Clinical Impact of Deep Learning Reconstruction in MRI.

Deep learning has been recognized as a paradigm-shifting tool in radiology. Deep learning reconstruction (DLR) has recently emerged as a technology used in the image reconstruction process of MRI, which is an essential procedure in generating MR images. Denoising, which is the first DLR application to be realized in commercial MRI scanners, improves signal-to-noise ratio. When applied to lower magnetic field-strength scanners, the signal-to-noise ratio can be increased without extending the imaging time, and image quality is comparable to that of higher-field-strength scanners. Shorter imaging times decrease patient discomfort and reduce MRI scanner running costs. The incorporation of DLR into accelerated acquisition imaging techniques, such as parallel imaging or compressed sensing, shortens the reconstruction time. DLR is based on supervised learning using convolutional layers and is divided into the following three categories: image domain, k-space learning, and direct mapping types. Various studies have reported other derivatives of DLR, and several have shown the feasibility of DLR in clinical practice. Although DLR efficiently reduces Gaussian noise from MR images, denoising makes image artifacts more prominent, and a solution to this problem is desired. Depending on the training of the convolutional neural network, DLR may change the imaging features of lesions and obscure small lesions. Therefore, radiologists may need to adopt the habit of questioning whether any information has been lost on images that appear clean. ©RSNA, 2023 Quiz questions for this article are available in the supplemental material.

Comparison of 1.5 T and 3 T magnetic resonance angiography for detecting cerebral aneurysms using deep learning-based computer-assisted detection software.

PURPOSE: To compare the diagnostic performance of 1.5 T versus 3 T magnetic resonance angiography (MRA) for detecting cerebral aneurysms with clinically available deep learning-based computer-assisted detection software (EIRL aneurysm® [EIRL_an]), which has been approved by the Japanese Pharmaceuticals and Medical Devices Agency. We also sought to analyze the causes of potential false positives. METHODS: In this single-center, retrospective study, we evaluated the MRA scans of 90 patients who underwent head MRA (1.5 T and 3 T in 45 patients each) in clinical practice. Overall, 51 patients had 70 aneurysms. We used MRI from a vendor not included in the dataset used to create the EIRL_an algorithm. Two radiologists determined the ground truth, the accuracy of the candidates noted by EIRL_an, and the causes of false positives. The sensitivity, number of false positives per case (FPs/case), and the causes of false positives were compared between 1.5 T and 3 T MRA. Pearson's χ2 test, Fisher's exact test, and the Mann‒Whitney U test were used for the statistical analyses as appropriate. RESULTS: The sensitivity was high for 1.5 T and 3 T MRA (0.875‒1), but the number of FPs/case was significantly higher with 3 T MRA (1.511 vs. 2.578, p < 0.001). The most common causes of false positives (descending order) were the origin/bifurcation of vessels/branches, flow-related artifacts, and atherosclerosis and were similar between 1.5 T and 3 T MRA. CONCLUSION: EIRL_an detected significantly more false-positive lesions with 3 T than with 1.5 T MRA in this external validation study. Our data may help physicians with limited experience with MRA to correctly diagnose aneurysms using EIRL_an.

Acceleration of knee magnetic resonance imaging using a combination of compressed sensing and commercially available deep learning reconstruction: a preliminary study.

PURPOSE: To evaluate whether deep learning reconstruction (DLR) accelerates the acquisition of 1.5-T magnetic resonance imaging (MRI) knee data without image deterioration. MATERIALS AND METHODS: Twenty-one healthy volunteers underwent MRI of the right knee on a 1.5-T MRI scanner. Proton-density-weighted images with one or four numbers of signal averages (NSAs) were obtained via compressed sensing, and DLR was applied to the images with 1 NSA to obtain 1NSA-DLR images. The 1NSA-DLR and 4NSA images were compared objectively (by deriving the signal-to-noise ratios of the lateral and the medial menisci and the contrast-to-noise ratios of the lateral and the medial menisci and articular cartilages) and subjectively (in terms of the visibility of the anterior cruciate ligament, the medial collateral ligament, the medial and lateral menisci, and bone) and in terms of image noise, artifacts, and overall diagnostic acceptability. The paired t-test and Wilcoxon signed-rank test were used for statistical analyses. RESULTS: The 1NSA-DLR images were obtained within 100 s. The signal-to-noise ratios (lateral: 3.27 ± 0.30 vs. 1.90 ± 0.13, medial: 2.71 ± 0.24 vs. 1.80 ± 0.15, both p < 0.001) and contrast-to-noise ratios (lateral: 2.61 ± 0.51 vs. 2.18 ± 0.58, medial 2.19 ± 0.32 vs. 1.97 ± 0.36, both p < 0.001) were significantly higher for 1NSA-DLR than 4NSA images. Subjectively, all anatomical structures (except bone) were significantly clearer on the 1NSA-DLR than on the 4NSA images. Also, in the former images, the noise was lower, and the overall diagnostic acceptability was higher. CONCLUSION: Compared with the 4NSA images, the 1NSA-DLR images exhibited less noise, higher overall image quality, and allowed more precise visualization of the menisci and ligaments.

