SPINNED: Simulation-based physics-informed neural network for deconvolution of dynamic susceptibility contrast MRI perfusion data.

PURPOSE: To propose the simulation-based physics-informed neural network for deconvolution of dynamic susceptibility contrast (DSC) MRI (SPINNED) as an alternative for more robust and accurate deconvolution compared to existing methods. METHODS: The SPINNED method was developed by generating synthetic tissue residue functions and arterial input functions through mathematical simulations and by using them to create synthetic DSC MRI time series. The SPINNED model was trained using these simulated data to learn the underlying physical relation (deconvolution) between the DSC-MRI time series and the arterial input functions. The accuracy and robustness of the proposed SPINNED method were assessed by comparing it with two common deconvolution methods in DSC MRI data analysis, circulant singular value decomposition, and Volterra singular value decomposition, using both simulation data and real patient data. RESULTS: The proposed SPINNED method was more accurate than the conventional methods across all SNR levels and showed better robustness against noise in both simulation and real patient data. The SPINNED method also showed much faster processing speed than the conventional methods. CONCLUSION: These results support that the proposed SPINNED method can be a good alternative to the existing methods for resolving the deconvolution problem in DSC MRI. The proposed method does not require any separate ground-truth measurement for training and offers additional benefits of quick processing time and coverage of diverse clinical scenarios. Consequently, it will contribute to more reliable, accurate, and rapid diagnoses in clinical applications compared with the previous methods including those based on supervised learning.

Comparative Performance of Susceptibility Map-Weighted MRI According to the Acquisition Planes in the Diagnosis of Neurodegenerative Parkinsonism.

OBJECTIVE: To evaluate the diagnostic performance of susceptibility map-weighted imaging (SMwI) taken in different acquisition planes for discriminating patients with neurodegenerative parkinsonism from those without. MATERIALS AND METHODS: This retrospective, observational, single-institution study enrolled consecutive patients who visited movement disorder clinics and underwent brain MRI and 18F-FP-CIT PET between September 2021 and December 2021. SMwI images were acquired in both the oblique (perpendicular to the midbrain) and the anterior commissure-posterior commissure (AC-PC) planes. Hyperintensity in the substantia nigra was determined by two neuroradiologists. 18F-FP-CIT PET was used as the reference standard. Inter-rater agreement was assessed using Cohen's kappa coefficient. The diagnostic performance of SMwI in the two planes was analyzed separately for the right and left substantia nigra. Multivariable logistic regression analysis with generalized estimating equations was applied to compare the diagnostic performance of the two planes. RESULTS: In total, 194 patients were included, of whom 105 and 103 had positive results on 18F-FP-CIT PET in the left and right substantia nigra, respectively. Good inter-rater agreement in the oblique (κ = 0.772/0.658 for left/right) and AC-PC planes (0.730/0.741 for left/right) was confirmed. The pooled sensitivities for two readers were 86.4% (178/206, left) and 83.3% (175/210, right) in the oblique plane and 87.4% (180/206, left) and 87.6% (184/210, right) in the AC-PC plane. The pooled specificities for two readers were 83.5% (152/182, left) and 82.0% (146/178, right) in the oblique plane, and 83.5% (152/182, left) and 86.0% (153/178, right) in the AC-PC plane. There were no significant differences in the diagnostic performance between the two planes (P > 0.05). CONCLUSION: There are no significant difference in the diagnostic performance of SMwI performed in the oblique and AC-PC plane in discriminating patients with parkinsonism from those without. This finding affirms that each institution may choose the imaging plane for SMwI according to their clinical settings.

Radiomics using non-contrast CT to predict hemorrhagic transformation risk in stroke patients undergoing revascularization.

