RESUMO
Background: The effect of region of interest (ROI) size variation on producing accurate noise levels is not yet studied. Objective: This study aimed to evaluate the influence of ROI sizes on the accuracy of noise measurement in computed tomography (CT) by using images of a computational and American College of Radiology (ACR) phantoms. Material and Methods: In this experimental study, two phantoms were used, including computational and ACR phantoms. A computational phantom was developed by using Matlab R215a software (Mathworks Inc., Natick, MA Natick, MA) with a homogeneously +100 Hounsfield Unit (HU) value and an added-Gaussian noise with various levels of 5, 10, 25, 50, 75, and 100 HU. The ACR phantom was scanned with a Philips MX-16 slice CT scanner in different slice thicknesses of 1.5, 3, 5, and 7 mm to obtain noise variation. Noise measurement was conducted at the center of the phantom images and four locations close to the edge of the phantom images using different ROI sizes from 3 × 3 to 41 × 41 pixels, with an increased size of 2 × 2 pixels. Results: The use of a minimum ROI size of 21 × 21 pixels shows noise in the range of ± 5% ground truth noise. The measured noise increases above the ± 5% range if the used ROI is smaller than 21 × 21 pixels. Conclusion: A minimum acceptable ROI size is required to maintain the accuracy of noise measurement with a size of 21 × 21 pixels.
RESUMO
The purpose of this study was to develop a computational phantom for validation of automatic noise calculations applied to all parts of the body, to investigate kernel size in determining noise, and to validate the accuracy of automatic noise calculation for several noise levels. The phantom consisted of objects with a very wide range of HU values, from -1000 to +950. The incremental value for each object was 10 HU. Each object had a size of 15 × 15 pixels separated by a distance of 5 pixels. There was no dominant homogeneous part in the phantom. The image of the phantom was then degraded to mimic the real image quality of CT by convolving it with a point spread function (PSF) and by addition of Gaussian noise. The magnitude of the Gaussian noises was varied (5, 10, 25, 50, 75 and 100 HUs), and they were considered as the ground truth noise (NG). We also used a computational phantom with added actual noise from a CT scanner. The phantom was used to validate the automated noise measurement based on the average of the ten smallest standard deviations (SD) from the standard deviation map (SDM). Kernel sizes from 3 × 3 up to 27 × 27 pixels were examined in this study. A computational phantom for automated noise calculations validation has been successfully developed. It was found that the measured noise (NM) was influenced by the kernel size. For kernels of 15 × 15 pixels or smaller, the NMvalue was much smaller than the NG. For kernel sizes from 17 × 17 to 21 × 21 pixels, the NMvalue was about 90% of NG. And for kernel sizes of 23 × 23 pixels and above, NMis greater than NG. It was also found that even with small kernel sizes the relationship between NMand NGis linear with R2more than 0.995. Thus accurate noise levels can be automatically obtained even with small kernel sizes without any concern regarding the inhomogeneity of the object.