Deep learning reconstruction for 1.5 T cervical spine MRI: effect on interobserver agreement in the evaluation of degenerative changes.

OBJECTIVES: To investigate whether deep learning reconstruction (DLR) provides improved cervical spine MR images using a 1.5 T unit in the evaluation of degenerative changes without increasing imaging time. METHODS: This study included 21 volunteers (age 42.4 ± 11.9 years; 17 males) who underwent 1.5 T cervical spine sagittal T2-weighted MRI. From the imaging data with number of acquisitions (NAQ) of 1 or 2, images were reconstructed with DLR (NAQ1-DLR) and without DLR (NAQ1) or without DLR (NAQ2), respectively. Two readers evaluated the images for depiction of structures, artifacts, noise, overall image quality, spinal canal stenosis, and neuroforaminal stenosis. The two readers read studies blinded and randomly. Values were compared between NAQ1-DLR and NAQ1 and between NAQ1-DLR and NAQ2 using the Wilcoxon signed-rank test. RESULTS: The analyses showed significantly better results for NAQ1-DLR compared with NAQ1 and NAQ2 (p < 0.023), except for depiction of disc and foramina by one reader and artifacts by both readers in the comparison between NAQ1-DLR and NAQ2. Interobserver agreements (Cohen's weighted kappa [97.5% confidence interval]) in the evaluation of spinal canal stenosis for NAQ1-DLR/NAQ1/NAQ2 were 0.874 (0.866-0.883)/0.778 (0.767-0.789)/0.818 (0.809-0.827), respectively, and those in the evaluation of neuroforaminal stenosis were 0.878 (0.872-0.883)/0.855 (0.849-0.860)/0.852 (0.845-0.860), respectively. CONCLUSIONS: DLR improved the 1.5 T cervical spine MR images in the evaluation of degenerative spine changes. KEY POINTS: • Two radiologists demonstrated that deep learning reconstruction reduced the noise in cervical spine sagittal T2-weighted MR images obtained using a 1.5 T unit. • Reduced noise in deep learning reconstruction images resulted in a clearer depiction of structures, such as the spinal cord, vertebrae, and zygapophyseal joint. • Interobserver agreement in the evaluation of spinal canal stenosis and foraminal stenosis on cervical spine MR images was significantly improved using deep learning reconstruction (0.874 and 0.878, respectively) versus without deep learning (0.778-0.818 and 0.852-0.855, respectively).

Radiomics with 3-dimensional magnetic resonance fingerprinting: influence of dictionary design on repeatability and reproducibility of radiomic features.

OBJECTIVES: We aimed to investigate the influence of magnetic resonance fingerprinting (MRF) dictionary design on radiomic features using in vivo human brain scans. METHODS: Scan-rescans of three-dimensional MRF and conventional T1-weighted imaging were performed on 21 healthy volunteers (9 males and 12 females; mean age, 41.3 ± 14.6 years; age range, 22-72 years). Five patients with multiple sclerosis (3 males and 2 females; mean age, 41.2 ± 7.3 years; age range, 32-53 years) were also included. MRF data were reconstructed using various dictionaries with different step sizes. First- and second-order radiomic features were extracted from each dataset. Intra-dictionary repeatability and inter-dictionary reproducibility were evaluated using intraclass correlation coefficients (ICCs). Features with ICCs > 0.90 were considered acceptable. Relative changes were calculated to assess inter-dictionary biases. RESULTS: The overall scan-rescan ICCs of MRF-based radiomics ranged from 0.86 to 0.95, depending on dictionary step size. No significant differences were observed in the overall scan-rescan repeatability of MRF-based radiomic features and conventional T1-weighted imaging (p = 1.00). Intra-dictionary repeatability was insensitive to dictionary step size differences. MRF-based radiomic features varied among dictionaries (overall ICC for inter-dictionary reproducibility, 0.62-0.99), especially when step sizes were large. First-order and gray level co-occurrence matrix features were the most reproducible feature classes among different step size dictionaries. T1 map-derived radiomic features provided higher repeatability and reproducibility among dictionaries than those obtained with T2 maps. CONCLUSION: MRF-based radiomic features are highly repeatable in various dictionary step sizes. Caution is warranted when performing MRF-based radiomics using datasets containing maps generated from different dictionaries. KEY POINTS: • MRF-based radiomic features are highly repeatable in various dictionary step sizes. • Use of different MRF dictionaries may result in variable radiomic features, even when the same MRF acquisition data are used. • Caution is needed when performing radiomic analysis using data reconstructed from different dictionaries.