OBJECTIVES: This study explores whether textural features from initial non-contrast CT scans of infarcted brain tissue are linked to hemorrhagic transformation susceptibility. MATERIALS AND METHODS: Stroke patients undergoing thrombolysis or thrombectomy from Jan 2012 to Jan 2022 were analyzed retrospectively. Hemorrhagic transformation was defined using follow-up magnetic resonance imaging. A total of 94 radiomic features were extracted from the infarcted tissue on initial NCCT scans. Patients were divided into training and test sets (7:3 ratio). Two models were developed with fivefold cross-validation: one incorporating first-order and textural radiomic features, and another using only textural radiomic features. A clinical model was also constructed using logistic regression with clinical variables, and test set validation was performed. RESULTS: Among 362 patients, 218 had hemorrhagic transformations. The LightGBM model with all radiomics features had the best performance, with an area under the receiver operating characteristic curve (AUROC) of 0.986 (95% confidence interval [CI], 0.971-1.000) on the test dataset. The ExtraTrees model performed best when textural features were employed, with an AUROC of 0.845 (95% CI, 0.774-0.916). Minimum, maximum, and ten percentile values were significant predictors of hemorrhagic transformation. The clinical model showed an AUROC of 0.544 (95% CI, 0.431-0.658). The performance of the radiomics models was significantly better than that of the clinical model on the test dataset (p < 0.001). CONCLUSIONS: The radiomics model can predict hemorrhagic transformation using NCCT in stroke patients. Low Hounsfield unit was a strong predictor of hemorrhagic transformation, while textural features alone can predict hemorrhagic transformation. CLINICAL RELEVANCE STATEMENT: Using radiomic features extracted from initial non-contrast computed tomography, early prediction of hemorrhagic transformation has the potential to improve patient care and outcomes by aiding in personalized treatment decision-making and early identification of at-risk patients. KEY POINTS: • Predicting hemorrhagic transformation following thrombolysis in stroke is challenging since multiple factors are associated. • Radiomics features of infarcted tissue on initial non-contrast CT are associated with hemorrhagic transformation. • Textural features on non-contrast CT are associated with the frailty of the infarcted tissue.

Diagnostic performance of a high-spatial-resolution voxelwise analysis of neuromelanin-sensitive imaging in early-stage idiopathic Parkinson's disease.

BACKGROUND: Quantitative assessments of neuromelanin (NM) of the substantia nigra pars compacta (SNpc) in neuromelanin-sensitive MRI (NM-MRI) to determine its abnormality have been conducted by measuring either the volume or contrast ratio (CR) of the SNpc. A recent study determined the regions in the SNpc that are significantly different between early-stage idiopathic Parkinson's disease (IPD) patients and healthy controls (HCs) using a high spatial-resolution NM-MRI template, which enables a template-based voxelwise analysis to overcome the susceptibility of CR measurement to inter-rater discrepancy. We aimed to assess the diagnostic performance, which has not been reported, of the CRs between early-stage IPD patients and HCs using a NM-MRI template. METHODS: We retrospectively enrolled early-stage IPD patients (n = 50) and HCs (n = 50) who underwent 0.8-mm isovoxel NM-MRI and dopamine-transporter PET as the standard of reference. A template-based voxelwise analysis revealed two regions in nigrosomes 1 and 2 (N1 and N2, respectively), with significant differences in each substantia nigra (SNpc) between IPD and HCs. The mean CR values of N1, N2, volume-weighted mean of N1 and N2 (N1 + N2), and whole SNpc on each side were compared between IPD and HC using the independent t-test or the Mann-Whitney U test. The diagnostic performance was compared in each region using receiver operating characteristic curves. RESULTS: The mean CR values in the right N1 (0.149459 vs. 0.194505), left N1 (0.133328 vs. 0.169160), right N2 (0.230245 vs. 0.278181), left N2 (0.235784 vs. 0.314169), right N1 + N2 (0.155322 vs. 0.278143), left N1 + N2 (0.140991 vs. 0.276755), right whole SNpc (0.131397 vs. 0.141422), and left whole SNpc (0.127099 vs. 0.137873) significantly differed between IPD patients and HCs (all p < 0.001). The areas under the curve of the left N1 + N2, right N1 + N2, left N1, right N1, left N2, right N2, left whole SNpc, and right whole SNpc were 0.994 (sensitivity, 98.0%; specificity, 94.0%), 0.985, 0.804, 0.802, 0.777, 0.766, 0.632, and 0.606, respectively. CONCLUSION: Our NM-MRI template-based CR measurements revealed significant differences between early-stage IPD patients and HCs. The CR values of the left N1 + N2 demonstrated the highest diagnostic performance.

Nigrosome 1 visibility and its association with nigrostriatal dopaminergic loss in Parkinson's disease.

BACKGROUND: Nigrosome 1 (NG1), a small cluster of dopaminergic cells in the substantia nigra and visible in the susceptibility map-weighted magnetic resonance image (SMwI), is severely affected in Parkinson's disease (PD). However, the degree of nigrostriatal degeneration according to the visibility of NG1 has not yet been well elucidated. METHODS: We consecutively recruited 138 PD and 78 non-neurodegenerative disease (non-ND) patients, who underwent both 18 F-FP-CIT positron emission tomography (PET) and SMwI. Three neurologists and one radiologist evaluated the visibility of NG1 in SMwI. The participants were thereby grouped into visible, intermediate, and non-visible groups. Nigrostriatal dopaminergic input was calculated using the specific binding ratio (SBR) of the 18 F-FP-CIT PET. We determined the threshold of regional SBR for discriminating NG1 visibility and the probability for NG1 visibility according to regional SBR. RESULTS: Visual rating of NG1 showed excellent interobserver agreements as well as high sensitivity and specificity to differentiate the PD group from the non-ND group. NG1 was visible in seven patients (5.1%) in the PD group, who had relatively short disease duration or less severe loss of striatal dopamine. The threshold of putaminal SBR reduction on the more affected side for the disappearance of NG1 was 45.5%, and the probability for NG1 visibility dropped to 50% after the reduction of putaminal SBR to 41% from the normal mean. CONCLUSIONS: Almost half loss of nigrostriatal dopaminergic input is required to dissipate the hyperintensity of NG1 on SMwI, suggesting its utility in diagnosing PD only after the onset of the motor symptoms.

