Performance of Machine Learning Methods Based on Multi-Sequence Textural Parameters Using Magnetic Resonance Imaging and Clinical Information to Differentiate Malignant and Benign Soft Tissue Tumors.

RATIONALE AND OBJECTIVES: To evaluate the performance of a machine learning method to differentiate malignant from benign soft tissue tumors based on textural features on multiparametric magnetic resonance imaging (mpMRI). MATERIALS AND METHODS: We enrolled 163 patients with soft tissue tumors whose diagnosis was pathologically proven (71 malignant, 92 benign). All patients underwent mpMRI. Twelve histographic and textural parameters were assessed on T1-weighted imaging (T1WI), T2-weighted imaging, diffusion-weighted imaging, apparent diffusion coefficient maps, and contrast-enhanced T1WI imaging. We compared mean signals of all sequences from the malignant and benign tumors using Welch's t-test. Prediction models were developed via a machine learning technique (support vector machine) using textural features of each sequence, clinical information (sex + age + tumor size), and the combined model incorporating all features. Areas under the receiver operating characteristic curves (AUCs) of these models were calculated using fivefold cross validation. RESULTS: The diagnostic ability of clinical information model (AUC 0.85) was not inferior to the model with textural features of each sequence (AUC 0.79-0.84). The combined model showed the highest diagnostic ability (AUC 0.89). The AUC of the combined model (0.89) was comparable to those of two board-certified radiologists (0.89 and 0.87). CONCLUSIONS: Machine learning methods based on textural features on mpMRI and clinical information offer adequate diagnostic performance to differentiate between malignant and benign soft tissue tumors.

Hybrid deep-learning-based denoising method for compressed sensing in pituitary MRI: comparison with the conventional wavelet-based denoising method.

OBJECTIVES: This study aimed to evaluate the efficacy of a combined wavelet and deep-learning reconstruction (DLR) method for under-sampled pituitary MRI. METHODS: This retrospective study included 28 consecutive patients who underwent under-sampled pituitary T2-weighted images (T2WI). Images were reconstructed using either the conventional wavelet denoising method (wavelet method) or the wavelet and DLR methods combined (hybrid DLR method) at five denoising levels. The signal-to-noise ratio (SNR) of the CSF, hypothalamic, and pituitary images and the contrast between structures were compared between the two image types. Noise quality, contrast, sharpness, artifacts, and overall image quality were evaluated by two board-certified radiologists. The quantitative and the qualitative analyses were performed with robust two-way repeated analyses of variance. RESULTS: Using the hybrid DLR method, the SNR of the CSF progressively increased as denoising levels increased. By contrast, with the wavelet method, the SNR of the CSF, hypothalamus, and pituitary did not increase at higher denoising levels. There was a significant main effect of denoising methods (p < 0.001) and denoising levels (p < 0.001), and an interaction between denoising methods and denoising levels (p < 0.001). For all five qualitative scores, there was a significant main effect of denoising methods (p < 0.001) and an interaction between denoising methods and denoising levels (p < 0.001). CONCLUSIONS: The hybrid DLR method can provide higher image quality for T2WI of the pituitary with compressed sensing (CS) than the wavelet method alone, especially at higher denoising levels. KEY POINTS: • The signal-to-noise ratios of cerebrospinal fluid progressively increased with the hybrid DLR method, with an increase in the denoising level for cerebrospinal fluid in pituitary T2WI with CS. • The signal-to-noise ratios of cerebrospinal fluid using the conventional wavelet method did not increase at higher denoising levels. • All qualitative scores of hybrid deep-learning reconstructions at all denoising levels were higher than those for the wavelet denoising method.

Virtual Monochromatic Image Quality from Dual-Layer Dual-Energy Computed Tomography for Detecting Brain Tumors.