Deep learning reconstruction for the evaluation of neuroforaminal stenosis using 1.5T cervical spine MRI: comparison with 3T MRI without deep learning reconstruction.

PURPOSE: To compare image quality and interobserver agreement in evaluations of neuroforaminal stenosis between 1.5T cervical spine magnetic resonance imaging (MRI) with deep learning reconstruction (DLR) and 3T MRI without DLR. METHODS: In this prospective study, 21 volunteers (mean age: 42.4 ± 11.9 years; 17 males) underwent cervical spine T2-weighted sagittal 1.5T and 3T MRI on the same day. The 1.5T and 3T MRI data were used to reconstruct images with (1.5T-DLR) and without (3T-nonDLR) DLR, respectively. Regions of interest were marked on the spinal cord to calculate non-uniformity (NU; standard deviation/signal intensity × 100), as an indicator of image noise. Two blinded radiologists evaluated the images in terms of the depiction of structures, artifacts, noise, overall image quality, and neuroforaminal stenosis. The NU value and the subjective image quality scores were compared between 1.5T-DLR and 3T-nonDLR using the Wilcoxon signed-rank test. Interobserver agreement in evaluations of neuroforaminal stenosis for 1.5T-DLR and 3T-nonDLR was evaluated using Cohen's weighted kappa analysis. RESULTS: The NU value for 1.5T-DLR was 8.4, which was significantly better than that for 3T-nonDLR (10.3; p < 0.001). Subjective image scores were significantly better for 1.5T-DLR than 3T-nonDLR images (p < 0.037). Interobserver agreement (95% confidence intervals) in the evaluations of neuroforaminal stenosis was significantly superior for 1.5T-DLR (0.920 [0.916-0.924]) than 3T-nonDLR (0.894 [0.889-0.898]). CONCLUSION: By using DLR, image quality and interobserver agreement in evaluations of neuroforaminal stenosis on 1.5T cervical spine MRI could be improved compared to 3T MRI without DLR.

[An improved driving ability after attention function training for safe driving via use of a day rehabilitation service in a patient with cerebral infarction].

In recent years, the assessment of and support for the safety of driving for people with higher brain dysfunction to allow them to resume car driving have become issues to be addressed in Japan. It is difficult to determine whether or not people with higher brain dysfunction may safely resume car driving; in addition, methods of supporting this resumption have not been established. To support people with higher brain dysfunction and allow them to live at home in areas where public means of transportation may be insufficient, initiatives promoting the resumption of car driving are necessary in healthcare sectors, including day rehabilitation facilities. We provided support to a patient with an attention disorder due to left thalamic infarction, with the aim of achieving sufficient independence to drive a car, in a day rehabilitation facility. We herein report this case from the perspective of a speech-language-hearing therapist. The patient was a right-handed man in his 60s who had higher brain dysfunction with attention disorder as the main symptom. No marked motor paralysis of the extremities was observed. Use of a day rehabilitation service was started approximately two months after the onset of symptoms. Rehabilitation and support aimed at the resumption of car driving were provided approximately one month after the start of the day rehabilitation service use. To determine whether or not the patient was fit to drive a car, higher brain function tests for the intellectual function, attention function, and frontal function, as well as a theoretical evaluation based on the Stroke Drivers' Screening Assessment Japanese Version (J-SDSA) and monitoring of daily behaviors were performed. In addition, after the patient was given permission from an attending physician to drive a car on the condition that the patient did not drive fast and the patient's wife always accompanied him while driving, a safety assessment was also performed. As a result, approximately 10 months later, the J-SDSA theoretical evaluation score showed a passing grade, in contrast to the failing grade he had previously earned. Furthermore, errors in performing household activities due to a decreased attention function became unremarkable with respect to daily behaviors; therefore, we determined, together with the attending physician, that the patient now had sufficient independence to drive a car. In our day rehabilitation facility, the number of requests for advice on car driving from people with higher brain dysfunction living in the community had been increasing. Multisectoral assessments, training, and instruction should be continued in collaboration with attending physicians, other facilities located within the community, and driving schools in order to support people with higher brain dysfunction and help them once again be able to drive a car.

Effects of Gadolinium Deposition in the Brain on Motor or Behavioral Function: A Mouse Model.