Correlation Between Neuromelanin-Sensitive MRI and 18F-FP-CIT PET in Early-Stage Parkinson's Disease: Utility of a Voxel-Wise Analysis by Using High-Spatial-Resolution MRI.

BACKGROUND AND PURPOSE: The correlation between dopamine transporter (DAT) imaging and neuromelanin-sensitive magnetic resonance imaging (NM-MRI) in early-stage Parkinson's disease (PD) has not yet been established. This study aimed to determine the correlation between NM-MRI and DAT positron-emission tomography (PET) in patients with early-stage PD. METHODS: Fifty drug-naïve patients with early-stage PD who underwent both 0.8-mm isovoxel NM-MRI and DAT PET were enrolled retrospectively. Using four regions of interest (nigrosome 1 and nigrosome 2 [N1 and N2] regions) from a previous study, the contrast ratios (CRs) of 12 regions were measured: N1, N2, flipped N1, flipped N2, combined N1 and N2, and whole substantia nigra pars compacta [SNpc] (all on both sides). The clinically more affected side was separately assessed. The standardized uptake value ratios (SUVRs) were measured in the striatum using DAT PET. A partial correlation analysis was performed between the SUVR and CR measurements. RESULTS: CR of the flipped left N1 region was significantly correlated with SUVR of the right posterior putamen (p=0.047), and CR values of the left N1 region, left N2 region, flipped right N1 region, and combined left N1 and N2 regions were significantly correlated with SUVR of the left posterior putamen (p=0.011, 0.038, 0.020, and 0.010, respectively). SUVR of the left anterior putamen was significantly correlated with CR of the left N2 region (p=0.027). On the clinically more affected side, the CR values of the N1 region, combined N1 and N2 regions, and the whole SNpc were significantly correlated with SUVR of the posterior putamen (p=0.001, 0.024, and 0.021, respectively). There were significant correlations between the SUVR of the anterior putamen and the CR values of the N1 region, combined N1 and N2 regions, and whole SNpc (p=0.027, 0.001, and 0.036, respectively). CONCLUSIONS: This study found that there were significant correlations between CR values in the SNpc on NM-MRI and striatal SUVR values on DAT PET on both sides in early-stage PD.

Outcome of epidural blood patch for imaging-negative spontaneous intracranial hypotension.

BACKGROUND: Spontaneous intracranial hypotension is diagnosed by an abnormal finding in brain MRI, spinal imaging, or lumbar puncture. However, the sensitivity of each test is low. We investigated whether patients with suspected spontaneous intracranial hypotension and negative imaging findings would respond to epidural blood patch. METHODS: We prospectively recruited patients with new-onset orthostatic headache admitted at the Samsung Medical Center from January 2017 to July 2021. In patients without abnormal imaging findings and no history of prior epidural blood patch, treatment outcome-defined as both 50% response in maximal headache intensity and improvement of orthostatic component-was collected at discharge and three months after epidural blood patch. RESULTS: We included 21 treatment-naïve patients with orthostatic headache and negative brain and spinal imaging results who received epidural blood patch. After epidural blood patch (mean 1.3 times, range 1-3), 14 (66.7%) and 19 (90.5%) patients achieved both 50% response and improvement of orthostatic component at discharge and three months post-treatment, respectively. Additionally, complete remission was reported in 11 (52.4%) patients at three-month follow-up, while most of the remaining patients had only mild headaches. Among nine (42.9%) patients who underwent lumbar puncture, none had an abnormally low opening pressure (median 13.8 cm H2O, range 9.2-21.5). CONCLUSION: Given the high responder rates of epidural blood patch in our study, empirical epidural blood patch should be considered to treat new-onset orthostatic headache, even when brain and spinal imaging are negative. The necessity of lumbar puncture is questionable considering the high response rate of epidural blood patch and low rate of "low pressure."

