fund of the National Natural Science Foundation of China (82302188), Shanghai Pujiang Program (22PJD069).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Shanghai Chest Hospital ethics committee

Author Block: A. W. Marka¹, M. Steinhardt¹, M. Graf¹, L. Rahn¹, K. Weiss², M. R. Makowski¹, D. C. Karampinos¹, J. Gawlitza¹, S. Ziegelmayer¹; ¹Munich/DE, ²Hamburg/DE
Purpose: Recent years have seen significant progress in pulmonary MR imaging for lung nodule detection through optimization and new sequences. This study evaluates the capabilities of a 3D gradient echo MRI sequence for detection and classification of pulmonary nodules, specifically in relation to the Lung CT Screening Reporting and Data System (Lung-RADS).
Methods or Background: In this prospective trial, patients with benign and malignant lung nodules admitted between December 2021 and July 2024 underwent low-dose chest CT and pulmonary MRI using a 3D gradient echo sequence, accelerated by parallel imaging, compressed sensing, and deep learning (CSAI). Three radiologists (4, 9, and 10 years of experience), blinded to clinical information, independently evaluated the MR images. Nodule detection, characterization (size, morphology), and Lung-RADS assessment were performed for all patients. To quantify interreader agreement, intraclass correlation coefficient (ICC) for nodule measurements and Cohen’s kappa for Lung-RADS classifications were calculated.
Results or Findings: A total of 75 patients (mean [SD] age, 65±12 years; 33 women [44%]) with 135 pulmonary nodules were included and analyzed. Nominal scan time was 3:53 min. The CSAI sequence achieved a detection rate of 96,3%, with 5 missed nodules all being ≤4mm. The mean nodule diameter for MRI deviated from CT by 0.1 mm (1.96 SD5.87mm; -1.96 SD-5.67mm). Nodule size for CT and MRI showed excellent inter-rater agreement (ICC-CT: 0.995, CI95: 0.993, 0.996; ICC-CSAI: 0.993, CI95: 0.991, 0.995). Lung-RADS category agreement between CT and MRI was almost perfect for Reader 2 (k=0.86) and Reader 3 (k=0.90), while Reader 1 showed substantial agreement (k=0.69).
Conclusion: Pulmonary MRI with an accelerated 3D gradient echo sequence showed high detection rates for pulmonary nodules with comparable Lung-RADs scores and morphological assessments to CT.
Limitations: -Heterogeneous cohort
-No follow-up scans
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: It was approved by the local ethical review board (protocol number 692/21S).

Author Block: Y. Ohno, H. Nagata, T. Ueda, M. Nomura, T. Yoshikawa, D. Takenaka, Y. Ozawa; Toyoake/JP
Purpose: To determine capability of Conjugate gradient reconstruction (CG-recon) and deep learning reconstruction (DLR) for reducing acquisition time with keeping image quality and nodule detection performance on UTE-MRI.
Methods or Background: 35 patients with lung nodule underwent UTE-MRI obtained with CG-recon and grid-reconstruction (Grid-recon) by original (UTEoriginal), 1/2 (UTE1/2) and 1/4 (UTE1/4) sampling spoke numbers at 1.5T and 3T systems. Then, each UTE-MRI was reconstructed with and without DLR. Standard protocol in this study was UTEoriginal obtained by Grid-recon and reconstructed without DLR. In each patient, standard reference for nodule was determined by thin-section CT. To determine the influence of sampling spoke number reduction and reconstruction method differences, signal-to-noise ratios (SNRs) of lung and nodule, overall image quality and nodule presence probability were assessed by ROI measurements or 5-point scales. SNRs and overall image quality were compared between each UTE-MRI and standard protocol by Student’s t-test or Wilcoxon’s signed rank test. Then, ROC analysis was performed to compare nodule detection capability between each UTE-MRI and standard protocol.
Results or Findings: DLR was significantly improved SNRs of all UTEoriginals and UTE1/2 obtained by CG-recon as compared with standard protocol (p<0.05). Overall image qualities of each UTE1/4 and all UTE1/2s except that obtained by CG-recon and reconstructed with DLR were significantly lower than that of standard protocol (p<0.05). Area under the curve (Az) of standard protocol (Az=0.97) was significantly larger than that of all UTE1/4s (0.82Conclusion: CG-recon and DLR can reduce acquisition time without degradation of image quality and nodule detection on UTE-MRI.
Limitations: Limited study population and nodule numbers
Funding for this study: Canon Medical Systems Corporation
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Fujita Health University Hospital