Background Recent studies showing gadolinium deposition in multiple organs have raised concerns about the safety of gadolinium-based contrast agents (GBCAs). Purpose To explore whether gadolinium deposition in brain structures will cause any motor or behavioral alterations. Materials and Methods This study was performed from July 2019 to December 2020. Groups of 17 female BALB/c mice were each repeatedly injected with phosphate-buffered saline (control group, group A), a macrocyclic GBCA (group B), or a linear GBCA (group C) for 8 weeks (5 mmol per kilogram of bodyweight per week for GBCAs). Brain MRI studies were performed every other week to observe the signal intensity change caused by the gadolinium deposition. After the injection period, rotarod performance test, open field test, elevated plus-maze test, light-dark anxiety test, locomotor activity assessment test, passive avoidance memory test, Y-maze test, and forced swimming test were performed to assess the locomotor abilities, anxiety level, and memory. Among-group differences were compared by using one-way or two-way factorial analysis of variance with Tukey post hoc testing or Dunnett post hoc testing. Results Gadolinium deposition in the bilateral deep cerebellar nuclei was confirmed with MRI only in mice injected with a linear GBCA. At 8 weeks, contrast ratio of group C (0.11; 95% CI: 0.10, 0.12) was higher than that of group A (-2.1 × 10-3; 95% CI: -0.011, 7.5 × 10-3; P < .001) and group B (2.7 × 10-4; 95% CI: -8.2 × 10-3, 8.7 × 10-3; P < .001). Behavioral analyses showed that locomotor abilities, anxiety level, and long-term or short-term memory were not different in mice injected with linear or macrocyclic GBCAs. Conclusion No motor or behavioral alterations were observed in mice with brain gadolinium deposition. Also, the findings support the safety of macrocyclic gadolinium-based contrast agents. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Chen in this issue.

[A patient with mesenteric lymphoma who developed amyotrophic lateral sclerosis and sepsis].

We treated a patient with mesenteric lymphoma who concomitantly developed amyotrophic lateral sclerosis (ALS). The patient died of urinary tract infection nine months after the onset of ALS. We herein report the changes in the patient's condition and the sequence of events until death from the viewpoint of a physiotherapist. The patient was a 69-year-old woman who developed mesenteric lymphoma in September of X year and perceived weakness in the toes in November of X year. She showed signs of upper and lower motor neuron disorders, and electrophysiologic testing revealed denervation in three areas of the spinal cord. In March of X+1 year, she was diagnosed with definite ALS based on the Awaji criteria. In April of X+1 year, she began to receive continuous home healthcare, specifically outpatient rehabilitation. No remarkable bulbar palsy was observed soon after the initiation of rehabilitation; however, manual muscle testing revealed strengths in the lower and upper limbs of 1 and 3-5, respectively, indicating muscle weakness and muscle atrophy. She developed exacerbation of neurological symptoms in the upper limbs, bulbar palsy, and respiratory muscle paralysis during rehabilitation. The ALS Functional Rating Scale-Revised indicated a decreased tendency to X [please define X]. In July of X+1 year, the mesenteric lymphoma had enlarged, resulting in the development of ureteric obstruction and ultimately causing hydronephrosis. Urinary tract infection and sepsis were diagnosed, and she was hospitalized. Although her fever temporarily subsided following ceftriaxone administration, she ultimately died due to a systemic inflammatory response syndrome three days after hospitalization. The mean period between the ALS onset and death is reportedly 40.6±33.1 months. The rate of ALS progression differs among individuals. Malignant tumors and paraneoplastic neurological syndrome may be involved in rapidly worsening neurological symptoms. Patients who concomitantly develop motor neuron disorders and malignant tumors are likely to have a higher risk of developing serious conditions associated with the exacerbation of neurological symptoms and complications. Our patient had several diseases that affected her survival prognosis; however, the sharing of information regarding her condition among healthcare professionals may have been insufficient. The primary physician responsible for treating each disease should cooperate with physiotherapists and other paramedical staff who have frequent opportunities to talk to patients in daily clinical practice. In geriatric patients in particular, such an environment is essential.

In vitro Demonstration of Melanoma Metastasis in Lymph Nodes of Prepared Specimens Using a Light-emitting Diode-based Multispectral Photoacoustic Ultrasound Imaging System.

Although an excellent photoacoustic (PA) ultrasound method has been reported for the evaluation of lymph node melanoma metastasis in animal experiments, it remains to be evaluated in clinical trials. Recently, we performed PA ultrasound assessment using light-emitting diodes to detect metastatic melanoma in the lymph nodes of specimens prepared for microscopic examination. The PA effect was not obvious in amelanotic melanoma, but was seen in melanotic melanoma by PA imaging (PAI) and histopathological correlation in cases of primary melanotic melanoma accompanied by metastatic lymph nodes, including the coexistence of amelanotic melanoma and melanotic melanoma. Clinical workup should be performed with not only PAI but also conventional ultrasonography in cases with metastasis related to amelanotic transformation, which would likely be missed by PAI alone.

