Deep learning reconstruction for improving the visualization of acute brain infarct on computed tomography.

PURPOSE: This study aimed to investigate the impact of deep learning reconstruction (DLR) on acute infarct depiction compared with hybrid iterative reconstruction (Hybrid IR). METHODS: This retrospective study included 29 (75.8 ± 13.2 years, 20 males) and 26 (64.4 ± 12.4 years, 18 males) patients with and without acute infarction, respectively. Unenhanced head CT images were reconstructed with DLR and Hybrid IR. In qualitative analyses, three readers evaluated the conspicuity of lesions based on five regions and image quality. A radiologist placed regions of interest on the lateral ventricle, putamen, and white matter in quantitative analyses, and the standard deviation of CT attenuation (i.e., quantitative image noise) was recorded. RESULTS: Conspicuity of acute infarct in DLR was superior to that in Hybrid IR, and a statistically significant difference was observed for two readers (p ≤ 0.038). Conspicuity of acute infarct with time from onset to CT imaging at < 24 h in DLR was significantly improved compared with Hybrid IR for all readers (p ≤ 0.020). Image noise in DLR was significantly reduced compared with Hybrid IR with both the qualitative and quantitative analyses (p < 0.001 for all). CONCLUSION: DLR in head CT helped improve acute infarct depiction, especially those with time from onset to CT imaging at < 24 h.

High-angular resolution diffusion imaging generation using 3d u-net.

PURPOSE: To investigate the effects on tractography of artificial intelligence-based prediction of motion-probing gradients (MPGs) in diffusion-weighted imaging (DWI). METHODS: The 251 participants in this study were patients with brain tumors or epileptic seizures who underwent MRI to depict tractography. DWI was performed with 64 MPG directions and b = 0 s/mm2 images. The dataset was divided into a training set of 191 (mean age 45.7 [± 19.1] years), a validation set of 30 (mean age 41.6 [± 19.1] years), and a test set of 30 (mean age 49.6 [± 18.3] years) patients. Supervised training of a convolutional neural network was performed using b = 0 images and the first 32 axes of MPG images as the input data and the second 32 axes as the reference data. The trained model was applied to the test data, and tractography was performed using (a) input data only; (b) input plus prediction data; and (c) b = 0 images and the 64 MPG data (as a reference). RESULTS: In Q-ball imaging tractography, the average dice similarity coefficient (DSC) of the input plus prediction data was 0.715 (± 0.064), which was significantly higher than that of the input data alone (0.697 [± 0.070]) (p < 0.05). In generalized q-sampling imaging tractography, the average DSC of the input plus prediction data was 0.769 (± 0.091), which was also significantly higher than that of the input data alone (0.738 [± 0.118]) (p < 0.01). CONCLUSION: Diffusion tractography is improved by adding predicted MPG images generated by an artificial intelligence model.

Deep learning reconstruction for high-resolution computed tomography images of the temporal bone: comparison with hybrid iterative reconstruction.

PURPOSE: We investigated whether the quality of high-resolution computed tomography (CT) images of the temporal bone improves with deep learning reconstruction (DLR) compared with hybrid iterative reconstruction (HIR). METHODS: This retrospective study enrolled 36 patients (15 men, 21 women; age, 53.9 ± 19.5 years) who had undergone high-resolution CT of the temporal bone. Axial and coronal images were reconstructed using DLR, HIR, and filtered back projection (FBP). In qualitative image analyses, two radiologists independently compared the DLR and HIR images with FBP in terms of depiction of structures, image noise, and overall quality, using a 5-point scale (5 = better than FBP, 1 = poorer than FBP) to evaluate image quality. The other two radiologists placed regions of interest on the tympanic cavity and measured the standard deviation of CT attenuation (i.e., quantitative image noise). Scores from the qualitative and quantitative analyses of the DLR and HIR images were compared using, respectively, the Wilcoxon signed-rank test and the paired t-test. RESULTS: Qualitative and quantitative image noise was significantly reduced in DLR images compared with HIR images (all comparisons, p ≤ 0.016). Depiction of the otic capsule, auditory ossicles, and tympanic membrane was significantly improved in DLR images compared with HIR images (both readers, p ≤ 0.003). Overall image quality was significantly superior in DLR images compared with HIR images (both readers, p < 0.001). CONCLUSION: Compared with HIR, DLR provided significantly better-quality high-resolution CT images of the temporal bone.

Automated classification of brain MRI reports using fine-tuned large language models.