OBJECTIVE: To evaluate the usefulness of virtual monochromatic images (VMIs) obtained using dual-layer dual-energy CT (DL-DECT) for evaluating brain tumors. MATERIALS AND METHODS: This retrospective study included 32 patients with brain tumors who had undergone non-contrast head CT using DL-DECT. Among them, 15 had glioblastoma (GBM), 7 had malignant lymphoma, 5 had high-grade glioma other than GBM, 3 had low-grade glioma, and 2 had metastatic tumors. Conventional polychromatic images and VMIs (40-200 keV at 10 keV intervals) were generated. We compared CT attenuation, image noise, contrast, and contrast-to-noise ratio (CNR) between tumor and white matter (WM) or grey matter (GM) between VMIs showing the highest CNR (optimized VMI) and conventional CT images using the paired t test. Two radiologists subjectively assessed the contrast, margin, noise, artifact, and diagnostic confidence of optimized VMIs and conventional images on a 4-point scale. RESULTS: The image noise of VMIs at all energy levels tested was significantly lower than that of conventional CT images (p < 0.05). The 40-keV VMIs yielded the best CNR. Furthermore, both contrast and CNR between the tumor and WM were significantly higher in the 40 keV images than in the conventional CT images (p < 0.001); however, the contrast and CNR between tumor and GM were not significantly different (p = 0.47 and p = 0.31, respectively). The subjective scores assigned to contrast, margin, and diagnostic confidence were significantly higher for 40 keV images than for conventional CT images (p < 0.01). CONCLUSION: In head CT for patients with brain tumors, compared with conventional CT images, 40 keV VMIs from DL-DECT yielded superior tumor contrast and diagnostic confidence, especially for brain tumors located in the WM.

Non-contrast renal MRA using multi-shot gradient echo EPI at 3-T MRI.

OBJECTIVES: The purpose of this study was to investigate the feasibility of non-contrast renal MRA using multi-shot gradient echo planar imaging (MSG-EPI) with a 3-T MRI system. METHODS: Seventeen healthy volunteers underwent non-contrast renal MRA using MSG-EPI and balanced steady-state free precession (b-SSFP) sequences on a 3-T MRI system. Two radiologists independently recorded the images' contrast, noise, sharpness, artifacts, and overall quality on 4-point scales. The signal-to-noise ratio (SNR) for the renal artery, the contrast ratio (CR) between the renal artery and erector spinae, and acquisition time were compared between the two sequences. RESULTS: The SNR and CR were significantly higher with MSG-EPI than with the b-SSFP sequence (17.80 ± 3.67 vs. 10.84 ± 2.86 and 0.77 ± 0.05 and 0.66 ± 0.09, respectively; p < 0.05), and the acquisition time was significantly lower (164.5 ± 34.0 vs. 261.5 ± 39.3 s, respectively; p < 0.05). There were significant differences in image contrast, noise, sharpness, artifacts, and overall image quality between the two sequences (p < 0.01). CONCLUSIONS: The MSG-EPI sequence is a promising technique that can shorten the scan time and improve the image quality of non-contrast renal MRA with a 3-T MRI system. KEY POINTS: • The multi-shot gradient echo planar imaging with an inversion pulse is a brand-new fast scan technique for an unenhanced renal MRA. • The image quality of multi-shot gradient echo planar imaging is better than that of b-SSFP for an unenhanced renal MRA.

Effects of Deep Learning Reconstruction Technique in High-Resolution Non-contrast Magnetic Resonance Coronary Angiography at a 3-Tesla Machine.

PURPOSE: To evaluate the effects of deep learning reconstruction (DLR) in qualitative and quantitative image quality of non-contrast magnetic resonance coronary angiography (MRCA). METHODS: Ten healthy volunteers underwent conventional MRCA (C-MRCA) and high-resolution (HR) MRCA on a 3T magnetic resonance imaging with a voxel size of 1.8 × 1.1 × 1.7 mm3 and 1.8 × 0.6 × 1.0 mm3, respectively, for C-MRCA and HR-MRCA. High-resolution magnetic resonance coronary angiography was also reconstructed with the DLR technique (DLR-HR-MRCA). We compared the contrast-to-noise ratio (CNR) and visual evaluation scores for vessel sharpness and traceability of proximal and distal coronary vessels on a 4-point scale among 3 image series. RESULTS: The vascular CNR value on the C-MRCA and the DLR-HR-MRCA was significantly higher than that on the HR-MRCA in the proximal and distal coronary arteries (13.9 ± 6.4, 11.3 ± 4.4, and 7.8 ± 2.6 for C-MRCA, DLR-HR-MRCA, and HR-MRCA, P < .05, respectively). Mean visual evaluation scores for the vessel sharpness and traceability of proximal and distal coronary vessels were significantly higher on the HR-DLR-MRCA than the C-MRCA (P < .05, respectively). CONCLUSION: Deep learning reconstruction significantly improved the CNR of coronary arteries on HR-MRCA, resulting in both higher visual image quality and better vessel traceability compared with C-MRCA.

