Semi-automatic computed tomography angiography quantification assessment is an alternative method to digital subtraction angiography in intracranial stenosis: a multicenter study.

Background: The recent randomized controlled trials studying intracranial atherosclerotic stenosis (ICAS) have used digital subtraction angiography (DSA) to quantify stenosis and enroll patients. However, some disadvantages of DSA such as invasive features, contrast agent overuse, and X-ray radiation overexposure, were not considered in these studies. This study aimed to explore whether computed tomography angiography (CTA) with semi-automatic analysis could be an alternative method to DSA in quantifying the absolute stenotic degree in clinical trials. Methods: Patients with 50-99% ICAS were consecutively screened, prospectively enrolled, and underwent CTA and DSA between March 2021 and December 2021 at 6 centers. This study was registered at www.chictr.org.cn (ChiCTR2100052925). The absolute stenotic degree of ICAS on CTA with semi-automatic analysis was calculated by several protocols using minimal/maximum/mean diameters of stenosis and reference site from a semi-automatic analysis software. Intraclass correlation coefficient (ICC) was used to evaluate the reliabilities of quantifying stenotic degree on CTA. The optimal protocol for quantifying ICAS on CTA was explored. The agreements of quantifying ICAS in calcified or non-calcified lesions and 50-69% or 70-99% stenosis on CTA and DSA were assessed. Results: A total of 191 participants (58.8±10.7 years; 148 men) with 202 lesions were enrolled. The optimal protocol for quantifying ICAS on CTA was calculated as (1 - the minimal diameter of stenosis/the mean diameter of reference) × 100% for its highest agreement with DSA [ICC, 0.955, 95% confidence interval (CI): 0.944-0.966, P<0.001]. Among the 202 lesions, 80.2% (162/202) exhibited severe stenosis on DSA. The accuracy of CTA in detecting severe ICAS was excellent (sensitivity =95.1%, positive predictive value =98.1%). The agreements between DSA and CTA in non-calcified lesions (ICC, 0.960 vs. 0.849) and severe stenosis (ICC, 0.918 vs. 0.841) were higher than those in calcified lesions and moderate stenosis. Conclusions: CTA with semi-automatic analysis demonstrated an excellent agreement with DSA in quantifying ICAS, making it promising to replace DSA for the measurement of absolute stenotic degree in clinical trials.

Impact of the Alberta Stroke Program CT Score subregions on long-term functional outcomes in acute ischemic stroke: Results from two multicenter studies in China.

Background and Objectives: The Alberta Stroke Program CT Score (ASPECTS) is a widely used rating system for assessing infarct extent and location. We aimed to investigate the prognostic value of ASPECTS subregions' involvement in the long-term functional outcomes of acute ischemic stroke (AIS). Materials and Methods: Consecutive patients with AIS and anterior circulation large-vessel stenosis and occlusion between January 2019 and December 2020 were included. The ASPECTS score and subregion involvement for each patient was assessed using posttreatment magnetic resonance diffusion-weighted imaging. Univariate and multivariable regression analyses were conducted to identify subregions related to 3-month poor functional outcome (modified Rankin Scale scores, 3-6) in the reperfusion and medical therapy cohorts, respectively. In addition, prognostic efficiency between the region-based ASPECTS and ASPECTS score methods were compared using receiver operating characteristic curves and DeLong's test. Results: A total of 365 patients (median age, 64 years; 70% men) were included, of whom 169 had poor outcomes. In the reperfusion therapy cohort, multivariable regression analyses revealed that the involvement of the left M4 cortical region in left-hemisphere stroke (adjusted odds ratio [aOR] 5.39, 95% confidence interval [CI] 1.53-19.02) and the involvement of the right M3 cortical region in right-hemisphere stroke (aOR 4.21, 95% CI 1.05-16.78) were independently associated with poor functional outcomes. In the medical therapy cohort, left-hemisphere stroke with left M5 cortical region (aOR 2.87, 95% CI 1.08-7.59) and caudate nucleus (aOR 3.14, 95% CI 1.00-9.85) involved and right-hemisphere stroke with right M3 cortical region (aOR 4.15, 95% CI 1.29-8.18) and internal capsule (aOR 3.94, 95% CI 1.22-12.78) affected were related to the increased risks of poststroke disability. In addition, region-based ASPECTS significantly improved the prognostic efficiency compared with the conventional ASPECTS score method. Conclusion: The involvement of specific ASPECTS subregions depending on the affected hemisphere was associated with worse functional outcomes 3 months after stroke, and the critical subregion distribution varied by clinical management. Therefore, region-based ASPECTS could provide additional value in guiding individual decision making and neurological recovery in patients with AIS.

Imaging Assessments and Clinical Significance of Brain Frailty in Moyamoya Disease.