PURPOSE: This study aimed to investigate the efficacy of fine-tuned large language models (LLM) in classifying brain MRI reports into pretreatment, posttreatment, and nontumor cases. METHODS: This retrospective study included 759, 284, and 164 brain MRI reports for training, validation, and test dataset. Radiologists stratified the reports into three groups: nontumor (group 1), posttreatment tumor (group 2), and pretreatment tumor (group 3) cases. A pretrained Bidirectional Encoder Representations from Transformers Japanese model was fine-tuned using the training dataset and evaluated on the validation dataset. The model which demonstrated the highest accuracy on the validation dataset was selected as the final model. Two additional radiologists were involved in classifying reports in the test datasets for the three groups. The model's performance on test dataset was compared to that of two radiologists. RESULTS: The fine-tuned LLM attained an overall accuracy of 0.970 (95% CI: 0.930-0.990). The model's sensitivity for group 1/2/3 was 1.000/0.864/0.978. The model's specificity for group1/2/3 was 0.991/0.993/0.958. No statistically significant differences were found in terms of accuracy, sensitivity, and specificity between the LLM and human readers (p ≥ 0.371). The LLM completed the classification task approximately 20-26-fold faster than the radiologists. The area under the receiver operating characteristic curve for discriminating groups 2 and 3 from group 1 was 0.994 (95% CI: 0.982-1.000) and for discriminating group 3 from groups 1 and 2 was 0.992 (95% CI: 0.982-1.000). CONCLUSION: Fine-tuned LLM demonstrated a comparable performance with radiologists in classifying brain MRI reports, while requiring substantially less time.

Transient interruption of contrast on CT pulmonary angiography: effect of mid-inspiratory vs. end-inspiratory respiration command.

PURPOSE: To investigate the effects of mid-inspiratory respiration commands and other factors on transient interruption of contrast (TIC) incidence on CT pulmonary angiography. METHODS: In this retrospective study, 824 patients (mean age, 66.1 ± 15.3 years; 342 males) who had undergone CT pulmonary angiography between January 2021 and February 2023 were included. Among them, 545 and 279 patients were scanned at end- and mid-inspiratory levels, respectively. By placing a circular region of interest, CT attenuation of the main pulmonary artery (CTMPA) was recorded. Associations between several factors, including patient age, body weight, sex, respiratory command vs. TIC and severe TIC incidence (defined as CTMPA < 200 and 150 HU, respectively), were assessed using logistic regression analyses with stepwise regression selection based on Akaike's information criterion. RESULTS: Mid-inspiratory respiration command, in addition to patient age and lighter body weight, had negative association with the incidence of TIC. Only patient age, lighter body weight, female sex, and larger cardiothoracic ratio were negatively associated with severe TIC incidence. Mid-inspiratory respiration commands helped reduce TIC incidence among patients aged < 65 years (p = 0.039) and those with body weight ≥ 75 kg (p = 0.005) who were at high TIC risk. CONCLUSION: Changing the respiratory command from end- to mid-inspiratory levels, as well as patient age and body weight, was significantly associated with TIC incidence.

Deep Learning Reconstruction Plus Single-Energy Metal Artifact Reduction for Supra Hyoid Neck CT in Patients With Dental Metals.

Purpose: We investigated the effect of deep learning reconstruction (DLR) plus single-energy metal artifact reduction (SEMAR) on neck CT in patients with dental metals, comparing it with DLR and with hybrid iterative reconstruction (Hybrid IR)-SEMAR. Methods: In this retrospective study, 32 patients (25 men, 7 women; mean age: 63 ± 15 years) with dental metals underwent contrast-enhanced CT of the oral and oropharyngeal regions. Axial images were reconstructed using DLR, Hybrid IR-SEMAR, and DLR-SEMAR. In quantitative analyses, degrees of image noise and artifacts were evaluated. In one-by-one qualitative analyses, 2 radiologists evaluated metal artifacts, the depiction of structures, and noise on five-point scales. In side-by-side qualitative analyses, artifacts and overall image quality were evaluated by comparing Hybrid IR-SEMAR with DLR-SEMAR. Results: Artifacts were significantly less with DLR-SEMAR than with DLR in quantitative (P < .001) and one-by-one qualitative (P < .001) analyses, which resulted in significantly better depiction of most structures (P < .004). Artifacts in side-by-side analysis and image noise in quantitative and one-by-one qualitative analyses (P < .001) were significantly less with DLR-SEMAR than with Hybrid IR-SEMAR, resulting in significantly better overall quality of DLR-SEMAR. Conclusions: Compared with DLR and Hybrid IR-SEMAR, DLR-SEMAR provided significantly better supra hyoid neck CT images in patients with dental metals.

