Evaluation of Low-contrast Detectability Using the Digital Phantom Creation Tool in the Late Arterial Phase to Detect Liver Mass Lesions

• Published in: Japanese Journal of Radiological Technology Vol. 79; no. 7; pp. 674 - 681
• Main Authors: Harada, Kohei, Imai, Tatsuya, Ohashi, Yoshiya, Chiba, Ayaka, Numasawa, Kanako, Hayasaka, Shun, Omori, Go
• Format: Journal Article
• Language: Japanese; English
• Published: Japan Japanese Society of Radiological Technology 01.01.2023; Japan Science and Technology Agency
• Subjects: Algorithms; Correlation coefficient; Correlation coefficients; digital phantom; Humans; Image acquisition; Image contrast; Image quality; Image reconstruction; Lesions; Liver; Liver - diagnostic imaging; low-contrast detectability; Phantoms, Imaging; Radiation Dosage; Radiographic Image Interpretation, Computer-Assisted - methods; Reproducibility of Results; Simulation; standardization; Tomography, X-Ray Computed - methods; Tumors; low-contrast detectability; image quality; standardization; digital phantom; liver
• ISSN: 0369-4305; 1881-4883
• DOI: 10.6009/jjrt.2023-1360

Purpose: Late arterial phase images of SD 8, SD 10, and SD 12 were acquired in the 3-phase dynamic study of the liver in combination with hybrid iterative reconstruction. We evaluated the low-contrast detectability by adding a simulated tumor to these images and aimed to formulate a standard image quality. Methods: We prepared images with and without signal for 60 series of 20 samples, each with 3 image quality types (total: 120 series). The continuous confidence method by 10 observers detected 60 simulated tumors. Results: The detection sensitivities were 0.765, 0.785, and 0.260 for SD 8, SD 10, and SD 12, respectively (p<0.001) with no significantly different specificities, and the areas under the curve were 0.901, 0.892, and 0.616 (p<0.001), respectively. The simulated mass detection rates were 74.5%, 75.0%, and 21.5% for SD 8, SD 10, and SD 12, respectively (p<0.001), and the intraclass correlation coefficients, which indicate interobserver reliability, were 0.697 at SD 10 without signal, and SD 12 without a signal significantly dropped to 0.185. Conclusion: Therefore, SD 12 images increase the possibility of overlooking lesions. Hence, image quality in the late arterial phase should be SD 10 or less.