BACKGROUND AND PURPOSE: Imaging assessment of brain frailty in ischemic stroke has been extensively studied, while the correlation between brain frailty and Moyamoya disease remains obscure. This study aimed to investigate the imaging characteristics of brain frailty and its clinical applications in Moyamoya disease. MATERIALS AND METHODS: This study included 60 patients with Moyamoya disease (107 hemispheres). All patients were divided into stroke and nonstroke groups based on clinical symptoms and imaging findings. The modified brain frailty score was adapted to consider 4 imaging signs: white matter hyperintensity, enlargement of perivascular space, old vascular lesions, and cerebral microbleed. The relative CBF of the MCA territory was quantified using pseudocontinuous arterial-spin labeling. Surgical outcome after revascularization surgery was defined by the Matsushima grade. RESULTS: The relative CBF of the MCA territory decreased as the modified brain frailty score and periventricular white matter hyperintensity grades increased (ρ = -0.22, P = .02; ρ = -0.27, P = .005). Clinically, the modified brain frailty score could identify patients with Moyamoya disease with stroke (OR = 2.00, P = .02). Although the modified brain frailty score showed no predictive value for surgical outcome, basal ganglia enlargement of the perivascular space had a significant correlation with the postoperative Matsushima grade (OR = 1.29, P = .03). CONCLUSIONS: The modified brain frailty score could reflect a cerebral perfusion deficit and clinical symptoms of Moyamoya disease, and its component basal ganglia enlargement of perivascular space may be a promising marker to predict surgical outcome and thus aid future clinical decision-making.

Deep medullary vein damage correlates with small vessel disease in small vessel occlusion acute ischemic stroke.

OBJECTIVES: We aim to investigate whether cerebral small vessel disease (cSVD) imaging markers correlate with deep medullary vein (DMV) damage in small vessel occlusion acute ischemic stroke (SVO-AIS) patients. METHODS: The DMV was divided into six segments according to the regional anatomy. The total DMV score (0-18) was calculated based on segmental continuity and visibility. The damage of DMV was grouped according to the quartiles of the total DMV score. Neuroimaging biomarkers of cSVD including white matter hyperintensity (WMH), cerebral microbleed (CMB), perivascular space (PVS), and lacune were identified. The cSVD score were further analyzed. RESULTS: We included 229 SVO-AIS patients, the mean age was 63.7 ± 23.1 years, the median NIHSS score was 3 (IQR, 2-6). In the severe DMV burden group (the 4th quartile), the NIHSS score grade (6 (3-9)) was significantly higher than other groups (p < 0.01). The grade scores for basal ganglia PVS (BG-PVS) were positively correlated with the degree of DMV (R = 0.67, p < 0.01), rather than centrum semivole PVS (CS-PVS) (R = 0.17, p = 0.1). In multivariate analysis, high CMB burden (adjusted odds ratio [aOR], 25.38; 95% confidence interval [CI], 1.87-345.23) was associated with severe DMV scores. In addition, BG-PVS was related to severe DMV burden in a dose-dependent manner: when BG-PVS score was 3 and 4, the aORs of severe DMV burden were 18.5 and 12.19, respectively. CONCLUSION: The DMV impairment was associated with the severity of cSVD, which suggests that DMV burden may be used for risk stratification in SVO-AIS patients. CLINICAL RELEVANCE STATEMENT: The DMV damage score, based on the association between small vessel disease and the deep medullary veins impairment, is a potential new imaging biomarker for the prognosis of small vessel occlusion acute ischemic stroke, with clinical management implications. KEY POINTS: • The damage to the deep medullary vein may be one mechanism of cerebral small vessel disease. • Severe burden of the basal ganglia perivascular space and cerebral microbleed is closely associated with significant impairment to the deep medullary vein. • The deep medullary vein damage score may reflect a risk of added vascular damage in small vessel occlusion acute ischemic stroke patients.

Synthesized 7T MPRAGE From 3T MPRAGE Using Generative Adversarial Network and Validation in Clinical Brain Imaging: A Feasibility Study.