Author Block: G. Agarwal¹, D. Gopalan¹, M. Naik¹, N. Soneji¹, B. Statton¹, B. Ariff¹, M. Tibiletti², G. Parker², S. Copley¹; ¹London/UK, ²Manchester/UK
Purpose: Chronic thromboembolic pulmonary hypertension (CTEPH) is an under-recognized condition associated with significant morbidity yet is surgically treatable. Oxygen-enhanced MRI (OE-MRI) is an emerging tool for quantifying and mapping regional gas delivery and uptake without the need for hyperpolarized gas. This study evaluates the accuracy of visual analysis of OE-MRI for diagnosing CTEPH compared to the current standard of care: V/Q SPECT and MR perfusion (MR-P).
Methods or Background: Prospective study conducted from 2018 to 2023. Participants with a clinical suspicion of CTEPH underwent T1-weighted dynamic OE-MRI on a 1.5T scanner, V/Q SPECT and MR-P. Four consultant radiologists (2 MRI, 2 NM specialists) scored relevant scans independently (positive, negative, or indeterminate) for CTEPH blinded to clinical data with differences resolved by consensus. The reference standard was the multidisciplinary team diagnosis of CTEPH.
Results or Findings: A total of 58 patients were included, with 49 undergoing both OE-MRI and V/Q SPECT, and 48 undergoing all three modalities. Studies considered indeterminate for CTEPH were excluded from sensitivity and specificity analyses (2/49 V/Q, 1/48 MR-P, and 1/49 OE-MRI).

The sensitivity of OE-MRI, V/Q SPECT and MR-P were 0.932 (95% CI 0.78-0.981), 0.964 (95% CI 0.823-0.994) and 0.931 (95% CI 0.78-0.931) respectively. Specificities were 0.789 (95% CI 0.567-0.915), 0.947 (95% CI 0.754-0.991) and 0.833 (95% CI 0.608-0.942) respectively. There was no statistically significant difference between OE-MRI and V/Q or OE-MRI and MR-P using McNemar test (P > .05).
Conclusion: Visual analysis of OE-MRI maps is a valuable adjunct in diagnosing CTEPH with similar sensitivity but lower specificity than VQ SPECT and MR-P. Further analysis of quantitative data is required to fully assess the role of this technique.
Limitations: Limitations are a relatively small sample size (n=58) and only visual (not quantitative) analysis.
Funding for this study: Study funded by NIHR Imperial Biomedical Research Centre (BRC) grant
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Informed consent obtained from all patients for the extra oxygen-enhanced MRI sequence, the other investigations were standard of care.

Author Block: S. S. Schalekamp, L. J. Oostveen, M. Simmelink, W-J. Van Der Woude, P. P. P. Van Der Tol, M. Prokop, E. J. Smit; Nijmegen/NL
Purpose: To assess the image quality of chest scans acquired using a photon-counting CT (PCCT) scanner with cadmium-zinc-telluride (CZT) detectors and super-high resolution deep-learning image reconstruction (SHR-DLR).
Methods or Background: We analyzed the chest images from two consecutive cohorts of 18 and 25 patients who underwent imaging on a prototype PCCT scanner for various indications. Images were reconstructed using both normal resolution (NR: 0.62mm sections, 512-matrix, hybrid-iterative-reconstruction) and super-high resolution (SHR: 0.21mm sections, 1024-matrix, deep-learning-reconstruction) protocols. An experienced chest radiologist assessed image quality using a 5-point scale (poor to excellent) for overall quality, sharpness, detail visibility, noise, and artifacts. A homogeneous region in the left ventricle was used to measure image noise. The number of visible bronchial branching generations was quantified in three lung regions: the upper-right-lobe(1R), the upper-left-lobe(5L), and the right-lower-lobe(10R). Bronchus volumes were automatically calculated (first cohort only). Statistical significance was determined using a signed rank test (p<0.05).
Results or Findings: The SHR-DLR images were rated to have higher overall image quality (4.7 vs 3.6), image sharpness (4.8 vs 3.3), detail visibility (4.7 vs 3.6) compared to the NR images, while having lower perceived image noise (4.3 vs 3.3; all p<0.01). No image artifacts were observed with either protocol (both 4.0). Although measured noise levels were similar between SHR-DLR (33.2 HU) and NR (34.2 HU), SHR-DLR images demonstrated a finer noise texture. SHR-DLR enabled visualization of finer bronchial details, with 1.2 more peripheral branches visible on average (p<0.01). Automatic segmentation showed larger bronchus volumes in the SHR-DLR images (54.1 mL) compared to the NR images (47.2 mL;p<0.01).
Conclusion: PCCT with CZT detectors and SHR-DLR reconstruction provides excellent spatial resolution and superior visualization of the bronchial system while maintaining low image noise.
Limitations: No comparison to conventional CT.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Waived