Deep-Learning Reconstruction of High-Resolution CT Improves Interobserver Agreement for the Evaluation of Pulmonary Fibrosis.

Objective: This study aimed to investigate whether deep-learning reconstruction (DLR) improves interobserver agreement in the evaluation of honeycombing for patients with interstitial lung disease (ILD) who underwent high-resolution computed tomography (CT) compared with hybrid iterative reconstruction (HIR). Methods: In this retrospective study, 35 consecutive patients suspected of ILD who underwent CT including the chest region were included. High-resolution CT images of the unilateral lung with DLR and HIR were reconstructed for the right and left lungs. A radiologist placed regions of interest on the lung and measured standard deviation of CT attenuation (i.e., quantitative image noise). In the qualitative image analyses, 5 blinded readers assessed the presence of honeycombing and reticulation, qualitative image noise, artifacts, and overall image quality using a 5-point scale (except for artifacts which was evaluated using a 3-point scale). Results: The quantitative and qualitative image noise in DLR was remarkably reduced compared to that in HIR (P < .001). Artifacts and overall DLR quality were significantly improved compared to those of HIR (P < .001 for 4 out of 5 readers). Interobserver agreement in the evaluations of honeycombing and reticulation for DLR (0.557 [0.450-0.693] and 0.525 [0.470-0.541], respectively) were higher than those for HIR (0.321 [0.211-0.520] and 0.470 [0.354-0.533], respectively). A statistically significant difference was found for honeycombing (P = .014). Conclusions: DLR improved interobserver agreement in the evaluation of honeycombing in patients with ILD on CT compared to HIR.

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.

Comparison of Deep-Learning Image Reconstruction With Hybrid Iterative Reconstruction for Evaluating Lung Nodules With High-Resolution Computed Tomography.

OBJECTIVE: This study aimed to investigate the impact of deep-learning reconstruction (DLR) on the detailed evaluation of solitary lung nodule using high-resolution computed tomography (HRCT) compared with hybrid iterative reconstruction (hybrid IR). METHODS: This retrospective study was approved by our institutional review board and included 68 consecutive patients (mean ± SD age, 70.1 ± 12.0 years; 37 men and 31 women) who underwent computed tomography between November 2021 and February 2022. High-resolution computed tomography images with a targeted field of view of the unilateral lung were reconstructed using filtered back projection, hybrid IR, and DLR, which is commercially available. Objective image noise was measured by placing the regions of interest on the skeletal muscle and recording the SD of the computed tomography attenuation. Subjective image analyses were performed by 2 blinded radiologists taking into consideration the subjective noise, artifacts, depictions of small structures and nodule rims, and the overall image quality. In subjective analyses, filtered back projection images were used as controls. Data were compared between DLR and hybrid IR using the paired t test and Wilcoxon signed-rank sum test. RESULTS: Objective image noise in DLR (32.7 ± 4.2) was significantly reduced compared with hybrid IR (35.3 ± 4.4) ( P < 0.0001). According to both readers, significant improvements in subjective image noise, artifacts, depictions of small structures and nodule rims, and overall image quality were observed in images derived from DLR compared with those from hybrid IR ( P < 0.0001 for all). CONCLUSIONS: Deep-learning reconstruction provides a better high-resolution computed tomography image with improved quality compared with hybrid IR.

Effect of Deep Learning Reconstruction on Evaluating Cervical Spinal Canal Stenosis With Computed Tomography.

OBJECTIVE: Magnetic resonance imaging (MRI) is commonly used to evaluate cervical spinal canal stenosis; however, some patients are ineligible for MRI. We aimed to assess the effect of deep learning reconstruction (DLR) in evaluating cervical spinal canal stenosis using computed tomography (CT) compared with hybrid iterative reconstruction (hybrid IR). METHODS: This retrospective study included 33 patients (16 male patients; mean age, 57.7 ± 18.4 years) who underwent cervical spine CT. Images were reconstructed using DLR and hybrid IR. In the quantitative analyses, noise was recorded by placing the regions of interest on the trapezius muscle. In the qualitative analyses, 2 radiologists evaluated the depiction of structures, image noise, overall image quality, and degree of cervical canal stenosis. We additionally evaluated the agreement between MRI and CT in 15 patients for whom preoperative cervical MRI was available. RESULTS: Image noise was less with DLR than hybrid IR in the quantitative ( P ≤ 0.0395) and subjective analyses ( P ≤ 0.0023), and the depiction of most structures was improved ( P ≤ 0.0052), which resulted in better overall quality ( P ≤ 0.0118). Interobserver agreement in the assessment of spinal canal stenosis with DLR (0.7390; 95% confidence interval [CI], 0.7189-0.7592) was superior to that with hybrid IR (0.7038; 96% CI, 0.6846-0.7229). As for the agreement between MRI and CT, significant improvement was observed for 1 reader with DLR (0.7910; 96% CI, 0.7762-0.8057) than hybrid IR (0.7536; 96% CI, 0.7383-0.7688). CONCLUSIONS: Deep learning reconstruction provided better quality cervical spine CT images in the evaluation of cervical spinal stenosis than hybrid IR.

