In addition, for this, a method to more completely parameterize the NPS was developed. Therefore, the purpose of this study was to investigate the just-noticeable differences (JNDs) in noise texture, as described by the shape of the downslope and the f peak of the NPS. Therefore, as a first step before such noise texture optimization can take place, it would be of interest to know what changes in noise texture are perceptible by a human observer. This may allow for additional manipulation of the noise texture during reconstruction, which could be of interest since noise texture influences detectability of lesions. Furthermore, using newly developed, deep-learning based, reconstruction algorithms, it seems possible to decouple resolution and noise texture from each other to a larger extent than in current reconstruction algorithms. Therefore, if the shape and extent of the NPS downslope have a significant influence on the perception of noise texture, then f peak is not a sufficient descriptor of it. However, the shape of the downslope of the NPS, past f peak, can vary considerably, independent of the value of f peak. Since the first section of the NPS, up to the f peak, is generally monotonically increasing, f peak can be assumed to sufficiently describe this portion of the NPS. Currently, it is common for the peak frequency ( f peak) to be used as a one-parameter descriptor of the NPS 2, 3.
The NPS of CT images reflects a ramp that dominates the lower spatial frequencies, and an apodization dominating the higher spatial frequencies 1. The noise texture in CT images is mainly dictated by the shape of the reconstruction kernel and can be quantified by the noise power spectrum (NPS).
In conclusion, both the peak frequency and the apodization section of the NPS influence the detectability of changes in image noise texture. This number changes if the apodization part of the NPS changes simultaneously. A change in only f peak of 0.2 lp/cm is below the detection threshold. The major radius makes an angle of 143° with the x-axis. The JND threshold ellipse is centered on the reference values and has a major and minor radius of 0.47 lp/cm and 0.12 lp/cm, respectively. Visibility thresholds for these changes were determined and an elliptical limiting detectability boundary was determined. A two alternative forced choice observer study was performed to determine the just noticeable- differences (JND) in f peak only, σ only, and both simultaneously. The σ of this Gaussian was used as the downslope descriptor of the NPS.
To quantify NPS downslope, half of a Gaussian function was fit through the NPS portion that lies beyond f peak.
NPSs were estimated using various reconstruction kernels on a commercial CT scanner. Therefore, we investigated the human-detectable differences in NPSs having different f peak and/or downslope parameters. However, if the downslope of the NPS beyond the f peak influences noise texture visibly, then f peak is insufficient as a single descriptor. The peak frequency of the NPS (f peak) is often used as a one-parameter metric for characterizing noise texture. Noise texture in CT images, commonly characterized by using the noise power spectrum (NPS), is mainly dictated by the shape of the reconstruction kernel.
0 kommentar(er)
