Using a deep-learning image reconstruction algorithm with dual-energy CT portal venography improves the technique’s performance, according to research published January 2 in Academic Radiology.
The findings could translate to better liver transplantation planning, wrote a team led by medical student Chong Meng of Affiliated Hospital of Xuzhou Medical University in Jiangsu, China.
“By providing consistent, high-quality imaging across multiple portal venous segments, deep-learning image reconstruction may offer a safer and more reliable approach for preoperative evaluation and postoperative monitoring in liver transplantation,” the group noted.
Deep-learning image reconstruction (DLIR) has shown promise for improving image quality in low-dose CT (LDCT) imaging, but its performance in dual-energy CT portal venography (DE-CTPV) — especially under reduced contrast medium volume and radiation dose conditions — is unclear, the investigators wrote.
To address this knowledge gap, Meng’s group conducted a study that compared the performance of DLIR and adaptive statistical iterative reconstruction (GE HealthCare‘s ASiR-V software) in dual-energy CT portal venography performed using dual-low dose protocols, tracking image quality across a range of vascular segments of the portal venous system.
Study participants’ dual-energy CT portal venography exams were reconstructed using DLIR medium (DLIR-M), DLIR high strength (DLIR-H), and ASiR-V (50%). Radiologist readers assessed image quality in the main portal vein, the left and right portal veins, the splenic vein, and the superior mesenteric vein, evaluating image noise, contrast-to-noise ratio, and signal-to-noise ratio. The investigators also compared radiation dose parameters (CT dose index volume, or CTDIvol; dose-length product, or DLP; and effective dose, or ED) and contrast medium volume findings to data from previous studies.
The researchers reported a mean CTDIvol of 9.79 mGy, a mean DLP of 326.26 mGy.cm, and a mean ED of 4.89 mSv. Mean contrast medium volume was 79.5 mL. They also reported that, overall, DLIR successfully reduced noise and provided clear, consistent vascular definition across all segments assessed in the study.
Meng and colleagues noted the following:
- DLIR high strength enhanced image quality across all vascular segments, substantially reducing image noise and increasing contrast-to-noise ratio and signal-to-noise ratio (p < 0.05).
- The interpreting radiologists also gave DLIR high strength top subjective ratings for overall image quality, image noise, vascular edge sharpness, and diagnostic confidence compared to ASiR-V 50%.
- Use of 55 keV virtual monoenergetic imaging improved iodine contrast effectiveness.
The takeaway? “Deep-learning image reconstruction substantially improves image quality in dual-energy CT portal venography compared with ASiR-V 50%, even when utilizing dual-low dose protocol,” the group concluded.
Access the full study here.