Virtual non-contrast images calculated from dual-energy CT shoulder arthrography improve the detection of intraarticular loose bodies

Christoph Stern, Dimitri N. Graf, Samy Bouaicha, Karl Wieser, Andrea B. Rosskopf, Reto SutterSkeletal 2022 · DOI

TL;DR

This study demonstrates that virtual non-contrast (VNC) images, generated from dual-energy CT (DECT) shoulder arthrography, significantly improve the detection of intraarticular loose bodies compared to standard DECT arthrography. This technique also enhances diagnostic confidence for periosteal calcifications, offering a valuable tool for shoulder imaging without additional radiation.

What Problem Does This Paper Address?

Shoulder injuries, such as those resulting from dislocations or osteoarthritis, can lead to the formation of intraarticular loose bodies (fragments of bone or cartilage within the joint). Detecting these fragments is crucial for accurate diagnosis and treatment planning. Conventional CT arthrography, which involves injecting contrast material into the joint, can obscure these small, often calcified, loose bodies due to the dense contrast. While dual-energy CT (DECT) offers the ability to differentiate materials, its application in shoulder arthrography for improved loose body detection using virtual non-contrast (VNC) imaging has not been extensively evaluated. This paper aims to fill that gap by assessing the image quality and diagnostic performance of VNC images derived from DECT shoulder arthrography for detecting periosteal calcifications and intraarticular loose bodies.

Methods

The researchers utilized a retrospective approach, analyzing DECT shoulder arthrography scans from 123 patients (129 shoulders). The DECT scans were performed using an 80 kV/140 kV sequential technique. Crucially, virtual non-contrast (VNC) images were generated from these DECT datasets using image postprocessing, specifically by subtracting iodine. This process effectively removes the iodinated contrast material, creating an image that mimics a true non-contrast scan without requiring an additional scan.

The study involved two radiologists independently evaluating both the VNC and the blended DECT arthrography images. They assessed image quality using a 5-point Likert scale and rated their diagnostic confidence for detecting periosteal calcifications and intraarticular loose bodies on a 4-point Likert scale. Quantitative analysis included measuring CT numbers (Hounsfield Units) for iodine, muscle, and joint fluid, as well as calculating image contrast and the percentage of iodine removal.

Datasets

The study utilized a dataset of 129 shoulders from 123 patients who underwent DECT shoulder arthrography. This dataset was collected retrospectively from Balgrist University Hospital. No new benchmark datasets were introduced; rather, the study evaluated the performance of VNC images generated from existing clinical DECT scans.

Key Insights and Ideas

  • Iodine Subtraction is Key: The core innovation is the application of iodine subtraction algorithms from DECT data to create VNC images. This technique effectively "removes" the contrast agent, which is a major limitation in standard CT arthrography for visualizing small intraarticular structures.
  • VNC Mimics True Non-Contrast: The generated VNC images provide a quality comparable to true non-contrast CT, allowing for better visualization of calcifications and loose bodies that might be masked by contrast.
  • Improved Detection of Loose Bodies: The study highlights a significant improvement in the detection rate of intraarticular loose bodies using VNC images compared to DECT arthrography. This is particularly important for small or subtle fragments.
  • Enhanced Confidence: Not only does VNC improve detection rates, but it also significantly increases the diagnostic confidence of radiologists when evaluating both loose bodies and periosteal calcifications.
  • Periosteal Calcifications vs. Labral Tears: VNC images are shown to be superior in differentiating periosteal calcifications from labral tears, which can appear similar on DECT arthrography due to the contrast material.

Key Results

  • Image Quality: VNC images were rated as good quality (median 4 out of 5).
  • Iodine Removal: An average of 91% of intraarticular iodine was removed on VNC images.
  • Loose Body Detection: The detection rate for intraarticular loose bodies was significantly higher on VNC images (10.9%) compared to DECT-A images (5.4%) (p=0.02).
  • Periosteal Calcification Detection: While detection rates were similar (19.4% for VNC vs. 16.3% for DECT-A), diagnostic confidence was significantly higher for VNC images (p=0.009).
  • Diagnostic Confidence: Confidence was significantly higher for both periosteal calcifications (p=0.009) and intraarticular loose bodies (p<0.001) on VNC images.
  • Radiation Dose: The effective dose of DECT-A scans was comparable to single-energy shoulder CT scans (mean 2.21 mSv).

Limitations and Future Work

  • Scanner Variability: The study used two different CT scanners, which could introduce variability, although acquisition parameters were adjusted.
  • Reconstruction Kernels: Different reconstruction kernels were used for DECT-A (bone) and VNC (soft tissue) images, which could influence visibility. The authors argue that a soft tissue kernel for DECT-A might have worsened visibility.
  • Blinding: Image reading was not fully blinded to image type (VNC vs. DECT-A), potentially introducing bias in confidence ratings.
  • Low Loose Body Count: The relatively low number of detected intraarticular loose bodies (14 on VNC, 7 on DECT-A) may limit the generalizability of these specific findings.
  • Lack of True Non-Contrast Comparison: For most patients, true unenhanced shoulder CT scans were not available for direct comparison with VNC images.

Future work could involve larger, multi-center studies with direct comparisons to true non-contrast CT and further exploration of VNC's utility in other musculoskeletal applications.

Summary updated

Abstract

VNC images from DECT shoulder arthrography are superior to DECT-A images for the detection of intraarticular loose bodies and increase the confidence in detecting periosteal calcifications.
BibTeX
@inproceedings{stern2022virtual,
  author = {Christoph Stern and Dimitri N. Graf and Samy Bouaicha and Karl Wieser and Andrea B. Rosskopf and Reto Sutter},
  title = {{Virtual non-contrast images calculated from dual-energy CT shoulder arthrography improve the detection of intraarticular loose bodies}},
  booktitle = {Skeletal},
  year = {2022},
  doi = {10.1007/s00256-022-04007-7},
}
Figure