RPS 1704 - The dark side of chest imaging

Purpose:

To investigate chest radiograph (CXR) reporting by radiographers.

Methods and materials:

A prospective, single-site trial. 12-month consecutive CXRs referred from primary care were independently reported by consultant radiologists (CR;n=13) and reporting radiographers (RR;n=3). Respiratory physicians, blinded to the reporter, compared reports for agreement. Discordant cases were reviewed by thoracic radiologists, blinded to the reporter, who reached the index diagnosis and graded report elements (observation, interpretation, recommendations, and usefulness). The number of CR and RR generated CT scans and lung cancers diagnosed were recorded.

Results:

8,685 of 9,136 (95.1%) CXRs were included. Agreement and insignificant disagreement between CR/RR reports occurred in 5,981 (68.9%) and 1,347 (15.5%) of cases, respectively. 1,357 (15.6%) of CR/RR reports had clinically significant disagreement. Thoracic radiology review has been performed for 908 of 1,357 (66.9%) discordant reports. Both reports were correct in 292 (32.2%), CR report correct in 255 (28.1%), RR report correct in 271 (29.8%), and neither report correct in 90 (9.9%). Thoracic radiologists were no more likely to agree with a CR or RR report (p=0.49;CI=-0.03,0.07). RR reports were non-inferior to CR reports for observation (764 vs 768;p=0.787,CI=-0.037-0.029), interpretation (758 vs 763;p=0.742,CI=-0.039-0.028), and recommendations (692 vs 656;p=0.06,CI=-0.003-0.082), and not clinically different for usefulness (870 vs 832;p<0.001,CI=0.02-0.063). 350 CT scans were generated by CRs or RRs; 103 both CR/RR, 149 CR only, and 98 RR only. 31 of 49 lung cancers were diagnosed on a radiology generated CT; n=22 both CR/RR, n=5 CR only, and n=4 RR only (CR PPV=10.7%;RR PPV=12.9%).

Conclusion:

CXR reporting by RR appears to be comparable to CR, with similar accuracy and use of further tests.

Limitations:

A single-site study with a small number of observers, mitigated by prospective data collection and a large consecutive clinical series.

Ethics committee approval

NHS-HRA:17/LO/0870.

Funding:

Cancer Research UK EDAG 2016.

Purpose:

Commonly in the developing world, photographs of medical images are sent through WhatsApp to radiologists for reporting in emergency situations with limited availability of PACS and trained radiologists. We studied the risks associated with such practice by replicating this process in a multi-reader chest x-ray (CXR) study.

Methods and materials:

200 CXRs from 4 outpatient imaging centres were pulled from PACS and anonymised and read by 3 radiologists; R1, R2, and R3 having 32-, 15-, and 6-years' experience, respectively. CXRs were read in two phases. In Phase-1 (P-1), photographs of CXRs, viewed on a non-medical screen, were taken using a smartphone (RedMi-Y1, 13MP camera) and sent to radiologists through WhatsApp who also read the images on smartphones (OnePlus-5T, iPhone 6, and RedMi-4). In Phase-2 (P-2), after a 2-week washout period, the radiologists reported the same CXRs on medical-grade monitors (27 inches, NEC). The findings in the CXRs were recorded under the labels: atelectasis, consolidation, infiltration, pneumothorax, oedema, emphysema, fibrosis, pleural thickening, effusion, cardiomegaly, nodule/mass, and normal study. Sensitivity and specificity were calculated for each label from P-1 for individual radiologists using their own findings and from P-2 as ground truth.

Results:

Average sensitivity and specificity across all findings were 80% and 97%. The highest average sensitivity was for pleural effusion (84%) and lowest for pleural thickening (15%). The highest average specificity was for pneumothorax and oedema (99% for both), and the lowest for infiltration (94%). Pneumothorax, pleural effusion, consolidation, and mass/nodules were missed in 8, 6, 1, and 2 cases by R1; 3, 14, 3, and 8 cases by R2; and 2, 2, 3 and 12 cases by R3.

Conclusion:

Reading CXRs on mobile phones can lead to significant errors.