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.

Assessing the Effects of Deep Learning Reconstruction on Abdominal CT Without Arm Elevation.

Purpose: To evaluate the effects of deep learning reconstruction (DLR) on image quality of abdominal computed tomography (CT) in patients without arm elevation compared with hybrid-iterative reconstruction (Hybrid-IR) and filtered back projection (FBP). Methods: In this retrospective study, axial images of 26 patients who underwent CT without arm elevation were reconstructed using DLR, Hybrid-IR, and FBP. Streak artifact index (SAI) was calculated by dividing the standard deviation of CT attenuation in the liver or spleen by that in fat. Two other blinded radiologists evaluated streak artifacts on images (in the liver, spleen, and kidney), depiction of liver vessels, subjective image noise, and overall quality. They were also asked to detect space-occupying lesions other than cysts in the liver, spleen, and kidney. Results: The SAI (liver/spleen) in DLR images was significantly reduced compared with Hybrid-IR and FBP. Regarding qualitative image analysis, streak artifacts in the 3 organs, qualitative image noise, and overall quality in DLR images were rated by both readers as significantly improved compared with Hybrid-IR (P ≤ .012) and FBP (P < .001). Both blinded readers detected more lesions on DLR images than on Hybrid-IR and FBP ones. Conclusion: DLR resulted in significantly better-quality abdominal CT images in patients scanned without elevating their arms with reducing streak artifacts compared with Hybrid-IR and FBP.

Detection of Vertebral Mass and Diagnosis of Spinal Cord Compression in Computed Tomography With Deep Learning Reconstruction: Comparison With Hybrid Iterative Reconstruction.

PURPOSE: To compare the impact of deep learning reconstruction (DLR) and hybrid-iterative reconstruction (hybrid-IR) on vertebral mass depiction, detection, and diagnosis of spinal cord compression on computed tomography (CT). METHODS: This retrospective study included 29 and 20 patients with and without vertebral masses. CT images were reconstructed using DLR and hybrid-IR. Three readers performed vertebral mass detection tests and evaluated the presence of spinal cord compression, the depiction of vertebral masses, and image noise. Quantitative image noise was measured by placing regions of interest on the aorta and spinal cord. RESULTS: Deep learning reconstruction tended to improve the performance of readers with less diagnostic imaging experience in detecting vertebral masses (area under the receiver operating characteristic curve [AUC] = .892-.966) compared with hybrid-IR (AUC = .839-.917). Diagnostic performance in evaluating spinal cord compression in DLR (AUC = .887-.995) also improved compared with that in hybrid-IR (AUC = .866-.942) for some readers. The depiction of vertebral masses and subjective image noise in DLR were significantly improved compared with those in hybrid-IR (P < .041). Quantitative image noise in DLR was also significantly reduced compared with that in hybrid-IR (P < .001). CONCLUSION: Deep learning reconstruction improved the depiction of vertebral masses, which resulted in a tendency to improve the performance of CT compared to hybrid-IR in detecting vertebral masses and diagnosing spinal cord compression for some readers.

Accelerated Two-Point Dixon MR Angiography Improves Diagnostic Performance for Cervical Artery Diseases.

