Three-dimensionally consolidated reality of image quality of CT images reconstructed by iterative reconstruction methods

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K09883
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Radiation science
• Research Institution: Tohoku University
• Principal Investigator: Mori Issei 東北大学, 医学系研究科, 大学院非常勤講師 (90375171)
• Co-Investigator(Kenkyū-buntansha): 町田 好男 東北大学, 医学系研究科, 教授 (30507083)
• Co-Investigator(Renkei-kenkyūsha): HOMMA NORIYASU 東北大学, 大学院医学系研究科, 教授 (30282023)
• Research Collaborator: GOTO MITSUNORI
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 医用画像工学 / X線CT

Outline of Final Research Achievements

Iterative reconstruction (IR) methods is said to reduce radiation dose while maintaining image quality. Numerous task-based image quality assessment, which mimics clinical tasks, support the superiority of IR methods over existing filtered backprojection (FBP) method. However, the test object is visually too distinct: it does not reflect clinically difficult/important task. Further, it ignores slice sensitivity profile (SSP).
We developed methods to measure in-plane resolution and SSP using indistinct test objects buried in noise. We also measured three-dimensional noise power spectrum (NPS) accurately. Using these methods, we evaluated matched filter SNR (MFSNR) of indistinct object for images of IR and FBP. Resultantly, none of three IR methods showed better MFSNR than FBP. Consequently, IR methods do not improve the detectability of indistinct object: radiation dose cannot be reduced. The predominant task-based evaluation method necessarily overvalues IR methods.

Free Research Field

医歯薬学