Prediction of bone mineral density from computed tomography: application of deep learning with a convolutional neural network.

OBJECTIVES: To investigate whether a deep learning model can predict the bone mineral density (BMD) of lumbar vertebrae from unenhanced abdominal computed tomography (CT) images. METHODS: In this Institutional Review Board-approved retrospective study, patients who received both unenhanced CT examinations and dual-energy X-ray absorptiometry (DXA) of the lumbar vertebrae, in two institutions (1 and 2), were included. Supervised deep learning was employed to obtain a convolutional neural network (CNN) model using axial CT images, including the lumbar vertebrae as input data and BMD values obtained with DXA as reference data. For this purpose, 1665 CT images from 183 patients in institution 1, which were augmented to 99,900 (= 1665 × 60) images (noise adding, parallel shift and rotation were performed), were used. Internal (by using data of 45 other patients in institution 1) and external validations (by using data of 50 patients in institution 2) were performed to evaluate the performance of the trained CNN model. Correlations and diagnostic performances were evaluated with Pearson's correlation coefficient (r) and area under the receiver operating characteristic curve (AUC), respectively. RESULTS: The estimated BMD values, according to the CNN model (BMDCNN), were significantly correlated with the BMD values obtained with DXA (r = 0.852 (p < 0.001) and 0.840 (p < 0.001) for the internal and external validation datasets, respectively). Using BMDCNN, osteoporosis was diagnosed with AUCs of 0.965 and 0.970 for the internal and external validation datasets, respectively. CONCLUSIONS: Using deep learning, the BMD of lumbar vertebrae could be predicted from unenhanced abdominal CT images. KEY POINTS: • By applying a deep learning technique, the bone mineral density (BMD) of lumbar vertebrae can be estimated from unenhanced abdominal CT images. • A strong correlation was observed between the estimated BMD and the BMD obtained with DXA. • By using the estimated BMD, osteoporosis could be diagnosed with high performance.

Deep learning to differentiate parkinsonian disorders separately using single midsagittal MR imaging: a proof of concept study.

OBJECTIVES: To evaluate the diagnostic performance of deep learning with the convolutional neural networks (CNN) to distinguish each representative parkinsonian disorder using MRI. METHODS: This clinical retrospective study was approved by the institutional review board, and the requirement for written informed consent was waived. Midsagittal T1-weighted MRI of a total of 419 subjects (125 Parkinson's disease (PD), 98 progressive supranuclear palsy (PSP), and 54 multiple system atrophy with predominant parkinsonian features (MSA-P) patients, and 142 normal subjects) between January 2012 and April 2016 was retrospectively assessed. To deal with the overfitting problem of deep learning, all subjects were randomly divided into training (85%) and validation (15%) data sets with the same proportions of each disease and normal subjects. We trained the CNN to distinguish each parkinsonian disorder using single midsagittal T1-weighted MRI with a training group to minimize the differences between predicted output probabilities and the clinical diagnoses; then, we adopted the trained CNN to the validation data set. Subjects were classified into each parkinsonian disorder or normal condition according to the final diagnosis of the trained CNN, and we assessed the diagnostic performance of the CNN. RESULTS: The accuracies of diagnostic performances regarding PD, PSP, MSA-P, and normal subjects were 96.8, 93.7, 95.2, and 98.4%, respectively. The areas under the receiver operating characteristic curves for distinguishing each condition from others (PD, PSP, MSA-P, and normal subjects) were 0.995, 0.982, 0.990, and 1.000, respectively. CONCLUSION: Deep learning with CNN enables highly accurate discrimination of parkinsonian disorders using MRI. KEY POINTS: • Deep learning convolution neural network achieves differential diagnosis of PD, PSP, MSA-P, and normal controls with an accuracy of 96.8, 93.7, 95.2, and 98.4%, respectively. • The areas under the curves for distinguishing between PD, PSP, MSA-P, and normality were 0.995, 0.982, 0.990, and 1.000, respectively. • CNN may learn important features that humans not notice, and has a possibility to perform previously impossible diagnoses.

Factors associated with the size of the adhesio interthalamica based on 3.0-T magnetic resonance images.

