Task-specific observer performance and image quality : direct and indirect relationships in low-dose CT images

Jessop, LMM

Abstract

Aim: The aim of this research was to examine image quality in low-dose CT images, as determined by a range of image quality measures (IQM), in addition to psychophysical assessment, and look at direct or indirect relationships.
Method: CT images of an anthropomorphic chest phantom were obtained using increasing tube current (mA) to vary image quality. Tube current was increased incrementally (15mA- 100mA)
and 200mA (reference image). Three sets of simulated lesions (sizes 5mm, 8mm and 10mm) of density 100HU, -630HU and -800HU were imaged one density at a time, using a 16 slice CT. A
normal series was also acquired. These image sets were repeated using attenuation jackets, to achieve further image degradation. Images (5mm slice thickness) were reconstructed using filtered back-projection. Image analyses were carried out on 235 images of which 39 were normal. IQMs used were signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), mean square error (MSE), peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), non-shift-edge ratio (NSER), texture analysis and 3D noise power spectrum (NPS). Visual grading characteristic (VGC) observer performance studies were performed with seven observers using a localisation task, and observer tasks involving visualisation of structures with how well images fulfilled international CT quality criteria, and the impact of noise on those decisions.
Results and conclusion: The relationship of mA to IQM values was mainly logarithmic, with patterns of undulations in graphed data indicating potential for reduction in radiation dose. Additionally, attenuation jackets significantly affected IQM values, as did regional variation. NSER, Energy(uniformity) and Homogeneity showed the strongest inter-correlation and an inverse relationship with Entropy. Observer performance studies demonstrated no significant
difference in lesion localisations at different amperage, however there was strong correlation between the impact of image noise and the fulfilment of visualisation criteria in CT images. NSER, Energy (uniformity), Homogeneity and Entropy had moderate to strong correlation with human observers. This research demonstrated the importance of simulating body habitus, and
using the appropriate IQM, when assessing image quality for the task intended.

Citation

Jessop, L. Task-specific observer performance and image quality : direct and indirect relationships in low-dose CT images. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Nov 9, 2021
• Publicly Available Date: Nov 9, 2021
• Award Date: Jan 16, 2021