BACKGROUND: Ultra-high field 7T MRI can provide excellent tissue contrast and anatomical details, but is often cost prohibitive, and is not widely accessible in clinical practice. PURPOSE: To generate synthetic 7T images from widely acquired 3T images with deep learning and to evaluate the feasibility of this approach for brain imaging. STUDY TYPE: Prospective. POPULATION: 33 healthy volunteers and 89 patients with brain diseases, divided into training, and evaluation datasets in the ratio 4:1. SEQUENCE AND FIELD STRENGTH: T1-weighted nonenhanced or contrast-enhanced magnetization-prepared rapid acquisition gradient-echo sequence at both 3T and 7T. ASSESSMENT: A generative adversarial network (SynGAN) was developed to produce synthetic 7T images from 3T images as input. SynGAN training and evaluation were performed separately for nonenhanced and contrast-enhanced paired acquisitions. Qualitative image quality of acquired 3T and 7T images and of synthesized 7T images was evaluated by three radiologists in terms of overall image quality, artifacts, sharpness, contrast, and visualization of vessel using 5-point Likert scales. STATISTICAL TESTS: Wilcoxon signed rank tests to compare synthetic 7T images with acquired 7T and 3T images and intraclass correlation coefficients to evaluate interobserver variability. P < 0.05 was considered significant. RESULTS: Of the 122 paired 3T and 7T MRI scans, 66 were acquired without contrast agent and 56 with contrast agent. The average time to generate synthetic images was ~11.4 msec per slice (2.95 sec per participant). The synthetic 7T images achieved significantly improved tissue contrast and sharpness in comparison to 3T images in both nonenhanced and contrast-enhanced subgroups. Meanwhile, there was no significant difference between acquired 7T and synthetic 7T images in terms of all the evaluation criteria for both nonenhanced and contrast-enhanced subgroups (P ≥ 0.180). DATA CONCLUSION: The deep learning model has potential to generate synthetic 7T images with similar image quality to acquired 7T images. LEVEL OF EVIDENCE: 2 TECHNICAL EFFICACY: Stage 1.

MRI Assessment of Brain Frailty and Clinical Outcome in Patients With Acute Posterior Perforating Artery Infarction.

BACKGROUND: Global brain health has gained increasing attention recently. Imaging markers of brain frailty have been related to functional outcomes in previous studies on anterior circulation; however, little data are available on imaging markers and posterior circulation. PURPOSE: To investigate the impact of brain frailty on functional outcomes in patients with acute perforating artery infarction (PAI) of the posterior circulation. STUDY TYPE: Prospective. POPULATION: One hundred patients (60.78 ± 9.51 years, 72% men) with acute posterior circulation PAI (determined by diffusion-weighted magnetic resonance imaging (MRI)/time-of-flight MR angiography). FIELD STRENGTH/SEQUENCE: T1- and T2-weighted fast spin echo, T2-weighted fluid-attenuated inversion recovery, diffusion-weighted echo planar, gradient echo (susceptibility-weight imaging), and 3D time-of-flight MR angiography sequences at 3.0 T. ASSESSMENT: Periventricular and deep white matter hyperintensities (WMH), enlarged perivascular spaces (EPVS) in the basal ganglia and centrum semiovale area, lacunes, cerebral microbleeds (CMB), and total brain frailty score by calculating the above imaging characters were rated visually by three radiologists with 9, 10, and 11 years of experience and one neuroradiologist with 12. Infarction volume was assessed using baseline diffusion-weighted imaging (DWI) data obtained within 24 hours of symptom onset. A modified Rankin Scale (mRS) score >1 on day 90 defined an adverse functional outcome. Associations between the imaging markers of brain frailty and functional outcomes were assessed. STATISTICAL TESTS: Fisher's exact test, Mann-Whitney U test, and multivariable binary logistic regression. A P value <0.05 was considered statistically significant. RESULTS: Adverse prognoses (mRS > 1) were observed in 34 (34%) patients. Infarction volume, periventricular WMH, deep WMH, basal ganglia EPVS, CMB, and the brain frailty score were significantly associated with adverse functional outcomes. An increased brain frailty score was significantly associated with unfavorable mRS score on day 90 (odds ratio 1.773, 95% confidence interval 1.237-2.541). DATA CONCLUSION: Advanced MRI imaging markers of brain frailty, individually or combined as a total brain frailty score, were associated with worse functional outcomes after acute posterior circulation PAI. LEVEL OF EVIDENCE: 3 TECHNICAL EFFICACY: Stage 3.

Generative Adversarial Network-based Noncontrast CT Angiography for Aorta and Carotid Arteries.