BACKGROUND: Adhesio interthalamica (AI) is a small structure connecting bilateral thalami. PURPOSE: To evaluate the effects of patient age, sex, and lateral diameter of the third ventricle on the long diameter of the AI using multivariate analyses based on magnetic resonance (MR) images obtained with 3.0-T scanners. MATERIAL AND METHODS: This clinical retrospective study included images of 153 patients who underwent MR examination using 3.0-T scanners. The long diameter of the AI and lateral diameter of the third ventricle were measured on images in the mid-sagittal plane and axial plane at the anterior commissure, respectively. Univariate and multivariate analyses were performed. RESULTS: AI was observed in 138 patients (70 men, 68 women; mean age = 63.7 ± 13.7 years; mean AI size =5.34 ± 1.63 mm). By univariate analyses, patient age (r = -0.262, P = 0.002), sex ( P = 0.010), and lateral diameter of the third ventricle (r = -0.642, P < 0.001) were significantly associated with the long diameter of the AI. With multiple linear regression analyses with a stepwise selection of parameters, only the lateral diameter of the third ventricle (estimate = -0.432, P < 0.001) was significantly associated with the long diameter of the AI. The lateral diameter of the third ventricle was longer in patients without AI (15 patients) than in those with AI ( P = 0.006). CONCLUSION: The lateral diameter of the third ventricle was a major factor negatively associated with the long diameter of the AI.

Liver Fibrosis: Deep Convolutional Neural Network for Staging by Using Gadoxetic Acid-enhanced Hepatobiliary Phase MR Images.

Purpose To investigate the performance of a deep convolutional neural network (DCNN) model in the staging of liver fibrosis using gadoxetic acid-enhanced hepatobiliary phase magnetic resonance (MR) imaging. Materials and Methods This retrospective study included patients for whom input data (hepatobiliary phase MR images, static magnetic field of the imaging unit, and hepatitis B and C virus testing results available, either positive or negative) and reference standard data (liver fibrosis stage evaluated from biopsy or surgical specimens obtained within 6 months of the MR examinations) were available were assigned to the training (534 patients) or the test (100 patients) group. For the training group (54, 53, 81, 113, and 233 patients with fibrosis stages F0, F1, F2, F3, and F4, respectively; mean patient age, 67.4 ± 9.7 years; 388 men and 146 women), MR images with three different section levels were augmented 90-fold (rotated, parallel-shifted, brightness-changed and contrast-changed images were generated; a total of 144 180 images). Supervised training was performed by using the DCNN model to minimize the difference between the output data (fibrosis score obtained through deep learning [FDL score]) and liver fibrosis stage. The performance of the DCNN model was evaluated in the test group (10, 10, 15, 20, and 45 patients with fibrosis stages F0, F1, F2, F3, and F4, respectively; mean patient age, 66.8 years ± 10.7; 71 male patients and 29 female patients) with receiver operating characteristic (ROC) analyses. Results The FDL score was correlated significantly with fibrosis stage (Spearman rank correlation coefficient: 0.63; P < .001). Fibrosis stages F4, F3, and F2 were diagnosed with areas under the ROC curve of 0.84, 0.84, and 0.85, respectively. Conclusion The DCNN model exhibited a high diagnostic performance in the staging of liver fibrosis. © RSNA, 2017 Online supplemental material is available for this article.

Deep Learning with Convolutional Neural Network for Differentiation of Liver Masses at Dynamic Contrast-enhanced CT: A Preliminary Study.

Purpose To investigate diagnostic performance by using a deep learning method with a convolutional neural network (CNN) for the differentiation of liver masses at dynamic contrast agent-enhanced computed tomography (CT). Materials and Methods This clinical retrospective study used CT image sets of liver masses over three phases (noncontrast-agent enhanced, arterial, and delayed). Masses were diagnosed according to five categories (category A, classic hepatocellular carcinomas [HCCs]; category B, malignant liver tumors other than classic and early HCCs; category C, indeterminate masses or mass-like lesions [including early HCCs and dysplastic nodules] and rare benign liver masses other than hemangiomas and cysts; category D, hemangiomas; and category E, cysts). Supervised training was performed by using 55 536 image sets obtained in 2013 (from 460 patients, 1068 sets were obtained and they were augmented by a factor of 52 [rotated, parallel-shifted, strongly enlarged, and noise-added images were generated from the original images]). The CNN was composed of six convolutional, three maximum pooling, and three fully connected layers. The CNN was tested with 100 liver mass image sets obtained in 2016 (74 men and 26 women; mean age, 66.4 years ± 10.6 [standard deviation]; mean mass size, 26.9 mm ± 25.9; 21, nine, 35, 20, and 15 liver masses for categories A, B, C, D, and E, respectively). Training and testing were performed five times. Accuracy for categorizing liver masses with CNN model and the area under receiver operating characteristic curve for differentiating categories A-B versus categories C-E were calculated. Results Median accuracy of differential diagnosis of liver masses for test data were 0.84. Median area under the receiver operating characteristic curve for differentiating categories A-B from C-E was 0.92. Conclusion Deep learning with CNN showed high diagnostic performance in differentiation of liver masses at dynamic CT. © RSNA, 2017 Online supplemental material is available for this article.