Limitations:

A small sample size.

Ethics committee approval

n/a

Funding:

No funding was received for this work.

Purpose:

To assess the dynamics of lung ventilation using xenon-enhanced dual-energy CT (Xe-DECT) of an animal in a normal and bronchoconstricted state.

Methods and materials:

Six anesthetised New Zealand white rabbits who received Xe-DECT scans using xenon gas as a contrast agent with a concentration of 70% Xe-30% oxygen, administered by mechanical ventilation via tracheostomy. Dynamic DECT scans (Revolution CT, GE Healthcare) were obtained over multiple breathing cycles during a Xe gas wash-in (WI) and wash-out (WO) sequence by performing 15 repetitive acquisitions with a 1,01s interscan delay. In addition, bronchoconstriction was induced by inhaled methacholine (40 mg/ml) and all scan sequences were repeated. Material decomposition was applied to isolate Xe from the image data and parametric ventilation maps were created to quantify Xe concentration in the segmented lungs. We compared the evolution of the regional Xe concentration (mg Xe/mL) in a normal and bronchoconstricted state.

Results:

In a normal state, the maximum Xe concentration in the lungs was 1.60±0.45 mg/mL, which was reached after 15.6±6.9 s. During WO, a 50% decrease in Xe concentration was observed after 12.1±1.3s. The parametric maps clearly showed the increase and decrease of Xe concentration during the WI and WO sequences. After induction of bronchoconstriction, local ventilation defects could be visually detected. Animals showed a reduced Xe concentration in affected areas (1.27±0.68 mg/mL) compared to a normal state (1.91±0.65 mg/mL), and also higher concentration in non-affected areas (2.26±0.71 mg/mL) compared to a normal state.

Conclusion:

This animal study suggests that Xe-DECT can be used to acquire spatial and dynamic data of lung ventilation as well as to detect hypoventilated regions due to bronchoconstriction.

Limitations:

Only six animals were used in this study.

Ethics committee approval

Animal board approval was obtained.

Funding:

Work supported by UZ Brussel department of Radiology.

Purpose:

To evaluate the feasibility of non-contrast-enhanced 3D-UTE MRI for pulmonary imaging in immunocompromised patients during haematopoietic stem cell transplantation (HSCT).

Methods and materials:

28 patients with an indication for allogeneic HSCT were included in this prospective single-centre study and underwent thoracic MRI before HSCT initiation, in case of suspected pneumonia, and before discharge following completion of HSCT. MRI was acquired using a prototypical stack-of-spirals 3D-UTE sequence within a single breath-hold. Clinically indicated MDCT and a clinically used T2 weighted sequence were used as diagnostic reference. 3D-UTE MR and T2 weighted MR image sets were separately reviewed by two radiologists regarding presence of pleural effusions (PE), ground-glass opacities (GGO), and consolidations. Each MDCT used as reference for MRI results was assessed regarding the same items in consensus reading.

Results:

A total of 62 MRI scans were acquired without periprocedural complications. 3D-UTE MRI allowed for sufficient imaging of pulmonary consolidations (sensitivity 78%, specificity 82%, PPV 84%, and NPV 76%). Regarding PE (sensitivity 54%, specificity 99%, PPV 92%, and NPV 89%) and GGO (sensitivity 49%, specificity 87%, PPV 45%, and NPV 88%), the diagnostic performance was lower. Overall consistency rate was high (79-90%). Inter-rater agreement was moderate to substantial (PE=0.62, GGO=0.57, consolidations=0.70). 3D-UTE was equivalent (54%) or superior (43%) when compared to T2 weighted MRI in detecting ground-glass opacity areas and consolidations.

Conclusion:

With an acquisition time similar to MDCT, radiation-free and contrast-free 3D-UTE MRI can be expected to play a major role in future pulmonary imaging, for example, in diagnostics and follow-up of immunocompromised patients with pneumonia.

Limitations:

n/a

Ethics committee approval

n/a

Funding:

The project was funded by Deutsche Forschungsgemeinschaft. The Department receives a research grant from Siemens Healthcare GmbH, which is not specifically directed towards authors.