Improving spatial resolution and diagnostic confidence with thinner slice and deep learning image reconstruction in contrast-enhanced abdominal CT

To investigate whether an adaptive statistical iterative reconstruction (ASIR) algorithm improves the image quality at low-tube-voltage (80-kVp), high-tube-current (675-mA) multidetector abdominal computed tomography (CT) during the late hepatic arterial phase. This prospective, single-center HIPAA-compliant study was institutional review board approved. Informed patient consent was obtained. Ten patients (six men, four women; mean age, 63 years; age range, 51-77 years) known or suspected to have hypervascular liver tumors underwent dual-energy 64-section multidetector CT. High- and low-tube-voltage CT images were acquired sequentially during the late hepatic arterial phase of contrast enhancement. Standard convolution FBP was used to reconstruct 140-kVp (protocol A) and 80-kVp (protocol B) image sets, and ASIR (protocol C) was used to reconstruct 80-kVp image sets. The mean image noise; contrast-to-noise ratio (CNR) relative to muscle for the aorta, liver, and pancreas; and effective dose with each protocol were assessed. A figure of merit (FOM) was computed to normalize the image noise and CNR for each protocol to effective dose. Repeated-measures analysis of variance with Bonferroni adjustment for multiple comparisons was used to compare differences in mean CNR, image noise, and corresponding FOM among the three protocols. The noise power spectra generated from a custom phantom with each protocol were also compared. When image noise was normalized to effective dose, protocol C, as compared with protocols A (P = .0002) and B (P = .0001), yielded an approximately twofold reduction in noise. When the CNR was normalized to effective dose, protocol C yielded significantly higher CNRs for the aorta, liver, and pancreas than did protocol A (P = .0001 for all comparisons) and a significantly higher CNR for the liver than did protocol B (P = .003). Mean effective doses were 17.5 mSv +/- 0.6 (standard error) with protocol A and 5.1 mSv +/- 0.3 with protocols B and C. Compared with protocols A and B, protocol C yielded a small but quantifiable noise reduction across the entire spectrum of spatial frequencies. Compared with standard FBP reconstruction, an ASIR algorithm improves image quality and has the potential to decrease radiation dose at low-tube-voltage, high-tube-current multidetector abdominal CT during the late hepatic arterial phase.

The purpose of this study was to compare the image quality of coronary computed tomography angiography (CTA) subjected to deep learning-based image restoration (DLR) method with images subjected to hybrid iterative reconstruction (IR).