Perfusion Maps Acquired From Dynamic Angiography MRI Using Deep Learning Approaches.

BACKGROUND: A typical stroke MRI protocol includes perfusion-weighted imaging (PWI) and MR angiography (MRA), requiring a second dose of contrast agent. A deep learning method to acquire both PWI and MRA with single dose can resolve this issue. PURPOSE: To acquire both PWI and MRA simultaneously using deep learning approaches. STUDY TYPE: Retrospective. SUBJECTS: A total of 60 patients (30-73 years old, 31 females) with ischemic symptoms due to occlusion or ≥50% stenosis (measured relative to proximal artery diameter) of the internal carotid artery, middle cerebral artery, or anterior cerebral artery. The 51/1/8 patient data were used as training/validation/test. FIELD STRENGTH/SEQUENCE: A 3 T, time-resolved angiography with stochastic trajectory (contrast-enhanced MRA) and echo planar imaging (dynamic susceptibility contrast MRI, DSC-MRI). ASSESSMENT: We investigated eight different U-Net architectures with different encoder/decoder sizes and with/without an adversarial network to generate perfusion maps from contrast-enhanced MRA. Relative cerebral blood volume (rCBV), relative cerebral blood flow (rCBF), mean transit time (MTT), and time-to-max (Tmax ) were mapped from DSC-MRI and used as ground truth to train the networks and to generate the perfusion maps from the contrast-enhanced MRA input. STATISTICAL TESTS: Normalized root mean square error, structural similarity (SSIM), peak signal-to-noise ratio (pSNR), DICE, and FID scores were calculated between the perfusion maps from DSC-MRI and contrast-enhanced MRA. One-tailed t-test was performed to check the significance of the improvements between networks. P values < 0.05 were considered significant. RESULTS: The four perfusion maps were successfully extracted using the deep learning networks. U-net with multiple decoders and enhanced encoders showed the best performance (pSNR 24.7 ± 3.2 and SSIM 0.89 ± 0.08 for rCBV). DICE score in hypo-perfused area showed strong agreement between the generated perfusion maps and the ground truth (highest DICE: 0.95 ± 0.04). DATA CONCLUSION: With the proposed approach, dynamic angiography MRI may provide vessel architecture and perfusion-relevant parameters simultaneously from a single scan. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 5.

Detection of Cerebral Microbleeds in MR Images Using a Single-Stage Triplanar Ensemble Detection Network (TPE-Det).

BACKGROUND: Cerebral microbleeds (CMBs) are microscopic brain hemorrhages with implications for various diseases. Automated detection of CMBs is a challenging task due to their wide distribution throughout the brain, small size, and visual similarity to their mimics. For this reason, most of the previously proposed methods have been accomplished through two distinct stages, which may lead to difficulties in integrating them into clinical workflows. PURPOSE: To develop a clinically feasible end-to-end CMBs detection network with a single-stage structure utilizing 3D information. This study proposes triplanar ensemble detection network (TPE-Det), ensembling 2D convolutional neural networks (CNNs) based detection networks on axial, sagittal, and coronal planes. STUDY TYPE: Retrospective. SUBJECTS: Two datasets (DS1 and DS2) were used: 1) 116 patients with 367 CMBs and 12 patients without CMBs for training, validation, and testing (70.39 ± 9.30 years, 68 women, 60 men, DS1); 2) 58 subjects with 148 microbleeds and 21 subjects without CMBs only for testing (76.13 ± 7.89 years, 47 women, 32 men, DS2). FIELD STRENGTH/SEQUENCE: A 3 T field strength and 3D GRE sequence scan for SWI reconstructions. ASSESSMENT: The sensitivity, FPavg (false-positive per subject), and precision measures were computed and analyzed with statistical analysis. STATISTICAL TESTS: A paired t-test was performed to investigate the improvement of detection performance by the suggested ensembling technique in this study. A P value < 0.05 was considered significant. RESULTS: The proposed TPE-Det detected CMBs on the DS1 testing set with a sensitivity of 96.05% and an FPavg of 0.88, presenting statistically significant improvement. Even when the testing on DS2 was performed without retraining, the proposed model provided a sensitivity of 85.03% and an FPavg of 0.55. The precision was significantly higher than the other models. DATA CONCLUSION: The ensembling of multidimensional networks significantly improves precision, suggesting that this new approach could increase the benefits of detecting lesions in the clinic. EVIDENCE LEVEL: 1 TECHNICAL EFFICACY: Stage 2.

Sandwich spatial saturation for neuromelanin-sensitive MRI: Development and multi-center trial.

