Transcatheter aortic valve implantation (TAVI) planning CT on 8-cm detector scanners: Proper dose control by combined use of two deep-learning reconstruction algorithms.

PURPOSE

To explore the feasibility of transcatheter aortic valve implantation (TAVI) planning computed tomography (CT) on single-source 8-cm detector scanners with proper dose control by using two deep-learning reconstruction algorithms.

METHODS

Reduced-dose TAVI planning CT was simulated by replacing routine aortic CT angiography (CTA) with a reduced-dose aortic CTA and a reduced-dose coronary CTA (Group A, n = 82), while keeping the total dose unchanged. Each of the two CTA scans was processed with a different deep-learning reconstruction algorithm. Routine-dose coronary CTA (Group B, n = 86) and routine-dose aortic CTA (Group C, n = 77) with hybrid iterative reconstruction were used as reference for evaluating the acceptability for surgical planning (A vs. B for aortic valve; A vs. C for access route) and for comparing both the diagnostic and objective image quality (A vs. B for coronary arteries; A vs. C for aortoiliac arteries).

RESULTS

The mean effective dose in Group A was 8.22 ± 0.83 mSv, representing a 57% reduction of the routine-dose TAVI planning CT, that is, a routine-dose coronary CTA plus a routine-dose aortic CTA on the same scanner model. With respect to B and C, images in A were scored higher for evaluating the aortic valve (p = 0.045) and the access route (p = 0.014) and for diagnosing the thoracic aorta and iliac segments (p < 0.050), while the diagnostic confidence were comparable on the coronary arteries (p > 0.050), abdominal aorta (p = 0.276), and femoral segment (p = 0.816). The image noise in A was found to be 21%-55% lower, leading to a significant increase in contrast-to-noise ratio (CNR) by 63%-114% (p < 0.050).

CONCLUSION

Reduced-dose TAVI planning CT is feasible on 8-cm detector scanners by using deep-learning reconstruction algorithms, showing promise of implementing the examination in imaging settings that are more commonly accessible.