Visual impression and texture analysis of advanced modeled iterative reconstruction (ADMIRE): improved assessment of image quality in CT for better estimation of dose reduction potential

Abstract

To evaluate the image quality (IQ) of advanced modeled iterative reconstruction (ADMIRE; Siemens Healthcare, Forchheim, Germany) applying image texture and image visual impression as a supplement to physical parameters such as noise level and spatial resolution. An ACR-phantom with four modules was examined at different radiation dose levels. To characterise the image texture, two Haralick texture parameters, contrast and entropy, were assessed at different dose levels and reconstruction algorithms. The visual impression of images and the low-contrast detectability were evaluated by the structural similarity index (SSIM). The spatial resolution was determined by the modulation transfer functions and the line spread function. The Haralick texture parameters, contrast and entropy, decreased with increasing ADMIRE levels. ADMIRE III, IV and V offered a comparable contrast and entropy to those calculated by filtered back projection (FBP) with a radiation dose reduction up to 50%. SSIM (low-contrast detectability) improved with increasing ADMIRE levels. SSIM calculated by ADMIRE IV and V revealed comparable IQ to FBP with a decreased CTDI_volup to 50%. Spatial resolution was retained up to 90% dose reduction. Compared to FBP at the same dose level, the image noise decreased up to 61% with higher ADMIRE levels (σ_FBP= 17.3 HU andσ_ADMIREV= 10.6 HU at 6.65 mGy). Taking texture analysis and visual perception into account, a more realistic assessment of the dose reduction potential of ADMIRE can be achieved than quality metrics based alone on physical measurements.

Keywords: ADMIRE; Haralick texture parameters; image texture analysis; image visual impression; low-contrast detectability; structural similarity.