Neuromelanin (NM)-sensitive MRI using a magnetization transfer (MT)-prepared T1-weighted sequence has been suggested as a tool to visualize NM contents in the brain. In this study, a new NM-sensitive imaging method, sandwichNM, is proposed by utilizing the incidental MT effects of spatial saturation RF pulses in order to generate consistent high-quality NM images using product sequences. The spatial saturation pulses are located both superior and inferior to the imaging volume, increasing MT weighting while avoiding asymmetric MT effects. When the parameters of the spatial saturation were optimized, sandwichNM reported a higher NM contrast ratio than those of conventional NM-sensitive imaging methods with matched parameters for comparability with sandwichNM (SandwichNM: 23.6 ± 5.4%; MT-prepared TSE: 20.6 ± 7.4%; MT-prepared GRE: 17.4 ± 6.0%). In a multi-vendor experiment, the sandwichNM images displayed higher means and lower standard deviations of the NM contrast ratio across subjects in all three vendors (SandwichNM vs. MT-prepared GRE; Vendor A: 28.4 ± 1.5% vs. 24.4 ± 2.8%; Vendor B: 27.2 ± 1.0% vs. 13.3 ± 1.3%; Vendor C: 27.3 ± 0.7% vs. 20.1 ± 0.9%). For each subject, the standard deviations of the NM contrast ratio across the vendors were substantially lower in SandwichNM (SandwichNM vs. MT-prepared GRE; subject 1: 1.5% vs. 8.1%, subject 2: 1.1 % vs. 5.1%, subject 3: 0.9% vs. 4.0%, subject 4: 1.1% vs. 5.3%), demonstrating consistent contrasts across the vendors. The proposed method utilizes product sequences, requiring no alteration of a sequence and, therefore, may have a wide practical utility in exploring the NM imaging.

A prospective multi-centre study of susceptibility map-weighted MRI for the diagnosis of neurodegenerative parkinsonism.

OBJECTIVES: This study aimed to compare susceptibility map-weighted imaging (SMwI) using various MRI machines (three vendors) with N-3-fluoropropyl-2-ß-carbomethoxy-3-ß-(4-iodophe nyl)nortropane (18F-FP-CIT) PET in the diagnosis of neurodegenerative parkinsonism in a multi-centre setting. METHODS: We prospectively recruited 257 subjects, including 157 patients with neurodegenerative parkinsonism, 54 patients with non-neurodegenerative parkinsonism, and 46 healthy subjects from 10 hospitals between November 2019 and October 2020. All participants underwent both SMwI and 18F-FP-CIT PET. SMwI was interpreted by two independent reviewers for the presence or absence of abnormalities in nigrosome 1, and discrepancies were resolved by consensus. 18F-FP-CIT PET was used as the reference standard. Inter-observer agreement was tested using Cohen's kappa coefficient. McNemar's test was used to test the agreement between the interpretations of SMwI and 18F-FP-CIT PET per participant and substantia nigra (SN). RESULTS: The inter-observer agreement was 0.924 and 0.942 per SN and participant, respectively. The diagnostic sensitivity of SMwI was 97.9% and 99.4% per SN and participant, respectively; its specificity was 95.9% and 95.2%, respectively, and its accuracy was 97.1% and 97.7%, respectively. There was no significant difference between the results of SMwI and 18F-FP-CIT PET (p > 0.05, for both SN and participant). CONCLUSIONS: This study demonstrated that the high diagnostic performance of SMwI was maintained in a multi-centre setting with various MRI scanners, suggesting the generalisability of SMwI for determining nigrostriatal degeneration in patients with parkinsonism. KEY POINTS: • Susceptibility map-weighted imaging helps clinicians to predict nigrostriatal degeneration. • The protocol for susceptibility map-weighted imaging can be standardised across MRI vendors. • Susceptibility map-weighted imaging showed diagnostic performance comparable to that of dopamine transporter PET in a multi-centre setting with various MRI scanners.

Continuous Conversion of CT Kernel Using Switchable CycleGAN With AdaIN.

X-ray computed tomography (CT) uses different filter kernels to highlight different structures. Since the raw sinogram data is usually removed after the reconstruction, in case there is additional need for other types of kernel images that were not previously generated, the patient may need to be scanned again. Accordingly, there exists increasing demand for post-hoc image domain conversion from one kernel to another without sacrificing the image quality. In this paper, we propose a novel unsupervised continuous kernel conversion method using cycle-consistent generative adversarial network (cycleGAN) with adaptive instance normalization (AdaIN). Even without paired training data, not only can our network translate the images between two different kernels, but it can also convert images along the interpolation path between the two kernel domains. We also show that the quality of generated images can be further improved if intermediate kernel domain images are available. Experimental results confirm that our method not only enables accurate kernel conversion that is comparable to supervised learning methods, but also generates intermediate kernel images in the unseen domain that are useful for hypopharyngeal cancer diagnosis.