Author Block: L. Lei, Y. Zhou; Zhengzhou City/CN
Purpose: To explore the feasibility of the virtual non-contrast images derived from Photon-Counting CT (PHCT) substitute for true non-contrast images.
Methods or Background: 40 patients underwent pre-and arterial-venous dual-phase post-contrast chest imagining on a PHCT and had previously undergone a chest CT with a standard energy-integrating detector system (EID-CT) scanner were retrospectively included in this study. The images were retrospectively analyzed. The arterial VNC images (VNC-A) and venous VNC images (VNC-V) were derived from raw datasets using dedicated software respectively. Two radiologists assessed image quality using a five-point Likert scale and performed measurements of vessels and lung parenchyma for signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and in the case of solid lung masses-to-lung parenchyma contrast ratio.
Results or Findings: The image noise of all tissues among the four kinds of images had significant differences, images noise of VNC-A and VNC-V images were lower than TNC images and EID-CT images (P＜0.05), and VNC-V images had the highest SNR and CNR. Good equivalence between VNC and TNC images was observed in all relevant tissues with Bland-Altman analysis. Image quality subjective scoring of EID-CT, TNC, VNC-A, and VNC-V were 5.00(1.00), 5.00(1.00), 5.00(0.75), 5.00(1.00), respectively which had no significant differences (P=0.20).
Conclusion: The VNC image derived from PHCT enhanced image might be used as a substitute for the TNC image, and the image quality is higher than conventional EID-CT images.
Limitations: Not applicable
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Human Research Ethics Committee.

Author Block: S. Sharma¹, S. Ross¹, T. Labno¹, R. Zhang¹, X. Zhan¹, R. Thompson², Z. Yu¹, A. Pourmorteza³; ¹Vernon Hills, IL/US, ²Cleveland, OH/US, ³Atlanta, GA/US
Purpose: To evaluate ultra-high-resolution spectral capabilities of CZT-based photon-counting detector CT (PCD-CT) for quantitative lung imaging.
Methods or Background: A COPDGene2 phantom, with three reference foams (20-lb, 12-lb, and 4-lb with HU-120kVp of -703, -824, and -937) and airways (inner-diameter (ID): 2.5-6 mm, wall-thickness (WT): 0.4-1.5 mm), was scanned on a CZT-based PCD-CT (120 kVp, 0.4x0.5 mm focal-spot, and CTDIvol =12.8 mGy). Scans were reconstructed in normal-resolution (NR) and ultra-high-resolution (UHR) spectral modes (pixel size (PS): 0.125 mm, slice thickness (ST): 0.2 and 0.6 mm, respectively), with a lung kernel (FC52) and iterative denoising, followed by generation of 40-150 keV VMIs. For evaluation, the following were quantified: (1) HU bias between UHR and NR VMIs (ST=3.0 mm), (2) contrast (C) and contrast-to-noise ratios (CNRs) for ground-glass nodules (GGNs) and emphysema (ES) (ST=3.0 mm) (20-lb, 12-lb, and 4-lb foams were surrogates for GGN, normal lung, and ES), and (3) IDs and WTs in VMI with maximum CNR(GGN) and CNR(ES) (ST=0.6 mm).
Results or Findings: Bias between UHR-VMI and NR-VMI was found to be <5 HU for all materials. Noise was greater in UHR-VMI (8.2-22.6%) than NR-VMI. C(GGN) and C(ES) improved in UHR-VMI compared to non-spectral UHR images at <60 keV (max: 4.0 HU) and >60 keV (max: 3.5 HU), respectively. The 70 keV VMI was optimal for both CNR(GGN) and CNR(ES) due to minimal noise amplification. For airways, errors (mean±σ) (mm) in 70 keV UHR-VMI were lower: (UHR/NR) 0.06±0.09/0.15±0.19 (WT) and -0.18±0.17/-0.29±0.31 (ID), with highest improvements for smaller airways.
Conclusion: Spectral-UHR imaging with CZT-based PCD-CT offers diagnostically-relevant contrast improvements between diseased and normal lung over non-spectral images, with reduced measurement errors for airways.
Limitations: Study used a phantom approximating compositional similarity to normal and diseased lung instead of real patients.
Funding for this study: N/A
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: N/A

