Evaluation of image quality between hybrid iterative reconstruction and deep learning reconstruction in low dose abdominopelvic CT in low body mass index individuals.
Authors
Affiliations (3)
Affiliations (3)
- Department of Medical Imaging Technology, Manipal College of Health Professions, Manipal Academy of Higher Education, Manipal, India.
- Department of Medical Imaging Technology, Manipal College of Health Professions, Manipal Academy of Higher Education, Manipal, Karnataka, India. [email protected].
- Radio-Diagnosis & Imaging, Department of Radiodiagnosis & Medical Imaging, Kasturba Medical College, Manipal Academy of Higher Education, Manipal, Karnataka, India.
Abstract
The study aimed to compare image quality between hybrid iterative reconstruction and deep learning image reconstruction (DLIR) in low dose contrast enhanced CT (CECT) abdomen and pelvis scans in low body mass index (BMI) individuals, to determine whether DLIR provides superior image quality at a given dose. Fifty patients with BMI < 18.5 kg/m<sup>2</sup> referred for CECT abdomen and pelvis were scanned on a 128-slice Philips CT scanner at 80 kVp. Images were reconstructed using iDose4 (level 4) and DLIR (precise image (PI), standard strength). Images were then analyzed quantitatively to estimate signal-to-noise ratio (SNR), contrast to noise ratio (CNR) and image noise by placing region of interest in aorta, liver, spleen and portal vein across the non-contrast, arterial and portal venous series. Qualitative analysis was performed by two radiologists using a 5-point Likert scale based on five image quality criteria and artifacts were assessed separately using 4-pont Likert scale. Paired t test/ Wilcoxon signed rank test was used to compare image quality parameters between both reconstruction algorithms. The mean attenuation (HU) did not significantly differ between the two reconstruction techniques for all the organs in the non-contrast, arterial and portal venous series. A significant increase in the SNR <sub>liver</sub> was noted using precise image-standard across the non-contrast (idose4: 6.3 ± 1.7 vs PI: 6.6 ± 1.7, p < 0.001), arterial (idose4: 7.2 ± 1.9 vs PI: 7.7 ± 1.9, p < 0.001) and portal venous (idose4: 12.9 ± 2.41 vs PI: 13.7 ± 2.43, p < 0.001) phases. Similarly, CNR <sub>liver</sub> significantly increased across the non-contrast [idose4: 0.09(- 0.4,0.6) vs PI: 0.4(- 0.3, 1), p < 0.001] and portal venous [idose4: 6 (4,7.17) vs PI: 6.47 (4.49, 7.62) p < 0.001] phases. Image noise was significantly lower with PI compared to iDose4 in the non- contrast (iDose4: 7.87 ± 0.97 vs PI: 7.59 ± 0.93, p value = 0.001), arterial (iDose4: 10.4 ± 1.42 vs PI: 9.83 ± 1.29, p < 0.001) and portal venous (idose4: 10.1 (9.6, 11.1) vs PI: 9.4 (8.9, 10.2), p < 0.001) phases. In qualitative analysis, Interobserver agreement between the two radiologists demonstrated good to excellent agreement (ICC range: 0.70-0.96). All the images were graded 3, 4 or 5, indicating good to excellent image quality with no artifacts. DLIR (Precise image-standard) offers a promising approach for improving image quality in low-dose abdominal and pelvic CT scans, particularly in individuals with a low BMI.