Early-stage Parkinson's disease: Abnormal nigrosome 1 and 2 revealed by a voxelwise analysis of neuromelanin-sensitive MRI.

Previous pathologic studies evaluated the substantia nigra pars compacta (SNpc) of a limited number of idiopathic Parkinson's disease (IPD) patients with relatively longer disease durations. Therefore, it remains unknown which region of the SNpc is most significantly affected in early-stage IPD. We hypothesized that a voxelwise analysis of thin-section neuromelanin-sensitive MRI (NM-MRI) may help determine the significantly affected regions of the SNpc in early-stage IPD and localize these areas in each nigrosome on high-spatial-resolution susceptibility map-weighted imaging (SMwI). Ninety-six healthy subjects and 50 early-stage IPD patients underwent both a 0.8 × 0.8 × 0.8 mm3 NM-MRI and a 0.5 × 0.5 × 1.0 mm3 multi-echo gradient-recalled echo imaging for SMwI. Both NM-MRI and SMwI templates were created by using image data from the 96 healthy subjects. Permutation-based nonparametric tests were conducted to investigate spatial differences between the two groups in NM-MRI, and the results were displayed on both NM-MRI and SMwI templates. The posterolateral and anteromedial regions of the SNpc in NM-MRI were significantly different between the two groups, corresponding to the nigrosome 1 and nigrosome 2 regions, respectively, on the SMwI template. There were the areas of significant spatial difference in the hypointense SN on SMwI between early-stage IPD patients and healthy subjects. These areas on SMwI were slightly greater than those on NM-MRI, including the areas showing group difference on NM-MRI. Our voxelwise analysis of NM-MRI suggests that two regions (nigrosome 1 and nigrosome 2) of the SNpc are separately affected in early-stage IPD.

Automated assessment of the substantia nigra on susceptibility map-weighted imaging using deep convolutional neural networks for diagnosis of Idiopathic Parkinson's disease.

OBJECTIVES: Despite its use in determining nigrostriatal degeneration, the lack of a consistent interpretation of nigrosome 1 susceptibility map-weighted imaging (SMwI) limits its generalized applicability. To implement and evaluate a diagnostic algorithm based on convolutional neural networks for interpreting nigrosome 1 SMwI for determining nigrostriatal degeneration in idiopathic Parkinson's disease (IPD). METHODS: In this retrospective study, we enrolled 267 IPD patients and 160 control subjects (125 patients with drug-induced parkinsonism and 35 healthy subjects) at our institute, and 24 IPD patients and 27 control subjects at three other institutes on approval of the local institutional review boards. Dopamine transporter imaging served as the reference standard for the presence or absence of abnormalities of nigrosome 1 on SMwI. Diagnostic performance was compared between visual assessment by an experienced neuroradiologist and the developed deep learning-based diagnostic algorithm in both internal and external datasets using a bootstrapping method with 10000 re-samples by the "pROC" package of R (version 1.16.2). RESULTS: The area under the receiver operating characteristics curve (AUC) (95% confidence interval [CI]) per participant by the bootstrap method was not significantly different between visual assessment and the deep learning-based algorithm (internal validation, .9622 [0.8912-1.0000] versus 0.9534 [0.8779-0.9956], P = .1511; external validation, 0.9367 [0.8843-0.9802] versus 0.9208 [0.8634-0.9693], P = .6267), indicative of a comparable performance to visual assessment. CONCLUSIONS: Our deep learning-based algorithm for assessing abnormalities of nigrosome 1 on SMwI was found to have a comparable performance to that of an experienced neuroradiologist.

Unpaired Training of Deep Learning tMRA for Flexible Spatio-Temporal Resolution.

Time-resolved MR angiography (tMRA) has been widely used for dynamic contrast enhanced MRI (DCE-MRI) due to its highly accelerated acquisition. In tMRA, the periphery of the k -space data are sparsely sampled so that neighbouring frames can be merged to construct one temporal frame. However, this view-sharing scheme fundamentally limits the temporal resolution, and it is not possible to change the view-sharing number to achieve different spatio-temporal resolution trade-offs. Although many deep learning approaches have been recently proposed for MR reconstruction from sparse samples, the existing approaches usually require matched fully sampled k -space reference data for supervised training, which is not suitable for tMRA due to the lack of high spatio-temporal resolution ground-truth images. To address this problem, here we propose a novel unpaired training scheme for deep learning using optimal transport driven cycle-consistent generative adversarial network (cycleGAN). In contrast to the conventional cycleGAN with two pairs of generator and discriminator, the new architecture requires just a single pair of generator and discriminator, which makes the training much simpler but still improves the performance. Reconstruction results using in vivo tMRA and simulation data set confirm that the proposed method can immediately generate high quality reconstruction results at various choices of view-sharing numbers, allowing us to exploit better trade-off between spatial and temporal resolution in time-resolved MR angiography.