A preliminary study of deep learning-based reconstruction specialized for denoising in high-frequency domain: usefulness in high-resolution three-dimensional magnetic resonance cisternography of the cerebellopontine angle.

PURPOSE: Deep learning-based reconstruction (DLR) has been developed to reduce image noise and increase the signal-to-noise ratio (SNR). We aimed to evaluate the efficacy of DLR for high spatial resolution (HR)-MR cisternography. METHODS: This retrospective study included 35 patients who underwent HR-MR cisternography. The images were reconstructed with or without DLR. The SNRs of the CSF and pons, contrast of the CSF and pons, and sharpness of the normal-side trigeminal nerve using full width at half maximum (FWHM) were compared between the two image types. Noise quality, sharpness, artifacts, and overall image quality of these two types of images were qualitatively scored. RESULTS: The SNRs of the CSF and pons were significantly higher with DLR than without DLR (CSF 21.81 ± 7.60 vs. 15.33 ± 4.03, p < 0.001; pons 5.96 ± 1.38 vs. 3.99 ± 0.48, p < 0.001). There were no significant differences in the contrast of the CSF and pons (p = 0.225) and sharpness of the normal-side trigeminal nerve using FWHM (p = 0.185) without and with DLR, respectively. Noise quality and the overall image quality were significantly higher with DLR than without DLR (noise quality 3.95 ± 0.19 vs. 2.53 ± 0.44, p < 0.001; overall image quality 3.97 ± 0.17 vs. 2.97 ± 0.12, p < 0.001). There were no significant differences in sharpness (p = 0.371) and artifacts (p = 1) without and with DLR. CONCLUSION: DLR can improve the image quality of HR-MR cisternography by reducing image noise without sacrificing contrast or sharpness.

Diagnostic Performance of 123I-FPCIT SPECT Specific Binding Ratio in Progressive Supranuclear Palsy: Use of Core Clinical Features and MRI for Comparison.

OBJECTIVE. Progressive supranuclear palsy (PSP) is listed as a core clinical feature in the Movement Disorder Society 2017 criteria, along with ocular motor dysfunction, postural instability, akinesia, and cognitive dysfunction. Imaging evidence shows predominant mid-brain atrophy and postsynaptic striatal dopaminergic degeneration as two supportive features. The purpose of this study was to investigate the diagnostic performance of 123I-N- ω-fluoropropyl-2ß-carbomethoxy-3ß-(4-iodophenyl) nortropane (123I-FP-CIT) SPECT by comparing it with evaluation of core clinical features and MRI in the diagnosis of PSP. MATERIALS AND METHODS. The study included 53 patients with clinically suspected PSP who had undergone 123I-FP-CIT SPECT and MRI examinations. MR parkinsonism index (MRPI) was used as the MRI index. For the 123I-FP-CIT SPECT index, specific binding ratio (SBR) was calculated as the average of the right and left SBRs. RESULTS. In regard to core clinical features, ocular motor dysfunction was present in 15 of 20 (75.0%) patients with the diagnosis of probable PSP (p < 0.0001). Calculation of the diagnostic performance of the imaging parameters showed that MRPI (cutoff > 11.6) had 85.0% sensitivity, 100% specificity, and 94.3% accuracy. SBR (cutoff < 3.7) had 95.0% sensitivity, 36.4% specificity, and 58.5% accuracy. CONCLUSION. Iodine-123-labeled FP-CIT SPECT has high sensitivity, and MRI has high specificity in the diagnosis of PSP. Because these tools have complementary roles, reach ing a more confident clinical diagnosis of PSP may be possible when both are used.

A primer for understanding radiology articles about machine learning and deep learning.

The application of machine learning and deep learning in the field of imaging is rapidly growing. Although the principles of machine and deep learning are unfamiliar to the majority of clinicians, the basics are not so complicated. One of the major issues is that commentaries written by experts are difficult to understand, and are not primarily written for clinicians. The purpose of this article was to describe the different concepts behind machine learning, radiomics, and deep learning to make clinicians more familiar with these techniques.

Trends in Diagnostic Imaging of Cardiac Amyloidosis: Emerging Knowledge and Concepts.