BACKGROUND: Nonenhanced MR angiography (MRA) studies are often used to manage acute and chronic large cervical artery disease, but lengthy scan times limit their clinical usefulness. PURPOSE: To develop an accelerated cervical MRA and test its diagnostic performance. STUDY TYPE: Prospective. POPULATION: Patients with cervical artery disease (n = 32, 17 males). FIELD STRENGTH/SEQUENCE: 3.0 T; accelerated two-point Dixon three-dimensional Cartesian spoiled gradient-echo (FLEXA) and conventional time-of-flight MRA (cMRA) sequences. ASSESSMENT: All patients underwent FLEXA (1'28″) and cMRA (6'47″) acquisitions. Quantitative evaluation (artery-to-background signal ratio and a blur metric) and qualitative evaluation using diagnostic performance measured by the sensitivity, specificity, and positive/negative predictive values (PPV/NPV), and vessel and plaque visualization scores from three board-certified radiologists' (with 10, 11, and 12 years of experience) independent readings using maximum intensity projection (MIP) for luminal diseases and axial images for plaque. The reference standards were contrast-enhanced angiography and fat-saturated T1-weighted images, respectively. STATISTICAL TESTS: All measures were compared between FLEXA and cMRA using the paired t, Wilcoxon signed-rank, McNemar's, or chi-squared test, as appropriate. Interreader agreement was assessed using Cohen's κ. P < 0.05 was considered statistically significant. RESULTS: The artery-to-background signal ratio was significantly higher for FLEXA (FLEXA: 7.20 ± 1.63 [fat]; 4.26 ± 0.52 [muscle]; cMRA: 2.57 ± 0.49 [fat]), while image blurring was significantly less (FLEXA: 0.24 ± 0.016; cMRA: 0.30 ± 0.029). In luminal disease detection, sensitivity (FLEXA: 0.97/0.91/0.91; cMRA:0.71/0.69/0.63), specificity (FLEXA: 0.98/0.93/0.98; cMRA:0.93/0.85/0.92), PPV (FLEXA: 0.92/0.86/0.86; cMRA: 0.64/0.5/0.58), and NPV (FLEXA: 0.99/0.98/0.98; cMRA: 0.92/0.91/0.9) were significantly higher for FLEXA. interreader agreement was substantial to almost perfect for FLEXA (κ = 0.82/0.86/0.78) and moderate to substantial for cMRA (κ = 0.67/0.56/0.57). MIP visualization scores were significantly higher for FLEXA, with substantial to almost perfect interreader agreement (FLEXA: κ = 0.83/0.86/0.82; cMRA: κ = 0.89/0.79/0.79). In plaque detection, sensitivity (FLEXA: 0.9/0.9/0.7; cMRA: 0.3/0.6/0.2) and specificity (FLEXA: 1/0.87/1; cMRA: 0.93/0.63/0.97) were significantly higher for FLEXA in two of three readers. The interreader plaque detection agreement was fair to substantial (FLEXA: κ = 0.63/0.69/0.48; cMRA: κ = 0.21/0.45/0.20). Side-by-side plaque and vessel wall visualization was superior for FLEXA in all readers, with moderate to substantial interreader agreement (plaque: κ = 0.73/0.73/0.77; vessel wall: κ = 0.57/0.40/0.39). DATA CONCLUSION: FLEXA enhanced visualization of the cervical arterial system and improved diagnostic performance for luminal abnormalities and plaques in patients with cervical artery diseases. LEVEL OF EVIDENCE: 1 TECHNICAL EFFICACY STAGE: 2.

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.

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.

Parkinson's disease: deep learning with a parameter-weighted structural connectome matrix for diagnosis and neural circuit disorder investigation.

PURPOSE: To investigate whether Parkinson's disease (PD) can be differentiated from healthy controls and to identify neural circuit disorders in PD by applying a deep learning technique to parameter-weighted and number of streamlines (NOS)-based structural connectome matrices calculated from diffusion-weighted MRI. METHODS: In this prospective study, 115 PD patients and 115 healthy controls were enrolled. NOS-based and parameter-weighted connectome matrices were calculated from MRI images obtained with a 3-T MRI unit. With 5-fold cross-validation, diagnostic performance of convolutional neural network (CNN) models using those connectome matrices in differentiating patients with PD from healthy controls was evaluated. To identify the important brain connections for diagnosing PD, gradient-weighted class activation mapping (Grad-CAM) was applied to the trained CNN models. RESULTS: CNN models based on some parameter-weighted structural matrices (diffusion kurtosis imaging (DKI)-weighted, neurite orientation dispersion and density imaging (NODDI)-weighted, and g-ratio-weighted connectome matrices) showed moderate performance (areas under the receiver operating characteristic curve (AUCs) = 0.895, 0.801, and 0.836, respectively) in discriminating PD patients from healthy controls. The DKI-weighted connectome matrix performed significantly better than the conventional NOS-based matrix (AUC = 0.761) (DeLong's test, p < 0.0001). Alterations of neural connections between the basal ganglia and cerebellum were indicated by applying Grad-CAM to the NODDI- and g-ratio-weighted matrices. CONCLUSION: Patients with PD can be differentiated from healthy controls by applying the deep learning technique to the parameter-weighted connectome matrices, and neural circuit disorders including those between the basal ganglia on one side and the cerebellum on the contralateral side were visualized.