PEG-poly(L-lysine)-based polymeric micelle MRI contrast agent: Feasibility study of a Gd-micelle contrast agent for MR lymphography.

PURPOSE: To investigate the feasibility of polymeric micelle of poly(ethyleneglycol) (PEG)-b-poly(L-lysine-DOTA) (Gd-micelle) as a contrast agent for magnetic resonance lymphography (MRL). MATERIALS AND METHODS: Twenty-four female BALB/c mice were randomly divided into four groups of six mice each. Among them, mice of two groups were injected of complete Freund's adjuvant to obtain inflamed lymph nodes. We subcutaneously injected 0.5 µmol Gd per mouse of Gd-micelle or gadofluorine P in the right rear footpad. Identical 3D T1 -weighted gradient-echo imaging (1T MRI system) were subsequently obtained to create time-intensity curves of the right popliteal, sacral, and lumbar-aortic lymph nodes and to measure the contrast ratios (CRs). The peak CR, area under the curve (AUC), and elimination half-life (T1/2 ) of CR of the popliteal lymph node were assessed by two-way factorial analysis of variance. We also performed a qualitative assessment of normal and inflamed lymph node at three timepoints. RESULTS: The mean peak CR of Gd-micelle was 2.64 and 1.89 for gadofluorine P in normal mice, and 3.48 and 2.73 in the inflamed lymph node. Statistically, peak CR was higher for Gd-micelle (P = 0.004). In addition, the AUC was larger (P < 0.001) and T1/2 was longer (P < 0.001) for Gd-micelle. In qualitative assessment, Gd-micelle demonstrated the same or higher scores in every lymph node, and demonstrated a higher score in lumbar-aortic lymph node of a 360-minute image (P = 0.006) and in inflamed lymph node of a 360-minute image (P = 0.009). CONCLUSION: Compared to gadofluorine P, Gd-micelle showed higher and more prolonged enhancement in MRL imaging in normal and inflamed lymph nodes. LEVEL OF EVIDENCE: 1 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2018;47:238-245.

Deep learning for staging liver fibrosis on CT: a pilot study.

OBJECTIVES: To investigate whether liver fibrosis can be staged by deep learning techniques based on CT images. METHODS: This clinical retrospective study, approved by our institutional review board, included 496 CT examinations of 286 patients who underwent dynamic contrast-enhanced CT for evaluations of the liver and for whom histopathological information regarding liver fibrosis stage was available. The 396 portal phase images with age and sex data of patients (F0/F1/F2/F3/F4 = 113/36/56/66/125) were used for training a deep convolutional neural network (DCNN); the data for the other 100 (F0/F1/F2/F3/F4 = 29/9/14/16/32) were utilised for testing the trained network, with the histopathological fibrosis stage used as reference. To improve robustness, additional images for training data were generated by rotating or parallel shifting the images, or adding Gaussian noise. Supervised training was used to minimise the difference between the liver fibrosis stage and the fibrosis score obtained from deep learning based on CT images (FDLCT score) output by the model. Testing data were input into the trained DCNNs to evaluate their performance. RESULTS: The FDLCT scores showed a significant correlation with liver fibrosis stage (Spearman's correlation coefficient = 0.48, p < 0.001). The areas under the receiver operating characteristic curves (with 95% confidence intervals) for diagnosing significant fibrosis (≥ F2), advanced fibrosis (≥ F3) and cirrhosis (F4) by using FDLCT scores were 0.74 (0.64-0.85), 0.76 (0.66-0.85) and 0.73 (0.62-0.84), respectively. CONCLUSIONS: Liver fibrosis can be staged by using a deep learning model based on CT images, with moderate performance. KEY POINTS: • Liver fibrosis can be staged by a deep learning model based on magnified CT images including the liver surface, with moderate performance. • Scores from a trained deep learning model showed moderate correlation with histopathological liver fibrosis staging. • Further improvement are necessary before utilisation in clinical settings.

Imaging prediction of nonalcoholic steatohepatitis using computed tomography texture analysis.