Cardiac amyloidosis (CA) has long been recognized as a rare disease. However, recent advances in cardiac imaging have led to increased identification of hidden CA in patients diagnosed with heart failure. This shift suggests that the actual incidence of CA is underestimated. The prognosis of CA is generally poor, especially in patients with advanced heart failure. However, recent developments in therapeutic interventions have improved the survival of patients with CA. An early diagnosis and interventions involving effective therapies are essential contributors to improved prognoses. Recent noninvasive diagnostic imaging modalities such as echocardiography, cardiac MRI, and nuclear imaging have facilitated the precise and early diagnosis of CA and enabled the initiation of appropriate management. The authors present an updated review of the clinical features of CA, including a discussion of current trends in noninvasive diagnostic imaging. ©RSNA, 2020.

Adrenal Adenomas versus Metastases: Diagnostic Performance of Dual-Energy Spectral CT Virtual Noncontrast Imaging and Iodine Maps.

Background Dual-energy CT allows virtual noncontrast (VNC) attenuation and iodine density measurements from contrast material-enhanced examination, potentially enabling adrenal lesion characterization. However, data regarding diagnostic performance remain limited, and combined diagnostic values have never been investigated. Purpose To determine whether VNC attenuation, iodine density, and combination of the two allow reliable differentiation between adrenal adenomas and metastases. Materials and Methods This retrospective study included patients with adrenal lesions who underwent unenhanced and portal venous phase dual-energy CT between January 2017 and December 2018. Unenhanced, contrast-enhanced, and VNC attenuation, as well as iodine density, were measured for each lesion. Agreement between unenhanced and VNC attenuation was assessed by using Wilcoxon rank-sum test, Pearson correlation coefficient, and Bland-Altman plot. The ratio of iodine density to VNC attenuation was calculated for lesions with positive VNC attenuation. Each parameter was compared between adenomas and metastases; diagnostic performance was evaluated by using the area under the receiver operating characteristic curve (AUC) with sensitivity and specificity. Results A total of 149 patients (mean age, 65 years ± 13 [standard deviation]; 89 men; 98 patients with 104 adenomas; 51 patients with 56 metastases) were evaluated. VNC attenuation showed strong positive correlation with unenhanced attenuation (r = 0.92) but resulted in overestimates of adenoma attenuation (mean bias, +11 HU; P < .001) and was less sensitive (P = .03) in the diagnosis of adenomas compared with unenhanced attenuation (sensitivity of 79% [81 of 102] [95% confidence interval {CI}: 70%, 87%] and specificity of 95% [53 of 56] [95% CI: 85%, 99%] versus sensitivity of 85% [87 of 102] [95% CI: 77%, 92%] and specificity of 96% [54 of 56] [95% CI: 88%, 100%], with thresholds of ≤29 HU and ≤22 HU, respectively). Contrast-enhanced attenuation had no discriminatory ability (AUC, 0.54; 95% CI: 0.45, 0.62). Iodine density yielded moderate performance (sensitivity of 78% [80 of 102] [95% CI: 69%, 86%] and specificity of 71% [40 of 56] [95% CI: 58%, 83%], with a threshold of ≥1.82 mg/mL). The iodine-to-VNC ratio was higher in adenomas than in metastases (mean, 14.5 vs 4.6; P < .001), with sensitivity of 95% (97 of 102; 95% CI: 89%, 98%) and specificity of 95% (53 of 56; 95% CI: 85%, 99%), with a threshold of 6.7 or greater. Conclusion Contrast-enhanced dual-energy CT during the portal venous phase enabled accurate differentiation between adrenal adenomas and metastases by combining virtual noncontrast attenuation and iodine density. Virtual noncontrast imaging alone led to overestimates of adenoma attenuation, and iodine density alone had limited discriminatory utility. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Hindman and Megibow in this issue.

A diagnostic strategy for Lewy body disease using DAT-SPECT, MIBG and Combined index.