Background Iodinated contrast agents (ICAs), which are widely used in CT angiography (CTA), may cause adverse effects in humans, and their use is time-consuming and costly. Purpose To develop an ICA-free deep learning imaging model for synthesizing CTA-like images and to assess quantitative and qualitative image quality as well as the diagnostic accuracy of synthetic CTA (Syn-CTA) images. Materials and Methods A generative adversarial network (GAN)-based CTA imaging model was trained, validated, and tested on retrospectively collected pairs of noncontrast CT and CTA images of the neck and abdomen from January 2017 to June 2022, and further validated on an external data set. Syn-CTA image quality was evaluated using quantitative metrics. In addition, two senior radiologists scored the visual quality on a three-point scale (3 = good) and determined the vascular diagnosis. The validity of Syn-CTA images was evaluated by comparing the visual quality scores and diagnostic accuracy of aortic and carotid artery disease between Syn-CTA and real CTA scans. Results CT scans from 1749 patients (median age, 60 years [IQR, 50-68 years]; 1057 male patients) were included in the internal data set: 1137 for training, 400 for validation, and 212 for testing. The external validation set comprised CT scans from 42 patients (median age, 67 years [IQR, 59-74 years]; 37 male patients). Syn-CTA images had high similarity to real CTA images (normalized mean absolute error, 0.011 and 0.013 for internal and external test set, respectively; peak signal-to-noise ratio, 32.07 dB and 31.58 dB; structural similarity, 0.919 and 0.906). The visual quality of Syn-CTA and real CTA images was comparable (internal test set, P = .35; external validation set, P > .99). Syn-CTA showed reasonable to good diagnostic accuracy for vascular diseases (internal test set: accuracy = 94%, macro F1 score = 91%; external validation set: accuracy = 86%, macro F1 score = 83%). Conclusion A GAN-based model that synthesizes neck and abdominal CTA-like images without the use of ICAs shows promise in vascular diagnosis compared with real CTA images. Clinical trial registration no. NCT05471869 © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Zhang and Turkbey in this issue.

Relay learning: a physically secure framework for clinical multi-site deep learning.

Big data serves as the cornerstone for constructing real-world deep learning systems across various domains. In medicine and healthcare, a single clinical site lacks sufficient data, thus necessitating the involvement of multiple sites. Unfortunately, concerns regarding data security and privacy hinder the sharing and reuse of data across sites. Existing approaches to multi-site clinical learning heavily depend on the security of the network firewall and system implementation. To address this issue, we propose Relay Learning, a secure deep-learning framework that physically isolates clinical data from external intruders while still leveraging the benefits of multi-site big data. We demonstrate the efficacy of Relay Learning in three medical tasks of different diseases and anatomical structures, including structure segmentation of retina fundus, mediastinum tumors diagnosis, and brain midline localization. We evaluate Relay Learning by comparing its performance to alternative solutions through multi-site validation and external validation. Incorporating a total of 41,038 medical images from 21 medical hosts, including 7 external hosts, with non-uniform distributions, we observe significant performance improvements with Relay Learning across all three tasks. Specifically, it achieves an average performance increase of 44.4%, 24.2%, and 36.7% for retinal fundus segmentation, mediastinum tumor diagnosis, and brain midline localization, respectively. Remarkably, Relay Learning even outperforms central learning on external test sets. In the meanwhile, Relay Learning keeps data sovereignty locally without cross-site network connections. We anticipate that Relay Learning will revolutionize clinical multi-site collaboration and reshape the landscape of healthcare in the future.

Automatic intracranial abnormality detection and localization in head CT scans by learning from free-text reports.

Deep learning has yielded promising results for medical image diagnosis but relies heavily on manual image annotations, which are expensive to acquire. We present Cross-DL, a cross-modality learning framework for intracranial abnormality detection and localization in head computed tomography (CT) scans by learning from free-text imaging reports. Cross-DL has a discretizer that automatically extracts discrete labels of abnormality types and locations from reports, which are utilized to train an image analyzer by a dynamic multi-instance learning approach. Benefiting from the low annotation cost and a consequent large-scale training set of 28,472 CT scans, Cross-DL achieves accurate performance, with an average area under the receiver operating characteristic curve (AUROC) of 0.956 (95% confidence interval: 0.952-0.959) in detecting 4 abnormality types in 17 regions while accurately localizing abnormalities at the voxel level. An intracranial hemorrhage classification experiment on the external dataset CQ500 achieves an AUROC of 0.928 (0.905-0.951). The model can also help review prioritization.

Resting-state functional magnetic resonance imaging study comparing tremor-dominant and postural instability/gait difficulty subtypes of Parkinson's disease.

PURPOSE: The symptom-specific intrinsic neural mechanisms underlying Parkinson's disease (PD) subtypes (tremor dominant [TD] and postural instability gait difficulty [PIGD]) remain unclarified. We examined spontaneous brain activity patterns in TD and PIGD. MATERIAL AND METHODS: We included 49 patients with PD (21 with TD/28 with PIGD) and 32 healthy controls (HCs) in this study. We conducted analysis of variance and post-hoc analyses of the amplitude of low-frequency fluctuation (ALFF) and regional homogeneity (ReHo) values of the three groups, with age, sex, and gray matter volume as covariates, and a relationship analysis of the ALFF and ReHo values with clinical variables. RESULTS: In comparison with HCs, PIGD PD patients had increased ALFF values in the right middle occipital gyrus and left superior occipital gyrus and decreased values primarily in the bilateral inferior frontal gyrus (triangular part). TD PD patients had lower ALFF values in the right inferior frontal gyrus (triangular part) and left insula. In comparison to TD PD patients, PIGD PD patients had higher ALFF values in the left middle occipital gyrus and left superior occipital gyrus. In contrast to HCs, TD PD patients demonstrated a reduction of ReHo values in the left middle temporal gyrus, and PIGD patients showed a decrease of ReHo values in the left inferior temporal gyrus. CONCLUSION: ALFF values increased in the occipital gyrus of the PIGD PD patients, thus providing evidence of a compensatory mechanism of altered motor function in comparison with the TD PD patients.

