Robustness of radiomics features of virtual unenhanced and virtual monoenergetic images in dual-energy CT among different imaging platforms and potential role of CT number variability

Impacts of Adaptive Statistical Iterative Reconstruction-V and Deep Learning Image Reconstruction Algorithms on Robustness of CT Radiomics Features: Opportunity for Minimizing Radiomics Variability Among Scans of Different Dose Levels

This study aims to investigate the influence of adaptive statistical iterative reconstruction-V (ASIR-V) and deep learning image reconstruction (DLIR) on CT radiomics feature robustness. A standardized phantom was scanned under single-energy CT (SECT) and dual-energy CT (DECT) modes at standard and low (20 and 10 mGy) dose levels. Images of SECT 120 kVp and corresponding DECT 120 kVp-like virtual monochromatic images were generated with filtered back-projection (FBP), ASIR-V at 40% (AV-40) and 100% (AV-100) blending levels, and DLIR algorithm at low (DLIR-L), medium (DLIR-M), and high (DLIR-H) strength levels. Ninety-four features were extracted via Pyradiomics. Reproducibility of features was calculated between standard and low dose levels, between reconstruction algorithms in reference to FBP images, and within scan mode, using intraclass correlation coefficient (ICC) and concordance correlation coefficient (CCC). The average percentage of features with ICC > 0.90 and CCC > 0.90 between the two dose levels was 21.28% and 20.75% in AV-40 images, and 39.90% and 35.11% in AV-100 images, respectively, and increased from 15.43 to 45.22% and from 15.43 to 44.15% with an increasing strength level of DLIR. The average percentage of features with ICC > 0.90 and CCC > 0.90 in reference to FBP images was 26.07% and 25.80% in AV-40 images, and 18.88% and 18.62% in AV-100 images, respectively, and decreased from 27.93 to 17.82% and from 27.66 to 17.29% with an increasing strength level of DLIR. DLIR and ASIR-V algorithms showed low reproducibility in reference to FBP images, while the high-strength DLIR algorithm provides an opportunity for minimizing radiomics variability due to dose reduction.

Experimental study on the effects of different exposure conditions and contrast agent concentrations on spectral computed tomography virtual non-contrast images

Background

The early diagnosis of thrombosis and fat embolism is important for subsequent treatment regimens. Spectral computed tomography (CT) virtual non-contrast (VNC) scanning can not only accurately diagnose thrombosis and medium fat embolism but can also reduce the radiation dose and scanning time. However, there is a relative paucity of studies on what contrast concentration and exposure conditions are best for the quality of VNC images. To address this issue, this study aimed to investigate the effects of different exposure conditions and contrast concentrations on the quality of VNC images of low-density substances in spectral CT.

Methods

Four solution groups [i.e., groups A (15 mgI/mL), B (10 mgI/mL), C (5 mgI/mL), and D (the control group)] were matched with normal saline and contrast agent groups. Four groups of solution, duck blood clots, and fat were injected into four sections of the pig large intestine, respectively. CT scans with different exposure amounts were performed under the condition of 120 KV. Comparing the true non-contrast (TNC) image based on solution D group with the VNC images of the other three solution groups. The differences in the CT values, standard deviation (SD) values, and contrast noise ratio (CNR) values of the duck blood and fat under different iodine concentrations and exposures were compared. The image quality was evaluated using a three-point method and the Kappa consistency test was performed. The consistency of the tissue CT values in the TNC and VNC images was analyzed by drawing Bland-Altman scatter plots.

Results

The CT values of the duck blood in the VNC20mAs and VNCC groups were lower than those in the TNC groups (P<0.05). Under different exposures and contrast agent concentrations, the CT value of the fat in the VNC group was higher than that in the TNC group (P<0.05). The SD values of the duck blood and fat in three groups (i.e., groups A, B, and C) were lower than those in the TNC group (P<0.05). The CNR value of the duck blood in the VNC20mAs group was lower than that in the TNC group (Z=-2.10, P=0.04), and the CNR values of the duck blood and fat in the VNC group were higher than those in the TNC groups in the remaining different exposure and concentration groups (P<0.05). The CT values of the lesions in the two groups were consistent, and there were no statistically significant differences between the subjective scores of the TNC and VNC images (z=-1.34, P=0.18); the subjective evaluations of the two physicians had good consistency (K=0.80).

Conclusions

Under the conditions of higher contrast agent concentrations and proper exposure conditions, the VNC images were better able to restore the CT values of the blood clots, reduce the SD values of the blood clots and fat. In addition, and improve the CNR values of the blood clots and fat. In addition, the quality of the two images was similar.

Robustness of Quantitative Diffusion Metrics from Four Models: A Prospective Study on the Influence of Scan-Rescans, Voxel Size, Coils, and Observers

BACKGROUND

Quantitative diffusion metrics provide additional microstructural information of diseases. The robustness of quantitative diffusion metrics should be established before clinical application.

PURPOSE

To evaluate the variability and reproducibility of quantitative diffusion MRI metrics.

STUDY TYPE

Prospective.

POPULATION

14 volunteers (7 men; median age, range, 28, 26-59 years).

FIELD STRENGTH/SEQUENCE

3.0-T/Diffusion spectrum imaging.

ASSESSMENT

Brain MRI studies were performed four times per subject: involving different combinations of coil types and voxel sizes. Regions of interest of 13 brain anatomical sites were drawn by one observer twice and another observer once to allow interobserver and intraobserver reproducibility assessment. Twenty-five quantitative metrics were calculated using four diffusion models.

STATISTICAL TESTS

The variability was evaluated with coefficients of variation (CV), and quartile coefficient of dispersion (QCD). The reproducibility was assessed with intraclass correlation coefficient (ICC), and concordance correlation coefficient (CCC). Wilcoxon signed rank test was used to compare the influence of factors on robustness of quantitative diffusion metrics. A two-tailed P < 0.05 was considered statistically significant.

RESULTS

The variability of quantitative diffusion metrics showed CV of 2.4%-68.2%, and QCD of 0.6%-48.2%, respectively. The reproducibility of scans using 20-channel coils with voxels of 2 × 2 × 2 mm3 and 3 × 3 × 3 mm3 , respectively (ICC 0.03-0.84, CCC 0.03-0.84) was significantly worse than that of repeated scans using a 20-channel coil with a voxel size of 2 × 2 × 2 mm3 (ICC of 0.74-0.97, CCC 0.74-0.97) and that of scans using 20- and 64-channel coils, respectively, with a voxel size of 2 × 2 × 2 mm3 (ICC 0.59-0.95, CCC 0.59-0.95). The intraobserver reproducibility (ICC 0.49-0.94, CCC 0.49-0.94) was significantly better than the interobserver reproducibility (ICC 0.28-0.91, CCC 0.28-0.91).

DATA CONCLUSION

Our study indicated that the voxel size has a greater influence on the reproducibility of quantitative diffusion metrics than scan-rescans and coils. The reproducibility within one observer was higher than that between two observers.

EVIDENCE LEVEL

2 TECHNICAL EFFICACY: Stage 1.