Validation of deep-learning image reconstruction for coronary computed tomography angiography: Impact on noise, image quality and diagnostic accuracy

Dominik C Benz; Georgios Benetos; Georgios Rampidis; Elia von Felten; Adam Bakula; Aleksandra Sustar; Ken Kudura; Michael Messerli; Tobias A Fuchs; Catherine Gebhard; Aju P Pazhenkottil; Philipp A Kaufmann; Ronny R Buechel

doi:10.1016/j.jcct.2020.01.002

Validation of deep-learning image reconstruction for coronary computed tomography angiography: Impact on noise, image quality and diagnostic accuracy

J Cardiovasc Comput Tomogr. 2020 Sep-Oct;14(5):444-451. doi: 10.1016/j.jcct.2020.01.002. Epub 2020 Jan 13.

Authors

Affiliations

¹ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: dominik.benz@usz.ch.
² Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: Georgios.benetos@usz.ch.
³ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: grampidi@outlook.com.
⁴ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: e.v.f@hotmail.com.
⁵ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: adam.bakula@usz.ch.
⁶ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: Aleksandra.sustar@usz.ch.
⁷ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: ken.kudura@usz.ch.
⁸ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: michael.messerli@usz.ch.
⁹ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: tobias.fuchs@usz.ch.
¹⁰ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: catherine.gebhard@usz.ch.
¹¹ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: aju.pazhenkottil@usz.ch.
¹² Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: pak@usz.ch.
¹³ Department of Nuclear Medicine, Cardiac Imaging, University and University Hospital Zurich, Ramistrasse 100, 8091, Zurich, Switzerland. Electronic address: ronny.buechel@usz.ch.

PMID: 31974008
DOI: 10.1016/j.jcct.2020.01.002

Abstract

Background: Advances in image reconstruction are necessary to decrease radiation exposure from coronary CT angiography (CCTA) further, but iterative reconstruction has been shown to degrade image quality at high levels. Deep-learning image reconstruction (DLIR) offers unique opportunities to overcome these limitations. The present study compared the impact of DLIR and adaptive statistical iterative reconstruction-Veo (ASiR-V) on quantitative and qualitative image parameters and the diagnostic accuracy of CCTA using invasive coronary angiography (ICA) as the standard of reference.

Methods: This retrospective study includes 43 patients who underwent clinically indicated CCTA and ICA. Datasets were reconstructed with ASiR-V 70% (using standard [SD] and high-definition [HD] kernels) and with DLIR at different levels (i.e., medium [M] and high [H]). Image noise, image quality, and coronary luminal narrowing were evaluated by three blinded readers. Diagnostic accuracy was compared against ICA.

Results: Noise did not significantly differ between ASiR-V SD and DLIR-M (37 vs. 37 HU, p = 1.000), but was significantly lower in DLIR-H (30 HU, p < 0.001) and higher in ASiR-V HD (53 HU, p < 0.001). Image quality was higher for DLIR-M and DLIR-H (3.4-3.8 and 4.2-4.6) compared to ASiR-V SD and HD (2.1-2.7 and 1.8-2.2; p < 0.001), with DLIR-H yielding the highest image quality. Consistently across readers, no significant differences in sensitivity (88% vs. 92%; p = 0.453), specificity (73% vs. 73%; p = 0.583) and diagnostic accuracy (80% vs. 82%; p = 0.366) were found between ASiR-V HD and DLIR-H.

Conclusion: DLIR significantly reduces noise in CCTA compared to ASiR-V, while yielding superior image quality at equal diagnostic accuracy.

Keywords: ASiR-V; Adaptive statistical iterative reconstruction-veo; Coronary CT angiography; DLIR; Deep-learning image reconstruction; Diagnostic accuracy; Image quality.

Publication types

Comparative Study
Validation Study

MeSH terms

Aged
Artifacts
Computed Tomography Angiography*
Coronary Angiography*
Coronary Artery Disease / diagnostic imaging*
Coronary Vessels / diagnostic imaging*
Deep Learning*
Diagnosis, Computer-Assisted*
Female
Humans
Male
Middle Aged
Predictive Value of Tests
Radiographic Image Interpretation, Computer-Assisted*
Registries
Reproducibility of Results
Retrospective Studies