[Reduction of Motion Artifacts in Liver MRI Using Deep Learning with High-pass Filtering].

PURPOSE: To investigate whether deep learning with high-pass filtering can be used to effectively reduce motion artifacts in magnetic resonance (MR) images of the liver. METHODS: The subjects were 69 patients who underwent liver MR examination at our hospital. Simulated motion artifact images (SMAIs) were created from non-artifact images (NAIs) and used for deep learning. Structural similarity index measure (SSIM) and contrast ratio (CR) were used to verify the effect of reducing motion artifacts in motion artifact reduction image (MARI) output from the obtained deep learning model. In the visual assessment, reduction of motion artifacts and image sharpness were evaluated between motion artifact images (MAIs) and MARIs. RESULTS: The SSIM values were 0.882 on the MARIs and 0.869 on the SMAIs. There was no statistically significant difference in CR between NAIs and MARIs. The visual assessment showed that MARIs had reduced motion artifacts and improved sharpness compared to MAIs. CONCLUSION: The learning model in this study is indicated to be reduced motion artifacts without decreasing the sharpness of liver MR images.

Quantitative evaluation of functional liver reserve using the liver-to-spleen ΔR1 ratio based on gadoxetic acid-enhanced MRI: corrected for the effect of gadoxetic acid dose.

BACKGROUND: The liver T1 reduction rate can be used to assess liver function. However, higher doses of gadoxetic acid may shorten the liver T1 value in the hepatobiliary phase and increase the T1 reduction rate in patients with severe liver dysfunction, potentially overestimating liver function. PURPOSE: To verify the relationship between the gadoxetic acid dose and the liver T1 reduction rate and ΔR1 of the liver and spleen, and to clarify whether the ΔR1 of hepatocytes, corrected for the effect of gadoxetic acid dose, could be used as an index of functional liver reserve. MATERIAL AND METHODS: We enrolled 13 patients with normal liver function (NLF); and 18, 8, and 3 patients with Child-Pugh classes A (CPA), B (CPB), and C (CPC) who underwent gadoxetic acid-enhanced magnetic resonance imaging. Phase-sensitive inversion recovery sequence was performed before and at 15 min after injection and T1 maps were calculated. Liver and spleen ΔR1, liver T1 reduction rate, and the liver-to-spleen ΔR1 ratio were calculated. RESULTS: Only the liver-to-spleen ΔR1 ratio showed no correlation with gadoxetic acid dose in any group. The T1 reduction rate was not significantly different between the CPA and CPB + CPC groups. The liver-to-spleen ΔR1 ratio significantly differed between all groups. CONCLUSION: The liver and spleen ΔR1 was dependent on the dose of gadoxetic acid in severe liver dysfunction. The liver-to-spleen ΔR1 ratio improves the delineation of the CPA and CPB + CPC groups.

[Effects of Changing the Source-to-image Receptor Distance Parameter of Scattered X-ray Correction after Exposure on Image Quality in Portable Chest Radiography].

PURPOSE: To validate the effects of changing the source-to-image receptor distance (SID) parameter of scattered X-ray correction after exposure on the image quality in portable chest radiography. METHODS: The actual SID and tube current-time product (mAs) were varied such that the direct X-ray dose to a flat panel detector (FPD) remained constant. We created two groups as follows: Group A (with the SID parameter unchanged) and Group B (with the SID parameter changed to the actual SID after a phantom chest exposure). The image contrast ratio and standard deviation (SD) were measured on the chest radiographs for physical assessment. Observer studies were performed by seven radiological technologists. Scheffé's (Ura) paired comparison methods were performed with image contrast, noise, and overall assessment as the assessment items. Receiver operating characteristic (ROC) analysis for lung nodules was performed. RESULTS: The image contrast ratio and SD in Group A changed, whereas the changes in Group B were less than those in Group A for both these properties. The observer study with Scheffé's methods showed a statistically significant difference (p<0.05) for all assessment items in Group A but not in Group B. The ROC analysis did not indicate any statistically significant differences in either group. CONCLUSION: Changing the SID parameter of scattered X-ray correction after exposure can possibly maintain image contrast and noise in portable chest radiography if the actual SID changes.

[Feasibility of Cerebrovascular Reserve Assessment Using Stretched Exponential Model in Major Cerebral Artery Steno-occlusive Disease: Comparison with SPECT].

PURPOSE: To clarify whether diffusion-weighted imaging using stretched exponential model can assess cerebrovascular reserve (CVR) in patients with major cerebral artery steno-occlusive disease, we compared stretched exponential parameters and single-photon emission computed tomography (SPECT). METHODS: Twenty-nine patients with unilateral major cerebral artery steno-occlusive disease (25 men and 4 women; age, 69±11 years) were analyzed in this study. The patients were divided into three groups: normal CVR (CVR≥30%), moderate CVR (10%≤CVR<30%), and severe CVR (CVR<10%). The distributed diffusion coefficient (DDC) and heterogeneity index (α) from the stretched exponential model, apparent diffusion coefficient (ADC) from the monoexponential model, and CVR and resting cerebral blood flow (CBF) from SPECT were measured in the bilateral middle cerebral artery territories, and ipsilateral-to-contralateral ratios (rDDC, rα, rADC, and rCBF) were obtained. RESULTS: The rDDC values in severe CVR were significantly higher than those in normal CVR (P=0.003). The rDDC values were significantly negatively correlated with ipsilateral CVR (rho=-0.31, P=0.009). The rDDC values were not significantly correlated with rCBF (P=0.34). CONCLUSION: We have shown that elevated rDDC values are associated with impaired CVR. Our results suggest that diffusion-weighted imaging using stretched exponential model has a potential to evaluate hemodynamic impairment.