PURPOSE: The functional imaging methods widely used for the diagnosis of Lewy body disease (LBD) are 123I-N-ω-fluoropropyl-2ß-carbomethoxy-3ß-(4-iodophenyl) nortropan (FP-CIT) with dopamine transporter single photon emission computed tomography (DAT-SPECT) and 123I-iodobenzylguanidine (MIBG) myocardial scintigraphy. The aim of this study was to determine whether DAT-SPECT or 123I-MIBG myocardial scintigraphy should be examined first and to evaluate whether the combined use of DAT-SPECT and MIBG myocardial scintigraphy is superior to using either modality alone for diagnosing suspected LBD. METHODS: In this retrospective study, a total of 117 patients suspected of having LBD underwent DAT-SPECT imaging followed by MIBG myocardial scintigraphy. The delayed heart-to-mediastinum (H/M) ratio of MIBG scintigraphy, and the specific binding ratio (SBR) of DAT-SPECT imaging, and Combined index (defined as SBR mean × H/M in the delayed phase) were used as semi-quantitative measures. The diagnostic ability was evaluated using these indexes. RESULTS: The sensitivity, specificity, and accuracy of diagnosing Lewy body disease were 59.6%, 71.4%, and 67.5% by SBR mean of DAT-SPECT, 85.1%, 91.4%, and 88.9% by delayed H/M ratio of MIBG myocardial scintigraphy, 76.6%, 74.3%, and 75.2% by Combined index, respectively. CONCLUSION: In the diagnosis of LBD, DAT-SPECT, MIBG myocardial scintigraphy, and Combined index may be reliable indices. In particular, MIBG myocardial scintigraphy was the specific modality for LBD diagnosis. Understanding the effectiveness and limits of DAT-SPECT and MIBG myocardial scintigraphy and using both properly will lead to a more accurate diagnosis and better treatment.

Impact of hydrostatic pressure on fractional flow reserve: in vivo experimental study of anatomical height difference of coronary arteries.

BACKGROUND: Although pressure equalization of the sensor-tipped guidewire and systemic pressure is mandatory in measuring fractional flow reserve (FFR), pressure in the distal artery (Pd) with wire advancement can be influenced by hydrostatic pressure related to the height difference between the catheter tip and the distal pressure sensor. We therefore analyzed the impact of hydrostatic pressure on FFR in vivo by modification of the height difference. METHODS: To reveal the anatomical height difference in human coronary arteries, measurement was performed during computed tomography angiography (CTA) of five consecutive patients. Utilizing the healthy coronary arteries of female swine, height difference diversity was reproduced by body rotation and vertical inclination. FFR measurements were performed during maximum hyperemia with adenosine. The height difference was calculated fluoroscopically with a contrast medium-filled balloon for reference. RESULTS: In human coronary CTA, height averages from the ostium in the left anterior descending artery (34.6 mm) were significantly higher than in the left circumflex (-15.5 mm, p = 0.008) and right coronary arteries (-2.3 mm, p = 0.008). In our swine model, reproduced height variation ranged from -7.2 cm to +6.5 cm. Mean FFR was significantly lower in positive sensor height and higher in negative sensor height compared to the reference height. Linear regression analyses revealed significant correlations between height difference and FFR, observed among all coronary arteries, as well as between the height difference and Pd-aortic pressure mismatch. Subtracting 0.622 mmHg/cm height difference from Pd could correct the expected hydrostatic pressure influence. CONCLUSION: Hydrostatic pressure variation resulting from sensor height influenced FFR values might affect interpretation during FFR assessment.

Diagnostic Performance of Dual-Layer Computed Tomography for Deep Vein Thrombosis in Indirect Computed Tomography Venography.

BACKGROUND: The aim of this study was to evaluate the quality and diagnostic performance of virtual monochromatic images (VMI) obtained with dual-layer dual-energy computed tomography (DL-DECT) during indirect CT venography (CTV) for deep vein thrombosis (DVT).Methods and Results:This retrospective study was approved by the Institutional Review Board, which waived the requirement for informed consent. We retrospectively enrolled 45 patients who underwent CTV with DL-DECT, and VMI were retrospectively generated. We compared the venous attenuation, noise, contrast, and contrast-to-noise ratio (CNR) between VMI with the highest CNR and conventional CT on paired t-test. Furthermore, we compared the pooled area under the curve (AUC) of each technique with Delong's test in 34 patients who underwent color Doppler ultrasonography. The 40-keV VMI had the best CNR. The noise was significantly lower on 40-keV (9.7±2.5 HU) than on 120-kVp VMI (10.5±2.5 HU; P<0.01). The contrast (120 kVp, 38.2±15.3 HU vs. 40 keV, 131.6±43.6 HU) and CNR (120 kVp, 3.8±1.7 vs. 40 keV, 14.4±6.1) were significantly higher in 40-keV VMI than in 120-kVp VMI (P<0.01). Furthermore, the pooled AUC was significantly higher for 40-keV (0.84) than for 120-kVp VMI (0.78; P=0.03). CONCLUSIONS: In indirect CTV, 40-keV VMI obtained with DL-DECT offers better image quality and diagnostic performance for DVT than conventional CT.