A Two Cascaded Network Integrating Regional-based YOLO and 3D-CNN for Cerebral Microbleeds Detection.

Cerebral Microbleeds (CMBs) are small chronic brain hemorrhages, which have been considered as diagnostic indicators for different cerebrovascular diseases including stroke, dysfunction, dementia, and cognitive impairment. In this paper, we propose a fully automated two-stage integrated deep learning approach for efficient CMBs detection, which combines a regional-based You Only Look Once (YOLO) stage for potential CMBs candidate detection and three-dimensional convolutional neural networks (3D-CNN) stage for false positives reduction. Both stages are conducted using the 3D contextual information of microbleeds from the MR susceptibility-weighted imaging (SWI) and phase images. However, we average the adjacent slices of SWI and complement the phase images independently and utilize them as a two- channel input for the regional-based YOLO method. The results in the first stage show that the proposed regional-based YOLO efficiently detected the CMBs with an overall sensitivity of 93.62% and an average number of false positives per subject (FPavg) of 52.18 throughout the five-folds cross-validation. The 3D-CNN based second stage further improved the detection performance by reducing the FPavg to 1.42. The outcomes of this work might provide useful guidelines towards applying deep learning algorithms for automatic CMBs detection.

Cortical thinning pattern according to differential nigrosome involvement in patients with Parkinson's disease.

The pathological hallmark of Parkinson's disease (PD) is the progressive degeneration of dopaminergic neurons in the substantia nigra pars compacta, where the dopaminergic neurons form five clusters called nigrosomes 1-5 (N1-N5). N1 is the largest and considered to be the most affected by PD, followed by N2, N4, N3, and N5. Recently, an MRI study suggested a sequential progression of loss from N1 to N4. As the extent of cortical thinning widens as PD progresses, we aimed to define cortical thinning patterns according to the differential involvement of N1 and N4 in PD patients. Cortical thickness was analyzed in 83 PD patients (29 with N1 loss on at least one side of the brain, but no N4 loss; and 54 with N4 loss on at least one side) and 35 healthy subjects with age, sex, disease duration, and intracranial volume as covariates. On patient-wise analysis, for areas with more cortical thinning than the controls, PD patients with N4 loss had wider cortical thinning involving more dorsolateral prefrontal cortex and temporal areas than PD patients with only N1 loss, but cortical thinning did not significantly differ between these two patient groups. However, cortical thinning was more apparent in hemisphere-level analysis with statistically significant clusters being found more in hemispheres with N4 loss than hemispheres with N1 loss in PD patients compared to normal hemispheres of the controls. Cortical thinning occurred in a similar propagation pattern to that seen with PD progression, supporting past hypotheses on the sequential progression of nigrosome loss from N1 to N4.

Multiphase MR Angiography Collateral Map: Functional Outcome after Acute Anterior Circulation Ischemic Stroke.

Background Collateral circulation determines tissue fate and affects treatment result in acute ischemic stroke. A precise method for collateral estimation in an optimal imaging protocol is necessary to make an appropriate treatment decision for acute ischemic stroke. Purpose To verify the value of multiphase collateral imaging data sets (MR angiography collateral map) derived from dynamic contrast material-enhanced MR angiography for predicting functional outcomes after acute ischemic stroke. Materials and Methods This secondary analysis of an ongoing prospective observational study included data from participants with acute ischemic stroke due to occlusion or stenosis of the unilateral internal carotid artery and/or M1 segment of the middle cerebral artery who were evaluated within 8 hours of symptom onset. Data were obtained from March 2016 through August 2018. The collateral grading based on the MR angiography collateral map was estimated by using six-scale MR acute ischemic stroke collateral (MAC) scores. To identify independent predictors of favorable functional outcomes, age, sex, risk factors, baseline National Institutes of Health Stroke Scale (NIHSS) score, baseline diffusion-weighted imaging (DWI) lesion volume, site of steno-occlusion, collateral grade, mode of treatment, and early reperfusion were evaluated with multiple logistic regression analyses. Results One hundred fifty-four participants (mean age ± standard deviation, 69 years ± 13; 99 men) were evaluated. Younger age (odds ratio [OR], 0.45; 95% confidence interval [CI]: 0.29, 0.70; P < .001), lower baseline NIHSS score (OR, 0.85; 95% CI: 0.78, 0.94; P < .001), MAC score of 3 (OR, 27; 95% CI: 4.0, 179; P < .001), MAC score of 4 (OR, 17; 95% CI: 2.1, 134; P = .007), MAC score of 5 (OR, 27; 95% CI: 2.5, 306; P = .007), and successful early reperfusion (OR, 7.5; 95% CI: 2.6, 22; P < .001) were independently associated with favorable functional outcomes in multivariable analysis. There was a linear negative association between collateral perfusion grades and functional outcomes (P < .001). Conclusion An MR angiography collateral map was clinically reliable for collateral estimation in patients with acute ischemic stroke. This map provided patient-specific pacing information for ischemic progression. © RSNA, 2020.