Apparent Diffusion Coefficient as Imaging Biomarker for Identifying IDH Mutation, 1p19q Codeletion, and MGMT Promoter Methylation Status in Patients With Glioma.

BACKGROUND: Glioma genotypes are of importance for clinical decision-making. This data can only be acquired through histopathological analysis based on resection or biopsy. Consequently, there is a need for alternative biomarkers that noninvasively provide reliable information for preoperatively identifying molecular characteristics. PURPOSE: To investigate apparent diffusion coefficient (ADC) as imaging biomarker for preoperatively identifying glioma genotypes based on the 2021 World Health Organization (WHO) classification of central nervous system (CNS) tumors. STUDY TYPE: Retrospective. SUBJECTS: One hundred and fifty-nine patients (47.6 ± 14.4 years) diagnosed with WHO grade 2-4 glioma including 93 males and 66 females. FIELD STRENGTH/SEQUENCE: A 3 T/spin echo echo planner imaging. ASSESSMENT: The ADC measurements were assessed by two neuroradiologists (both with 6 years of experience). Three different lowest portions inside the tumors without overlap were manually drawn on the ADC maps as regions of interest (ROIs). The mean ADC value of the three ROIs was defined as the minimum ADC value (ADCmin ). An ROI was placed in the contralateral normal appearing white matter (NAWM) to obtain the ADC value (ADCNAWM ). The ADCmin to ADCNAWM ratio (ADCratio ) was calculated. Genetics results were retrospectively recorded from pathologic and genetic test reports. STATISTICAL TESTS: Two-sample independent t-tests, receiver operating characteristic curve analysis, and intraclass correlation coefficient analysis were used. Statistical significance was set at P < 0.05. RESULTS: Isocitrate dehydrogenase (IDH)-mutated glioma showed higher ADCmin and ADCratio than IDH wild-type glioma. Among IDH-mutated glioma, higher ADCmin and ADCratio were found in 1p19q intact glioma than in 1p19q codeletion glioma. ADC parameters enabled differentiation of IDH mutation status with area under the curve (AUC) of 0.84 and 0.86. DATA CONCLUSION: ADC has potential discriminative value for IDH mutation and 1p19q codeletion status. EVIDENCE LEVEL: 3. TECHNICAL EFFICACY: Stage 2.

Small vessel disease burden predicts functional outcomes in patients with acute ischemic stroke using machine learning.

AIMS: Our purpose is to assess the role of cerebral small vessel disease (SVD) in prediction models in patients with different subtypes of acute ischemic stroke (AIS). METHODS: We enrolled 398 small-vessel occlusion (SVO) and 175 large artery atherosclerosis (LAA) AIS patients. Functional outcomes were assessed using the modified Rankin Scale (mRS) at 90 days. MRI was performed to assess white matter hyperintensity (WMH), perivascular space (PVS), lacune, and cerebral microbleed (CMB). Logistic regression (LR) and machine learning (ML) were used to develop predictive models to assess the influences of SVD on the prognosis. RESULTS: In the feature evaluation of SVO-AIS for different outcomes, the modified total SVD score (Gain: 0.38, 0.28) has the maximum weight, and periventricular WMH (Gain: 0.07, 0.09) was more important than deep WMH (Gain: 0.01, 0.01) in prognosis. In SVO-AIS, SVD performed better than regular clinical data, which is the opposite of LAA-AIS. Among all models, eXtreme gradient boosting (XGBoost) method with optimal index (OI) has the best performance to predict excellent outcome in SVO-AIS. [0.91 (0.84-0.97)]. CONCLUSIONS: Our results revealed that different SVD markers had distinct prognostic weights in AIS patients, and SVD burden alone may accurately predict the SVO-AIS patients' prognosis.

Arterial Spin Labeling-Based MRI Estimation of Penumbral Tissue in Acute Ischemic Stroke.