Machine Learning to Predict the Rapid Growth of Small Abdominal Aortic Aneurysm.

OBJECTIVE: The purpose of this study was to determine whether computed tomography (CT) angiography with machine learning (ML) can be used to predict the rapid growth of abdominal aortic aneurysm (AAA). MATERIALS AND METHODS: This retrospective study was approved by our institutional review board. Fifty consecutive patients (45 men, 5 women, 73.5 years) with small AAA (38.5 ± 6.2 mm) had undergone CT angiography. To be included, patients required at least 2 CT scans a minimum of 6 months apart. Abdominal aortic aneurysm growth, estimated by change per year, was compared between patients with baseline infrarenal aortic minor axis. For each axial image, major axis of AAA, minor axis of AAA, major axis of lumen without intraluminal thrombi (ILT), minor axis of lumen without ILT, AAA area, lumen area without ILT, ILT area, maximum ILT area, and maximum ILT thickness were measured. We developed a prediction model using an ML method (to predict expansion >4 mm/y) and calculated the area under the receiver operating characteristic curve of this model via 10-fold cross-validation. RESULTS: The median aneurysm expansion was 3.0 mm/y. Major axis of AAA and AAA area correlated significantly with future AAA expansion (r = 0.472, 0.416 all P < 0.01). Machine learning and major axis of AAA were a strong predictor of significant AAA expansion (>4 mm/y) (area under the receiver operating characteristic curve were 0.86 and 0.78). CONCLUSIONS: Machine learning is an effective method for the prediction of expansion risk of AAA. Abdominal aortic aneurysm area and major axis of AAA are the important factors to reflect AAA expansion.

Usefulness of Virtual Monochromatic Dual-Layer Computed Tomographic Imaging for Breast Carcinoma.

OBJECTIVE: This study aimed to evaluate virtual monochromatic images (VMIs) obtained using dual-layer dual-energy computed tomography (CT) for breast carcinoma. METHODS: We retrospectively enrolled 28 patients with breast cancer who were pathologically diagnosed using dual-layer dual-energy CT. Virtual monochromatic images (40-200 keV) were generated. We compared CT number, image noise, contrast, and contrast-to-noise ratio (CNR) between VMIs with the highest CNR and conventional CT images. We performed qualitative image analysis between VMIs at optimized energy and conventional CT images. RESULTS: Image noise of VMIs was not significantly different from that of the conventional CT images. As the x-ray energy decreased, CNR increased. The 40-keV VMIs were highest CNR and higher than that of the conventional CT images. In qualitative image analysis, the 40-keV images were significantly higher than conventional CT images. CONCLUSION: Both qualitative and quantitative analyses showed that the image quality of VMIs at 40 keV was significantly higher than that of conventional CT images.

Prospective Comparison of 70-kVp Single-Energy CT versus Dual-Energy CT: Which is More Suitable for CT Angiography with Low Contrast Media Dosage?

RATIONALE AND OBJECTIVES: To compare the objective and subjective image qualities between single-energy computed tomography (CT) at 70 kVp and virtual monoenergetic imaging (VMI) of dual-source dual-energy CT for CT angiography with 180 mgI/kg. MATERIALS AND METHODS: Total 63 patients scanned with 180 mgI/kg were randomly divided into two groups: Group A (32 patients) underwent CT angiography at 70-kVp, and Group B (31 patients) underwent dual-energy CT. VMI sets were generated at 10-keV increments between 40 and 100 keV. We calculated aortic attenuation, contrast-to-noise-ratio (CNR), signal-to-noise-ratio, figure of merit of CNR, and effective dose for each protocol. Three radiologists scored overall image quality and various arteries' visibility using a four-point scale. Quantitative and qualitative comparisons between 70 kVp and VMI with the highest CNR were performed with the two-tailed t test or Kruskal-Wallis test. RESULTS: The 40-keV images offered the highest CNR among VMIs. Aortic attenuation at 70 kVp was significantly lower than that at 40 keV (p < 0.001). However, the signal-to-noise-ratio, CNR, and figure of merit of CNR were significantly higher at 70 kVp than those at 40-keV (p < 0.001, p < 0.05, and p < 0.05, respectively). The effective dose of each group was almost equal. The qualitative visibility scores for various arteries, except the ascending and upper-abdominal aorta, were also better at 70 kVp than those at 40 keV. CONCLUSION: Aortic attenuation at 70 kVp with 180 mg I/kg was lower than that of VMI at 40 keV, and the objective and subjective image qualities were higher at 70 kVp than those at 40 keV.