Author Block: L. G. Murkes, M. Lidegran, M. Sund, P. Hillergren, S. Diaz; Stockholm/SE
Purpose: To introduce ultra low dose photon counting CT (ULDPCCT) as the main diagnostic follow up method in patients with cystic fibrosis (CF) and thereby reduce radiation dose with preserved or improved diagnostic value.
Methods or Background: Patients with CF undergo lifelong yearly follow up alternating CT and chest radiography from an early age. CT is the method of choice and provides important information about the course of the disease, making low and ultralow dose methods imperative.
This prospective study included 71 CF patients between 7 and 66 years of age. A specific study protocol was set up on a photon counting detector CT, Siemens Naeotom Alpha. All patients included underwent an inhaled and exhaled ULDPCCT and low dose PCCT (LDPCCT) examinations at their yearly follow up. Radiation doses were collected for each scan and patient. The median was also calculated. Images from all scans were assessed separately by two paediatric radiologists with different years of experience using a modified Bhalla scoring system. Interobserver agreement was calculated with Cohens’ kappa coefficient. P-values of <0.05 were considered statistically significant.
Results or Findings: The effective dose median (IQR) was 0,11 mSv (0.1-0.13) for the ULDPCCT and 0,77 mSv (0.66-0.87) for the LDPCCT respectively. There was no statistically significant difference between the Bhalla scoring when comparing ULDPCCT versus LDPCCT scans with p-value 0,71 (0,37;1,04). Interobserver agreement was substantial (Kappa value 0.65 for ULDPCCT and 0.71 for LDPCCT)
Conclusion: A tailored ULDPCCT scan protocol might be used as yearly follow up diagnostic tool in CF patients with detailed diagnostic value and reduction of the radiation dose at approximately 1/7th of a regular LDCT, thereby reducing the accumulative dose contribution to the patient.
Limitations: Relatively limited amount of patients.
Funding for this study: Funding was provided by " Riksförbundet Cystisk Fibros " for the statistical analysis
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the swedish ethics commitee. According to the declaration of Helsinkii. Dnr 2023-01227-01

Author Block: B. K. Budai¹, S. Hettinger¹, V. M. Wagner¹, V. Palm¹, R. Hosch², F. Nensa², O. Von Stackelberg¹, H-U. Kauczor¹, J. Biederer¹; ¹Heidelberg/DE, ²Essen/DE
Purpose: This study focused on AI-based CT body composition analysis (BCA) as an alternative to bioelectrical impedance analysis (BIA) for identifying COPD patients at high risk of sarcopenia. We aimed to construct CT-based linear regression models for predicting the patients’ fat mass (FM), fat-free mass (FFM), skeletal muscle mass (SMM), and total muscle mass (MM).
Methods or Background: A total of 571 COPD patients (349 males (61.1%), aged 65.5 ± 8.6y) from a prospective multicentric study (COSYCONET) underwent baseline chest CT scans and BIA. The AI-based BCA of inspiratory chest CTs was performed by the “Body and Organ Analysis” (BOA) algorithm. Volumes of muscles, bones, and fatty tissues were converted to mass in kg using standard human tissue densities. Linear regression with estimated thorax weight fitted to patient weight was used to extract residuals which combined with the respective CT-based measures, age, sex, weight, and height were used for predicting BIA-based results. The reliability of CT-predicted body composition measures was evaluated with intraclass correlation coefficients (ICC). The performance of the CT-based FFMI in identifying high-risk sarcopenia patients was assessed with ROC curve analysis.
Results or Findings: The CT-based estimated body composition measures correlated well with BIA with ICCs of 0.90 (FM), 0.94 (FFM), 0.92 (SMM), and 0.92 (MM). The CT-based FFMI achieved an ICC=0.88 and predicted high-risk sarcopenia patients with an AUC, accuracy, sensitivity, and specificity of 0.903, 88.3%, 88.7%, and 88.2%, respectively. Gwet’s AC1 of 0.82 suggested excellent agreement between the two approaches.
Conclusion: AI-based body composition analysis of chest CT scans could be used to assess BIA-based body composition measures of COPD patients and to identify patients at high risk of sarcopenia.
Limitations: The limitation of the study is that no external validation was performed.
Funding for this study: Funding was provided by the German Federal Ministry of Education and Research (BMBF) (Projektträger: DLR e.V. Bonn, Funding ref. 01GI0884)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the local ethics committee and the central ethics committee of the multicenter study.