BACKGROUND: Arterial spin labeling (ASL) has shown potential for the assessment of penumbral tissue in patients with acute ischemic stroke (AIS). The postlabeling delay (PLD) parameter is sensitive to arterial transit delays and influences cerebral blood flow measurements. PURPOSE: To assess the impact of ASL acquisition at different PLDs for penumbral tissue quantification and to compare their performance regarding assisting patient selection for endovascular treatment with dynamic susceptibility contrast MRI (DSC-MRI) as the reference method. STUDY TYPE: Retrospective. POPULATION: A total of 53 patients (59.98 ± 12.60 years, 32% women) with AIS caused by internal carotid or middle cerebral artery occlusion. FIELD STRENGTH/SEQUENCE: A 3-T, three-dimensional pseudo-continuous ASL with fast-spin echo readout. ASSESSMENT: Hypoperfusion volume was measured using DSC-MRI and ASL with PLDs of 1.500 msec and 2.500 msec, respectively. Eligibility for endovascular treatment was retrospectively determined according to the imaging criteria of the Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke trial (DEFUSE 3). STATISTICAL TESTS: Kruskal-Wallis tests, Bland-Altman plots, Cohen's kappa, and receiver operating characteristic analyses were used. The threshold for statistical significance was set at P Ë‚ 0.05. RESULTS: Hypoperfusion volume for ASL with a PLD of 1.500 msec was significantly larger than that for DSC-MRI, while the hypoperfusion volume for a PLD of 2.500 msec was not significantly different from that of DSC-MRI (P = 0.435). Bland-Altman plots showed that the mean volumetric error between the hypoperfusion volume measured by DSC-MRI and ASL with PLDs of 1.500/2.500 msec was -107.0 mL vs. 4.49 mL. Cohen's kappa was 0.679 vs. 0.773 for DSC-MRI and ASL, respectively, with a PLD of 1.500/2.500 msec. The sensitivity and specificity for ASL with a PLD of 1.500/2.500 msec in identifying patients eligible for treatment were 89.74% vs. 97.44% and 92.86% vs. 64.29%, respectively. DATA CONCLUSION: In AIS, PLDs for ASL acquisition may have a considerable impact on the quantification of the hypoperfusion volume. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 2.

Evaluation of postcontrast images of intracranial tumors at 7T and 3T MRI: An intra-individual comparison study.

AIM: This study aimed to evaluate the diagnostic value of ultrahigh-field magnetic resonance imaging (MRI) for brain tumors in clinical practice. METHODS: Thirty patients with brain tumors underwent 7- and 3-T MRI. The performance and diagnostic confidence of 7- and 3-T MRI in the visualization of tumor details such as internal structure and feeding artery were evaluated by radiologists. Contrast-enhanced region performance and tumor detail diagnostic confidence score (DCS) were calculated and compared between 7 and 3T using Wilcoxon rank sum test. RESULTS: In 19 with obvious enhancement and 11 cases without obvious enhancement, 7- and 3-T MRI showed similar performance. The tumors' internal structure and feeding artery were more clearly depicted by 7-T MRI (62.2% and 54.4%, respectively) than by 3-T MRI (2.2% and 6.7%, respectively). Furthermore, the mean DCSs of both internal structure and feeding artery were higher at 7T than at 3T (internal structure: 16.29 ± 9.67 vs. -5.79 ± 4.12, p = 0.028; feeding artery: 21.96 ± 6.93 vs. 4.46 ± 7.07, p = 0.028). The DCS was more significantly improved in the senior radiologist group. CONCLUSION: Better visualization of brain tumor details and higher tumor detail diagnostic confidence can be obtained with 7-T MRI.

Association of fluid-attenuated inversion recovery vascular hyperintensity with ischaemic events in internal carotid artery or middle cerebral artery occlusion.

BACKGROUND AND PURPOSE: Individuals with intracranial artery occlusion have high rates of ischaemic events and recurrence. It has been challenging to identify patients who had high-risk stroke using a simple, valid and non-invasive screening approach. This study aimed to investigate whether fluid-attenuated inversion recovery (FLAIR) vascular hyperintensity (FVH), a specific imaging sign on the FLAIR sequence, could be a predictor of ischaemic events in a population with internal carotid artery (ICA) or middle cerebral artery (MCA) occlusion. METHODS: We retrospectively analysed 147 patients (mean 60.43±12.83 years) with 149 lesions, including 37 asymptomatic and 112 symptomatic cases of ICA or MCA occlusion. Symptomatic occlusion was considered if ischaemic events were present in the relevant territory within 90 days. FVH Alberta Stroke Program Early Computed Tomography Score (FVH-ASPECTS: 0-7, with 0 indicating absence of FVH and 7 suggesting prominent FVH) and collateral circulation grade were assessed for each participant. Multivariable logistic regression analysis was performed to detect independent markers associated with symptomatic status. RESULTS: A lower FVH-ASPECTS was associated with a more favourable collateral circulation grade (rho=-0.464, p<0.0001). The FVH-ASPECTS was significantly lower in the asymptomatic occlusion group than in the symptomatic occlusion group (p<0.0001). FVH-ASPECTS (Odd ratio, 2.973; 95% confidence interval, 1.849 to 4.781; p<0.0001) was independently associated with symptomatic status after adjustment for age, sex, lesion location and collateral circulation grade in the multivariate logistic regression. The area under the curve was 0.861 for the use of FVH-ASPECTS to identify symptomatic occlusion. CONCLUSIONS: The ability to discriminate symptomatic from asymptomatic occlusion suggests that FVH may be a predictor of stroke. As a simple imaging sign, FVH may serve as a surrogate for haemodynamic impairments and can be used to identify high-risk stroke cases early in ICA or MCA occlusion.