Deep Learning Based Noise Reduction for Brain MR Imaging: Tests on Phantoms and Healthy Volunteers.

PURPOSE: To test whether our proposed denoising approach with deep learning-based reconstruction (dDLR) can effectively denoise brain MR images. METHODS: In an initial experimental study, we obtained brain images from five volunteers and added different artificial noise levels. Denoising was applied to the modified images using a denoising convolutional neural network (DnCNN), a shrinkage convolutional neural network (SCNN), and dDLR. Using these brain MR images, we compared the structural similarity (SSIM) index and peak signal-to-noise ratio (PSNR) between the three denoising methods. Two neuroradiologists assessed the image quality of the three types of images. In the clinical study, we evaluated the denoising effect of dDLR in brain images with different levels of actual noise such as thermal noise. Specifically, we obtained 2D-T2-weighted image, 2D-fluid-attenuated inversion recovery (FLAIR) and 3D-magnetization-prepared rapid acquisition with gradient echo (MPRAGE) from 15 healthy volunteers at two different settings for the number of image acquisitions (NAQ): NAQ2 and NAQ5. We reconstructed dDLR-processed NAQ2 from NAQ2, then compared with SSIM and PSNR. Two neuroradiologists separately assessed the image quality of NAQ5, NAQ2 and dDLR-NAQ2. Statistical analysis was performed in the experimental and clinical study. In the clinical study, the inter-observer agreement was also assessed. RESULTS: In the experimental study, PSNR and SSIM for dDLR were statistically higher than those of DnCNN and SCNN (P < 0.001). The image quality of dDLR was also superior to DnCNN and SCNN. In the clinical study, dDLR-NAQ2 was significantly better than NAQ2 images for SSIM and PSNR in all three sequences (P < 0.05), except for PSNR in FLAIR. For all qualitative items, dDLR-NAQ2 had equivalent or better image quality than NAQ5, and superior quality to that of NAQ2 (P < 0.05), for all criteria except artifact. The inter-observer agreement ranged from substantial to near perfect. CONCLUSION: dDLR reduces image noise while preserving image quality on brain MR images.

Hybrid of Compressed Sensing and Parallel Imaging Applied to Three-dimensional Isotropic T2-weighted Turbo Spin-echo MR Imaging of the Lumbar Spine.

PURPOSE: The hybrid compressed sensing (hybrid-CS) technique can shorten the acquisition time compared with the sensitivity encoding (SENSE) technique in lumbar MRI. To evaluate the feasibility of a hybrid-CS technique in comparison with 3D isotropic T2-weighted turbo spin-echo (3D volume isotropic turbo spin-echo acquisition [VISTA]) MRI of the lumbar spine. MATERIALS AND METHODS: The Institutional Review Board approved this study and informed consent was obtained from participants prior to study entry. Sixteen healthy volunteers underwent lumbar spine 3D VISTA with conventional parallel imaging for SENSE and hybrid-CS at 3T. We recorded the image acquisition times of SENSE and hybrid-CS. We compared the signal-to-noise ratio (SNR) in spine, cerebrospinal fluid (CSF), lumbar disc, epidural fat, and erector spinae muscle, and the contrast of spine, CSF, and disc, and performed qualitative image analysis assessment, between the two image sequences. RESULTS: The image acquisition time for hybrid-CS was 39.2% shorter than that of SENSE (218.4/358.8 s). The contrast of CSF and SNR of the spine was significantly higher with hybrid-CS than with SENSE (P < 0.05). The SNR of the disc and muscle was significantly higher with SENSE than with hybrid-CS (P < 0.05). There were no significant differences in the contrast of spine, disc, and fat, and SNR of CSF and fat between hybrid-CS and SENSE. There were no significant differences in the qualitative evaluation between hybrid-CS and SENSE. CONCLUSION: Compared with SENSE, hybrid-CS for 3D VISTA can shorten image acquisition time without sacrificing image quality.

