Diagnostic Accuracy of Subtraction Coronary CT Angiography in Severely Calcified Segments: Comparison Between Readers With Different Levels of Experience

Fang Li, Qing He, Lixue Xu, Yan Zhou, Yufei Sun, Zhenchang Wang, Yinghao Xu, Zhenghan Yang, Yi HeFrontiers in Cardiovascular Medicine 2022 · DOI

TL;DR

Coronary artery calcification often creates "blooming" artifacts that lead radiologists to overestimate heart disease severity. This study demonstrates that subtraction coronary CT angiography (CCTA)—a technique that digitally removes calcium from images—significantly boosts diagnostic accuracy and confidence for radiologists across all experience levels, from novices to seasoned experts.

What Problem Does This Paper Address?

Coronary Computed Tomography Angiography (CCTA) is a standard non-invasive tool for diagnosing heart disease. However, it has a "calcium problem." When arteries are heavily calcified, the dense calcium deposits create "blooming" and "beam-hardening" artifacts. On a scan, these look like bright, oversized glows that obscure the actual opening (lumen) of the blood vessel.

Because of this glare, radiologists often overestimate how narrow an artery is, leading to high "false positive" rates. This challenge is particularly daunting for less experienced doctors. This paper investigates whether a subtraction technique—taking a "before" and "after" contrast scan to digitally erase the calcium—can level the playing field and improve diagnostic precision.

Methods

The researchers conducted a prospective study using a 320-detector row CT scanner. The process involved:

  1. Dual Scanning: Patients underwent a non-contrast scan followed by a contrast-enhanced scan.
  2. Digital Subtraction: Using specialized software, the non-contrast image (showing only calcium) was registered and subtracted from the contrast image, theoretically leaving only the blood-filled lumen visible.
  3. Multi-Level Reader Comparison: Eight radiologists were divided into three groups based on their career experience:
  4. Novices: 50–100 CCTA interpretations.
  5. Juniors: ~500 interpretations.
  6. Seniors: >1,000 interpretations.
  7. Validation: All results were compared against Invasive Coronary Angiography (ICA), the clinical "gold standard" for measuring vessel stenosis.

Datasets

The study utilized a primary clinical dataset collected between October 2019 and June 2020:

  • Participants: 47 patients with suspected coronary artery disease.
  • Target Segments: 134 severely calcified arterial segments.
  • Reference: Each patient also underwent ICA within one month of the CCTA for direct comparison.

Key Insights and Ideas

  • The Specificity Breakthrough: The most dramatic finding was the surge in specificity. In conventional CCTA, radiologists of all levels were "over-calling" disease because of calcium glare. Subtraction allowed them to see that many "blocked" vessels were actually clear.
  • The Sensitivity Trade-off: Interestingly, sensitivity (the ability to catch every actual blockage) decreased slightly for novices and juniors after subtraction. This suggests that without the "scary-looking" calcium, less experienced readers might occasionally miss subtle narrowings, whereas senior readers maintained high sensitivity.
  • Experience Still Matters: While subtraction helped everyone, senior radiologists still achieved the highest overall accuracy (AUC of 0.85). The technology is a powerful assistant, but it doesn't entirely replace the "trained eye" developed over thousands of cases.
  • Confidence Boost: Diagnostic confidence scores improved significantly across the board. Radiologists felt much more certain about their diagnoses when the calcium "noise" was removed.

Key Results

  • Specificity: Jumped from a dismal range of 22.4%–42.2% (conventional) to a robust 81.3%–85.7% (subtraction) across all groups.
  • Overall Accuracy (AUC):
    • Novices: Improved from 0.55 to 0.73.
  • Juniors: Improved from 0.56 to 0.76.
  • Seniors: Improved from 0.63 to 0.85.
  • Uninterpretable Segments: The number of segments deemed "impossible to read" dropped to zero for senior readers and was significantly reduced for novices and juniors when using subtraction.

Limitations and Future Work

  • Misregistration: The biggest technical hurdle is "misregistration artifacts." If a patient moves or breathes differently between the two scans, the subtraction won't align perfectly, potentially creating new errors. In this study, 35% of segments had some degree of misregistration.
  • Radiation Dose: Because subtraction requires two scans (non-contrast and contrast), there is a potential for higher radiation exposure, though the authors used low-dose protocols to mitigate this.
  • Sample Size: The study featured a relatively small number of readers (8) and patients (47). The authors suggest that larger, multi-center studies are needed to confirm these findings in broader clinical practice.
  • Visual vs. Quantitative: Calcification was defined by visual assessment rather than Agatston scores (a standard calcium scoring metric), which might introduce some subjectivity in segment selection.

Summary updated

Abstract

Subtraction CCTA could improve the diagnostic accuracy of radiologists at all experience levels of CCTA interpretation.
BibTeX
@inproceedings{li2022diagnostic,
  author = {Fang Li and Qing He and Lixue Xu and Yan Zhou and Yufei Sun and Zhenchang Wang and Yinghao Xu and Zhenghan Yang and Yi He},
  title = {{Diagnostic Accuracy of Subtraction Coronary CT Angiography in Severely Calcified Segments: Comparison Between Readers With Different Levels of Experience}},
  booktitle = {Frontiers in Cardiovascular Medicine},
  year = {2022},
  doi = {10.3389/fcvm.2022.828751},
}
Figure