Collateral-Core Ratio as a Novel Predictor of Clinical Outcomes in Acute Ischemic Stroke.

The interaction effect between collateral circulation and ischemic core size on stroke outcomes has been highlighted in acute ischemic stroke (AIS). However, biomarkers that assess the magnitude of this interaction are still lacking. We aimed to present a new imaging marker, the collateral-core ratio (CCR), to quantify the interaction effect between these factors and evaluate its ability to predict functional outcomes using machine learning (ML) in AIS. Patients with AIS caused by anterior circulation large vessel occlusion (LVO) were recruited from a prospective multicenter study. CCR was calculated as collateral perfusion volume/ischemic core volume. Functional outcomes were assessed using the modified Rankin Scale (mRS) at 90 days. An ML model was built and tested with a tenfold cross-validation using nine clinical and four imaging variables with mRS score 3-6 as unfavorable outcomes. Among 129 patients, CCR was identified as the most important variable. The prediction model incorporating clinical factors, ischemic core volume, collateral perfusion volume, and CCR showed better discriminatory power in predicting unfavorable outcomes than the model without CCR (mean C index 0.853 ± 0.108 versus 0.793 ± 0.133, P = 0.70; mean net reclassification index 52.7% ± 32.7%, P < 0.05). When patients were divided into two groups based on their CCR value with a threshold of 0.73, unfavorable outcomes were significantly more prevalent in patients with CCR ≤ 0.73 than in those with CCR > 0.73. CCR is a robust predictor of functional outcomes, as identified by ML, in patients with acute LVO. The prediction model that incorporated CCR improved the model's ability to identify unfavorable outcomes. ClinicalTrials.gov Identifier: NCT02580097.

Evaluation of an automated intracranial aneurysm detection and rupture analysis approach using cascade detection and classification networks.

Intracranial aneurysm is commonly found in human brains especially for the elderly, and its rupture accounts for a high rate of subarachnoid hemorrhages. However, it is time-consuming and requires special expertise to pinpoint small aneurysms from computed tomography angiography (CTA) images. Deep learning-based detection has helped improve much efficiency but false-positives still render difficulty to be ruled out. To study the feasibility of deep learning algorithms for aneurysm analysis in clinical applications, this paper proposes a pipeline for aneurysm detection, segmentation, and rupture classification and validates its performance using CTA images of 1508 subjects. A cascade aneurysm detection model is employed by first using a fine-tuned feature pyramid network (FPN) for candidate detection and then applying a dual-channel ResNet aneurysm classifier to further reduce false positives. Detected aneurysms are then segmented by applying a traditional 3D V-Net to their image patches. Radiomics features of aneurysms are extracted after detection and segmentation. The machine-learning-based and deep learning-based rupture classification can be used to distinguish ruptured and un-ruptured ones. Experimental results show that the dual-channel ResNet aneurysm classifier utilizing image and vesselness information helps boost sensitivity of detection compared to single image channel input. Overall, the proposed pipeline can achieve a sensitivity of 90 % for 1 false positive per image, and 95 % for 2 false positives per image. For rupture classification the area under curve (AUC) of 0.906 can be achieved for the testing dataset. The results suggest feasibility of the pipeline for potential clinical use to assist radiologists in aneurysm detection and classification of ruptured and un-ruptured aneurysms.

Deep learning with weak annotation from diagnosis reports for detection of multiple head disorders: a prospective, multicentre study.