Improvement in the contrast-to-noise ratio and quantitative measurement of T1 and T2* values for carotid atherosclerotic plaque using multi-echo phase-sensitive inversion recovery.

Quantitative magnetic resonance imaging is required to accurately evaluate carotid plaque vulnerability. We prospectively evaluated the potential for fast quantitative black-blood carotid vessel wall imaging using a three-dimensional (3D) multi-echo phase-sensitive inversion recovery (mPSIR) sequence. Forty-nine patients with carotid atherosclerotic plaques were examined. Two-dimensional (2D) turbo spin-echo (TSE), 3D volumetric isotropic turbo spin-echo, and 3D mPSIR imaging were performed. The contrast-to-noise ratios (CNRs) between the carotid plaque and adjacent muscle were compared for the three imaging methods. The T1 and T2* values of the carotid plaques were measured using 3D mPSIR images. These values were compared with those of symptomatic and asymptomatic plaques. For carotid plaques with a signal intensity ratio ≥ 1.55, between the carotid plaque and adjacent muscle in 2D TSE images, the CNR of the mPSIR images was significantly higher than that of the other sequences. T1 values for symptomatic and asymptomatic plaques were 544.0 ± 258.0 and 569.1 ± 301.7, respectively. T2* values for symptomatic and asymptomatic plaques were 34.0 ± 33.0 and 21.8 ± 20.3 ms, respectively. There were no significant differences in the T1 and T2* values between the symptomatic and asymptomatic plaques. 3D mPSIR improves the CNR of T1-weighted images for carotid plaques and the adjacent muscle while simultaneously providing the T1 and T2* values of the carotid plaque. This improved CNR may be useful for assessing the vulnerability of carotid plaques.

Quantitative evaluation of focal liver lesions with T1 mapping using a phase-sensitive inversion recovery sequence on gadoxetic acid-enhanced MRI.

PURPOSE: To determine the usefulness of T1 values measured using a phase-sensitive inversion recovery (PSIR) sequence for the diagnosis of focal liver lesions. METHOD: The study enrolled 87 patients who underwent gadoxetic acid-enhanced magnetic resonance imaging (MRI) for assessment of 38 hepatocellular carcinomas, 33 hepatic hemangiomas, 30 metastatic liver tumors, and 14 hepatic cysts. PSIR was performed before and 15 min after contrast agent administration, and then the respective T1 values were measured and the T1 reduction rate was calculated. Wilcoxon matched-pairs signed-rank test was used to compare T1 values pre- and post-contrast administration in each tumor. The Kruskal-Wallis test and Dunn's post-hoc test were used to compare T1 values among all tumors pre- and post-contrast administration and the T1 reduction rate among all tumors. RESULTS: The T1 values measured before and after contrast enhancement were 1056 ± 292 ms and 724 ± 199 ms for hepatocellular carcinoma, 1757 ± 723 ms and 1033 ± 406 ms for metastatic liver tumor, 2524 ± 908 ms and 1071 ± 390 ms for hepatic hemangioma, and 3793 ± 207 ms and 3671 ± 241 ms for liver cysts, respectively. The T1 values obtained before and after contrast administration showed significant differences for all tumors except liver cysts (P < 0.0001). T1 reduction rate was not significantly different between hepatocellular carcinoma and metastatic liver tumor, but was significantly different among other tumors (P < 0.05). CONCLUSIONS: T1 mapping using the PSIR sequence is useful to differentiate focal liver lesions.

Comparison of Arterial Spin-labeling Perfusion Images at Different Spatial Normalization Methods Based on Voxel-based Statistical Analysis.

PURPOSE: Spatial normalization is a significant image pre-processing operation in statistical parametric mapping (SPM) analysis. The purpose of this study was to clarify the optimal method of spatial normalization for improving diagnostic accuracy in SPM analysis of arterial spin-labeling (ASL) perfusion images. METHODS: We evaluated the SPM results of five spatial normalization methods obtained by comparing patients with Alzheimer's disease or normal pressure hydrocephalus complicated with dementia and cognitively healthy subjects. We used the following methods: 3DT1-conventional based on spatial normalization using anatomical images; 3DT1-DARTEL based on spatial normalization with DARTEL using anatomical images; 3DT1-conventional template and 3DT1-DARTEL template, created by averaging cognitively healthy subjects spatially normalized using the above methods; and ASL-DARTEL template created by averaging cognitively healthy subjects spatially normalized with DARTEL using ASL images only. RESULTS: Our results showed that ASL-DARTEL template was small compared with the other two templates. Our SPM results obtained with ASL-DARTEL template method were inaccurate. Also, there were no significant differences between 3DT1-conventional and 3DT1-DARTEL template methods. In contrast, the 3DT1-DARTEL method showed higher detection sensitivity, and precise anatomical location. CONCLUSIONS: Our SPM results suggest that we should perform spatial normalization with DARTEL using anatomical images.