Author Block: L. Giuliani, G. M. Masci, N. Landini, P. Giuliani, V. Panebianco, C. Catalano; Rome/IT
Purpose: To investigate whether AI represents an added value for chest X-ray (CXR) interpretation.
Methods or Background: A dataset of CXR performed between March 2023 and January 2024 were retrospectively selected from the institutional PACS by a senior thoracic radiologist. All CXR were evaluated by two young radiologists with 1 year of experience in chest imaging, who assessed the presence of several findings (consolidation, nodule, atelectasis, fibrosis, calcification, pneumothorax, cardiomegaly, pleural effusion, mediastinal enlargement, pneumoperitoneum). All examinations were then analyzed with an AI tool (Lunit INSIGHT CXR, Version 3.110) which assessed the same features. Finally, an additional AI-assisted evaluation was performed by the two radiologists. The ground truth was established by the radiologist in charge of image selection who classified the abnormalities as either visible or not visible on the radiograph.
Results or Findings: A total of 548 CXR examinations were selected. From the 500 CXR analyzed, a total of 876 findings were reported either from the radiologists or from the AI tool. The two radiologists showed a sensitivity/specificity of 80.6%/93.2% and 87.2%/95.3%, respectively. AI showed a sensitivity/specificity of 96.9%/64.3%. With AI assistance, the sensitivity of the two radiologists increased to 82.9% (+2.3%) and 89.3% (+2.1%), while specificity decreased to 87.2% (-6%) and 89.5% (-5.8%). The abnormalities for which the radiologists showed higher disagreement with AI were fibrosis and calcification (p<0.0001), whilst the abnormalities for which the radiologists more often changed interpretation after AI evaluation were nodules, calcification, and pneumothorax (p<0.0001).
Conclusion: AI does not show significant increase of the diagnostic performance compared to standard radiological evaluation of CXR. Particularly, AI slightly increases sensitivity but at cost of a significant decrease in specificity. Therefore, standard radiological interpretation still remains the gold standard for CXR.
Limitations: It was conducted retrospectively
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Sapienza University, Rome, Rif.7226, Prot.0473/2024

Author Block: F. T. Gassert, H. Bast, T. Urban, M. Lochschmidt, L. Kaster, T. Koehler, M. R. Makowski, F. Pfeiffer, D. Pfeiffer; Munich/DE
Purpose: Conventional imaging techniques have limitations in early detection of pneumothorax, particularly for small pneumothoraces. Therefore, the purpose of this study was to evaluate the potential of dark-field chest radiography in improving the detection and assessment of pneumothorax.
Methods or Background: This study included 100 participants, comprising 36 patients with clinically diagnosed pneumothorax and 64 healthy controls. All participants underwent dark-field X-ray chest radiography using a prototype system that simultaneously acquires attenuation-based and dark-field images. Sensitivity, specificity, accuracy, reading time, and diagnostic confidence were compared between attenuation-based radiographs and the combination of attenuation-based radiographs with dark-field images (dark-field overlays).
Results or Findings: Dark-field radiography increased sensitivity for pneumothorax detection from 84.2% (attenuation-based radiographs) to 87.4% (dark-field overlays; p = .26), while specificity remained constant (97.3% vs. 97.4%). The median reading time was significantly reduced from 30.8 seconds to 10.3 seconds (p < .001), and diagnostic confidence improved significantly across all readers (p < .001).
Conclusion: Dark-field chest radiography enhances the detection of pneumothorax, significantly reducing reading time and increasing diagnostic confidence without compromising specificity.
Limitations: Inclusion criteria and the imaging modality they were assessed on were different for pneumothorax patients and controls. While for pneumothorax patients, a conventional radiograph showing a pneumothorax was sufficient, controls had to show a normal CT scan to be included.
Funding for this study: We acknowledge financial support through the European Research Council (AdG 695045), the Center for Advanced Laser Applications (CALA), the Federal Ministry of Education and Research (BMBF) and the Free State of Bavaria under the Excellence Strategy of the Federal Government and the Länder, the German Research Foundation (GRK2274), as well as by the Technical University of Munich–Institute for Advanced Study. This work was carried out with the support of the Karlsruhe Nano Micro Facility (KNMF, KNMF, www.kit.edu/knmf), a Helmholtz Research Infrastructure at Karlsruhe Institute of Technology (KIT).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee of the Technical University of Munich