Myocardial extracellular volume quantification in cardiac CT: comparison of the effects of two different iterative reconstruction algorithms with MRI as a reference standard.

OBJECTIVES: To compare the effects of hybrid iterative reconstruction (HIR) and model-based iterative reconstruction (MBIR) that incorporates a beam-hardening model for myocardial extracellular volume (ECV) quantification by cardiac CT using MRI as a reference standard. METHODS: In this retrospective study, a total of 34 patients were evaluated using cardiac CT and MRI. Paired CT image sets were created using HIR and MBIR with a beam-hardening model. We calculated mean absolute differences and correlations between the global mid-ventricular ECV derived from CT and MRI via Pearson correlation analysis. In addition, we performed qualitative analysis of image noise and beam-hardening artifacts on postcontrast images using a four-point scale: 1 = extensive, 2 = strong, 3 = mild, and 4 = minimal. RESULTS: The mean absolute difference between the ECV derived from CT and MRI for MBIR was significantly smaller than that for HIR (MBIR 3.74 ± 3.59%; HIR 4.95 ± 3.48%, p = 0.034). MBIR improved the correlation between the ECV derived from CT and MRI when compared with HIR (MBIR, r = 0.60, p < 0.001; HIR, r = 0.47, p = 0.006). In qualitative analysis, MBIR significantly reduced image noise and beam-hardening artifacts when compared with HIR ([image noise, MBIR 3.4 ± 0.7; HIR 2.1 ± 0.8, p < 0.001], [beam-hardening artifacts, MBIR 3.8 ± 0.4; HIR 2.6 ± 1.0, p < 0.001]). CONCLUSIONS: MBIR with a beam-hardening model effectively reduced image noise and beam-hardening artifacts and improved myocardial ECV quantification when compared with HIR using MRI as a reference standard. KEY POINTS: • MBIR with a beam-hardening model effectively reduced image noise and beam-hardening artifacts. • The mean absolute difference between the global mid-ventricular ECV derived from CT and MRI for MBIR was significantly smaller than that for conventional HIR. • MBIR provided more accurate myocardial CT number and improved ECV quantification when compared with HIR.

Clinical usefulness of quantification of myocardial blood flow and flow reserve using CZT-SPECT for detecting coronary artery disease in patients with normal stress perfusion imaging.

BACKGROUND: Relative myocardial perfusion imaging can misdiagnose "balanced" ischemia caused by coronary artery disease (CAD). We assessed the feasibility of myocardial blood flow (MBF) and myocardial perfusion reserve (MPR) using dynamic single-photon emission computed tomography (SPECT) with a cadmium-zinc-telluride (CZT) camera for estimating underlying CAD in patients with normal stress myocardial perfusion SPECT (MPS). METHODS: 125 patients with normal stress MPS (summed stress score ≤3) were enrolled. All patients underwent coronary angiography (CAG) and stress/rest 201Tl dynamic SPECT for MBF and MPR calculation. The diagnostic accuracy of both these quantitative values and other clinical risk factors for predicting occult CAD were validated by CAG. RESULTS: MPR was 2.85 in patients with no CAD, 2.47 with 1-, 1.98 with 2-, and 1.76 with 3-vessel CAD. The patient's age, morbidity of diabetes mellitus (DM), chronic kidney disease (CKD), stress MBF, and MPR were significantly associated with the presence of CAD (age, p=0.02; DM, p=0.005; CKD, p=0.005; creatinine level, p= 0.012, stress MBF, p=0.019, and MPR, p<0.001). Independent predictors in the multivariate regression analysis were as follows: DM, p=0.011, CKD, p=0.028, and MPR, p<0.001. The combined index was calculated from three independent predictors. Area under the receiver operating characteristic curve was 0.75 for MPR and 0.81 for the combined index. To identify CAD, sensitivity, and specificity for MPR were 77% and 66%, and for the combined index they were 79% and 66%, respectively. CONCLUSION: Quantification of MPR and MBF using dynamic SPECT with a CZT camera can be useful to identify balanced ischemia caused by occult CAD in patients with normal stress MPS findings.