Congress:
EuroSafe Imaging 2020
Keywords:
Retrospective, Quality assurance, Acute, Structured reporting, Diagnostic procedure, Computer Applications-General, Image manipulation / Reconstruction, CT, Conventional radiography, Lung, Emergency, Abdomen, Action 3 - Image quality assessment based on clinical indications, Not applicable, Performed at one institution
Authors:
P. Rogalla, S. Kandel, S. Carey, J. Kavanagh, A. Kielar, B. Hoppel
DOI:
10.26044/esi2020/ESI-12717
Description of activity and work performed
Materials and Methods
The increased dose efficiency in image reconstruction has permitted dose reduction in CT to the level of X-ray doses, thus allowing us to compare conventional X-ray (CXR) to ultra-low-dose CT imaging (ULDCT) in both the chest and abdomen for emergency room patients presenting with nontraumatic acute pain.
Abdominal Imaging (abdominal tomograms)
- A new protocol at our institution has replaced abdominal x-ray (AXR) in the ED with abdominal ULDCT. 462 patients underwent a CT of the abdomen (135 kV, 20-40 mA weight-based, 0.5 s rotation time, 0.5*80 detector rows, 1.0 mSv reference dose). Images were reconstructed with AIDR (current is standard of care), and Deep learning reconstruction (AiCE). Readers graded image quality on a Likert scale (1=excellent, 10=low quality). Image noise was measured in external and intra-abdominal air and liver.
Chest Imaging (thoracic tomograms)
- Twenty-two patients who had undergone a conventional chest X-ray (2 views) and same-day standard-of-care non-contrast chest-CT (mean DLP 44.1) were included. Scanner specific noise was added to the sinogram to simulate a target dose of 0.18 mSv. Coronal and sagittal isotropic reformats were post-processed by applying a voxel-based, locally normalized weighted-intensity projection to generate 2 cm thick slabs with 1 cm overlap. As in maximum intensity projections (MIP), the voxels with the largest CT numbers in V receive the largest weight. The voxels with the lowest CT numbers also receive relatively higher weight than the middle voxels of volume. ROC analyses were performed, the time for interpretation and reader’s image quality score (1=worst, 10=best) were compared.
Structured reporting
- Structured reporting was used for both abdominal and chest ULDCT scans. Abdomen reports focused on bowel obstruction, constipation and urinary stones, whereas chest report answered the clinical questions of Interstitial Lung Disease, nodules, effusion, heart size, vessels, and pneumothorax. Free air and fractures were noted in both regions (Figure 1)
Results
Abdominal tomograms provide better sensitivity and specificity than conventional x-ray (mean, 2.4 views) for many indications and decrease the mean effective radiation dose, calculated to be 1.6 ± 0.6 mSv and 2.1 ± 2.4 mSv for CXR(mean of 2.3 views).
- Additionally, DLR further decreased the standard deviation in ULDCT from 23.1 HU in the liver to 9.3 HU, which may allow for further dose decrease.
The low sensitivity of CXR compared to standard-dose CT(Aquilion One, Canon Medical) has been previously demonstrated for pneumonia and pulmonary opacities in general [1], pneumothorax [2,3], other indications [4]. The results show that thoracic tomograms can provide higher sensitivity and specificity than CXR for air-space opacities. The dose for ULDCT was 0.18 ± 0.3 mSv and 0.16 ± 0.8 mSv (mean of 2 views) in the chest.
- For focal lung disease (pneumonic infiltrates, nodules 5 mm), the area under the ROC curve (AUC) was significantly higher for thoracic tomograms than CXR (0.803 versus 0.648, respectively, p=0.02).
- An example can be seen in Figure 2, which shows a comparison between a CXR and ULDCT of the chest. The tumour in the ULDCT is much more clearly defined than on the CXR.
- The mean time for interpretation for ULDCT was 36.9s ± 0.857s while CXR is 24.0s ± 5.9s. However, the mean image quality score for ULDCT in chest was 8.2 ± 0.97 while CXR was 7.8 ± 1.6.
A survey confirmed 100% adoption of structured reporting for both abdominal and thoracic tomograms; emergency physicians unanimously prefer the new reports. In three years after implementation, no phonecall was received seeking clarification of the report content.