Exploring linearity of deep neural network trained QSM: QSMnet.

Recently, deep neural network-powered quantitative susceptibility mapping (QSM), QSMnet, successfully performed ill-conditioned dipole inversion in QSM and generated high-quality susceptibility maps. In this paper, the network, which was trained by healthy volunteer data, is evaluated for hemorrhagic lesions that have substantially higher susceptibility than healthy tissues in order to test "linearity" of QSMnet for susceptibility. The results show that QSMnet underestimates susceptibility in hemorrhagic lesions, revealing degraded linearity of the network for the untrained susceptibility range. To overcome this limitation, a data augmentation method is proposed to generalize the network for a wider range of susceptibility. The newly trained network, which is referred to as QSMnet+, is assessed in computer-simulated lesions with an extended susceptibility range (-1.4 â€‹ppm to +1.4 â€‹ppm) and also in twelve hemorrhagic patients. The simulation results demonstrate improved linearity of QSMnet+ over QSMnet (root mean square error of QSMnet+: 0.04 â€‹ppm vs. QSMnet: 0.36 â€‹ppm). When applied to patient data, QSMnet+ maps show less noticeable artifacts to those of conventional QSM maps. Moreover, the susceptibility values of QSMnet+ in hemorrhagic lesions are better matched to those of the conventional QSM method than those of QSMnet when analyzed using linear regression (QSMnet+: slope â€‹= â€‹1.05, intercept â€‹= â€‹-0.03, R2 â€‹= â€‹0.93; QSMnet: slope â€‹= â€‹0.68, intercept â€‹= â€‹0.06, R2 â€‹= â€‹0.86), consolidating improved linearity in QSMnet+. This study demonstrates the importance of the trained data range in deep neural network-powered parametric mapping and suggests the data augmentation approach for generalization of network. The new network can be applicable for a wide range of susceptibility quantification.

Automated detection of cerebral microbleeds in MR images: A two-stage deep learning approach.

Cerebral Microbleeds (CMBs) are small chronic brain hemorrhages, which have been considered as diagnostic indicators for different cerebrovascular diseases including stroke, dysfunction, dementia, and cognitive impairment. However, automated detection and identification of CMBs in Magnetic Resonance (MR) images is a very challenging task due to their wide distribution throughout the brain, small sizes, and the high degree of visual similarity between CMBs and CMB mimics such as calcifications, irons, and veins. In this paper, we propose a fully automated two-stage integrated deep learning approach for efficient CMBs detection, which combines a regional-based You Only Look Once (YOLO) stage for potential CMBs candidate detection and three-dimensional convolutional neural networks (3D-CNN) stage for false positives reduction. Both stages are conducted using the 3D contextual information of microbleeds from the MR susceptibility-weighted imaging (SWI) and phase images. However, we average the adjacent slices of SWI and complement the phase images independently and utilize them as a two-channel input for the regional-based YOLO method. This enables YOLO to learn more reliable and representative hierarchal features and hence achieve better detection performance. The proposed work was independently trained and evaluated using high and low in-plane resolution data, which contained 72 subjects with 188 CMBs and 107 subjects with 572 CMBs, respectively. The results in the first stage show that the proposed regional-based YOLO efficiently detected the CMBs with an overall sensitivity of 93.62% and 78.85% and an average number of false positives per subject (FPavg) of 52.18 and 155.50 throughout the five-folds cross-validation for both the high and low in-plane resolution data, respectively. These findings outperformed results by previously utilized techniques such as 3D fast radial symmetry transform, producing fewer FPavg and lower computational cost. The 3D-CNN based second stage further improved the detection performance by reducing the FPavg to 1.42 and 1.89 for the high and low in-plane resolution data, respectively. The outcomes of this work might provide useful guidelines towards applying deep learning algorithms for automatic CMBs detection.