OBJECTIVES: To determine if texture analysis of non-contrast-enhanced CT (NECT) images is able to predict nonalcoholic steatohepatitis (NASH). METHODS: NECT images from 88 patients who underwent a liver biopsy for the diagnosis of suspected NASH were assessed and texture feature parameters were obtained without and with filtration. The patient population was divided into a predictive learning dataset and a validation dataset, and further divided into groups according to the prediction of liver fibrosis as assessed by hyaluronic acid levels. The reference standard was the histological result of a liver biopsy. A predictive model for NASH was developed using parameters derived from the learning dataset that demonstrated areas under the receiver operating characteristic curve (AUC) of >0.65. The resulting model was then applied to the validation dataset. RESULTS: In patients without suspected fibrosis, the texture parameter mean without filter and skewness with a 2-mm filter were selected for the NASH prediction model. The AUC of the predictive model for the validation dataset was 0.94 and the accuracy was 94%. In patients with suspicion of fibrosis, the mean without filtration and kurtosis with a 4-mm filter were selected for the NASH prediction model. The AUC for the validation dataset was 0.60 and the accuracy was 42%. CONCLUSIONS: In patients without suspicion of fibrosis, NECT texture analysis effectively predicted NASH. KEY POINTS: • In patients without suspicion of fibrosis, NECT texture analysis effectively predicted NASH. • The mean without filtration and skewness with a 2-mm filter were modest predictors of NASH in patients without suspicion of liver fibrosis. • Hepatic fibrosis masks the characteristic texture features of NASH.

Gadoxetate disodium-induced tachypnoea and the effect of dilution method: a proof-of-concept study in mice.

OBJECTIVES: To directly investigate the rapid respiratory effect of gadoxetate disodium in an experimental study using mice. METHODS: After confirming the steady respiratory state under general anaesthesia, eight mice were injected with all test agents in the following order: phosphate-buffered saline (A, control group), 1.25 mmol/kg of gadoteridol (B) or gadopentetate dimeglumine (C), or 0.31 mmol/kg of gadoxetate disodium (D, E). The experimenter was not blinded to the agents. The injection dose was fixed as 100 µL for Groups A-D and 50 µL for Group E. We continuously monitored and recorded respiratory rate (RR), peripheral oxygen saturation (SpO2), and heart rate. The time-series changes from 0 to 30 s were compared by the linear mixed method RESULTS: Groups D and E showed the largest RR increase (20.6 and 20.3 breaths/min, respectively) and were significantly larger compared to Group A (3.36 breaths/min, both P<0.001). RR change of Groups D and E did not differ. RR change of Groups B and C was smaller (0.72 and 12.4 breaths/min, respectively) and did not differ statistically with Group A. Significant bradycardia was observed only in Group C (P<0.001). SpO2 was constant in all groups. CONCLUSIONS: Gadoxetate disodium causes a rapid tachypnoea without significant change of SpO2 and heart rate regardless of the dilution method. KEY POINTS: • Injection of gadoxetate disodium causes tachypnoea. • Dilution method did not alter the rapid respiratory effect of gadoxetate disodium. • The respiratory effect of gadoxetate disodium was larger than other contrast agents.

The inhibitory effect of gadoxetate disodium on hepatic transporters: a study using indocyanine green.

OBJECTIVES: To assess the inhibitory effect of gadoxetate disodium on the transporter system using indocyanine green (ICG). MATERIALS AND METHODS: Groups of six female B6 Albino mice were injected with the test agent (0.62 mmol/kg gadoxetate disodium) or phosphate-buffered saline (control) 10 min before injection of ICG. Identical fluorescence images were subsequently obtained to create time-efficiency curves of liver parenchymal uptake. The study was performed on hypothermic and normothermic mice. The logarithms of the absorption rate constants (logKa values) and of the elimination rate constants (logKe values) were calculated for each experimental condition, and between-group differences were compared using Student's t-test. RESULTS: The logKe values of the test group were lower than those of the control group at both temperatures (-6.52 vs. -5.87 under hypothermic conditions and -4.54 vs. -4.14 under normothermic conditions), and both differences were statistically significant (p = 0.037, 0.015 respectively). In terms of the logKa values, although the difference did not reach statistical significance (p = 0.052), the test group had lower values than the control group under hypothermic conditions (-0.771 vs. -0.376). In normothermic mice, the logKa values for the test and control groups were 0.037 and 0.277 respectively, thus not significantly different (p = 0.404). CONCLUSIONS: Gadoxetate disodium inhibited ICG excretion. Thus, gadoxetate disodium inhibited the ATP-binding cassette sub-family C member 2 transporter. KEY POINTS: • Gadoxetate disodium inhibited ICG excretion. • Gadoxetate disodium tended to inhibit hepatic ICG uptake. • Drug-drug interactions of gadoxetate disodium need further investigation.