Evaluation of non-linear blending in dual-energy computed tomography

Abstract

Dual-energy CT scanning has significant potential for disease identification and classification. However, it dramatically increases the amount of data collected and therefore impacts the clinical workflow. One way to simplify image review is to fuse CT datasets of different tube energies into a unique blended dataset with desirable properties.A non-linear blending method based on a modified sigmoid function was compared to a standard 0.3 linear blending method. The methods were evaluated in both a liver phantom and patient study. The liver phantom contained six syringes of known CT contrast which were placed in a bovine liver. After scanning at multiple tube currents (45, 55, 65, 75, 85, 95, 105, and 115 mAs for the 140-kV tube), the datasets were blended using both methods. A contrast-to-noise (CNR) measure was calculated for each syringe. In addition, all eight scans were normalized using the effective dose and statistically compared. In the patient study, 45 dual-energy CT scans were retrospectively mixed using the 0.3 linear blending and modified sigmoid blending functions. The scans were compared visually by two radiologists.For the 15, 45, and 64 HU syringes, the non-linear blended images exhibited similar CNR to the linear blended images; however, for the 79, 116, and 145 HU syringes, the non-linear blended images consistently had a higher CNR across dose settings. The radiologists qualitatively preferred the nonlinear blended images of the phantom. In the patient study, the radiologists preferred non-linear blending in 31 of 45 cases with a strong preference in bowel and liver cases.Non-linear blending of dual energy data can provide an improvement in CNR over linear blending and is accompanied by a visual preference for non-linear blended images. Further study on selection of blending parameters and lesion conspicuity in non-linear blended images is being pursued.

Maximizing Iodine Contrast-to-Noise Ratios in Abdominal CT Imaging through Use of Energy Domain Noise Reduction and Virtual Monoenergetic Dual-Energy CT

“…In general, DE CT is performed by acquiring projection data with two different x-ray beam spectra. A number of different image types can be generated from the resulting data, including images corresponding to the low- and high-energy spectra, mixed images that represent a linear or nonlinear blending of the low- and high-energy images, material-specific images, and virtual monoenergetic images (17–20). …”

Renal Masses

“…In particular, the high-intensity regions corresponding to iodine contrast-enhancement are preferentially extracted by 80-kVp data, to maintain the high image contrast in regions of high iodine concentration. By contrast, low-intensity levels, which correspond to air, water, and soft tissues, are less noisy in 140-kV scan and should be preferred in the blending process [49,50]. The result is an image in which iodine signal is magnified, while optimizing contrast differences and minimizing image noise.…”

Evaluating optimal CNR as a preset criteria for nonlinear moidal blending of dual energy CT data

“…The 80 and 140 kV datasets were blended using the moidal blending function. Every combination of moidal level and width was applied to the data, and the CNR was calculated using the method from [8].…”

“…A review of several of these functions can be found in [7]. One approach, a modified sigmoid blending referred to as "Moidal Blending," was viewed favorably by radiologists in [8]. The moidal blending function, shown in Figure 1, requires two parameters.…”