BACKGROUND: A large training dataset with high-quality annotations is necessary for building an accurate and generalisable deep learning system, which can be difficult and expensive to prepare in medical applications. We present a novel deep-learning-based system, requiring no annotator but weak annotation from a diagnosis report, for accurate and generalisable performance in detecting multiple head disorders from CT scans, including ischaemia, haemorrhage, tumours, and skull fractures. METHODS: Our system was developed on 104 597 head CT scans from the Chinese PLA General Hospital, with associated textual diagnosis reports. Without expert annotation, we used keyword matching on the reports to automatically generate disorder labels for each scan. The labels were inaccurate because of the unreliable annotator-free strategy and inexact because of scan-level annotation. We proposed RoLo, a novel weakly supervised learning algorithm, with a noise-tolerant mechanism and a multi-instance learning strategy to address these issues. RoLo was tested on retrospective (2357 scans from the Chinese PLA General Hospital), prospective (650 scans from the Chinese PLA General Hospital), cross-centre (1525 scans from the Brain Hospital of Hunan Province), cross-equipment (1484 scans from the Chinese PLA General Hospital), and cross-nation (CQ500 public dataset from India) test datasets. Four radiologists were tested on the prospective test dataset before and after viewing system recommendations to assess whether the system could improve diagnostic performance. FINDINGS: The area under the receiver operating characteristic curve for detecting the four disorder types was 0·976 (95% CI 0·976-0·976) for retrospective, 0·975 (0·974-0·976) for prospective, 0·965 (0·964-0·966) for cross-centre, and 0·971 (0·971-0·972) for cross-equipment test datasets, and 0·964 (0·964-0·966) for CQ500 (with only haemorrhage and fracture). The system achieved similar performance to four radiologists and helped to improve sensitivity and specificity by 0·109 (95% CI 0·086-0·131) and 0·022 (0·017-0·026), respectively. INTERPRETATION: Without expert annotated data, our system achieved accurate and generalisable performance for head disorder detection. The system improved the diagnostic performance of radiologists. Because of its accuracy and generalisability, our computer-aided diganostic system could be used in clinical practice to improve the accuracy and efficiency of radiologists in different hospitals. FUNDING: National Key R&D Program of China, National Natural Science Foundation of China, and Beijing Natural Science Foundation.

FLAIR vessel hyperintensities predict functional outcomes in patients with acute ischemic stroke treated with medical therapy.

OBJECTIVES: The prognostic value of fluid-attenuated inversion recovery vessel hyperintensity (FVH) remains controversial in acute ischemic stroke (AIS). The objective was to investigate whether the presence of FVH could predict long-term functional outcomes in patients with AIS receiving medical therapy. METHODS: Consecutive AIS patients with anterior circulation large vessel stenosis (LVS) in multiple centers between January 2019 and December 2020 were studied. Presence of FVH was identified and evaluated as FVH (+). Quantification of FVH was performed using an FVH-Alberta Stroke Program Early CT Score (ASPECTS) system and divided into grades: FVH-ASPECTS of 0 = grade 0; 1-2 = grade 1; 3-7 = grade 2. Poor functional outcome was defined as modified Rankin scale > 2 at 3 months. RESULTS: Overall, 175 patients were analyzed (age, 64.31 ± 13.47 years; men, 65.1%), and 78.9% patients presented with FVH. Larger infarct volume (19.90 mL vs. 5.50 mL, p < 0.001), higher rates of FVH (+) (92.0% vs. 65.9%, p < 0.001), and higher FVH grades (grade 2, 34.5% vs. 10.2%, p < 0.001) were more prone to be observed in patients with poor functional outcomes. FVH (+) with infarct volume larger than 6.265 mL (adjusted odds ratio [aOR] 6.03, 95% confidence interval [CI] 1.82-19.98) and FVH grade (grade 1, aOR 3.07, 95% CI 1.12-8.43; grade 2, aOR 5.80, 95% CI 1.59-21.11) were independently associated with poor functional outcomes. CONCLUSION: FVH (+) combined with large infarct volume and high FVH grade can predict poor long-term functional outcomes in patients with LVS who receive medical therapy. KEY POINTS: • FVH is expected to be a contrast agent-independent alternative for assessing hemodynamic status in the acute stage of stroke. • FVH (+) and high FVH grade, quantified by FVH-ASPECTS rating system and grades, are associated with large infarct volume. • The combination of FVH and DWI-based infarct volume has independent predictive value for long-term functional outcomes in AIS patients with large artery stenosis treated with medical therapy.

Accelerating susceptibility-weighted imaging with deep learning by complex-valued convolutional neural network (ComplexNet): validation in clinical brain imaging.

OBJECTIVES: Susceptibility-weighted imaging (SWI) is crucial for the characterization of intracranial hemorrhage and mineralization, but has the drawback of long acquisition times. We aimed to propose a deep learning model to accelerate SWI, and evaluate the clinical feasibility of this approach. METHODS: A complex-valued convolutional neural network (ComplexNet) was developed to reconstruct high-quality SWI from highly accelerated k-space data. ComplexNet can leverage the inherently complex-valued nature of SWI data and learn richer representations by using complex-valued network. SWI data were acquired from 117 participants who underwent clinical brain MRI examination between 2019 and 2021, including patients with tumor, stroke, hemorrhage, traumatic brain injury, etc. Reconstruction quality was evaluated using quantitative image metrics and image quality scores, including overall image quality, signal-to-noise ratio, sharpness, and artifacts. RESULTS: The average reconstruction time of ComplexNet was 19 ms per section (1.33 s per participant). ComplexNet achieved significantly improved quantitative image metrics compared to a conventional compressed sensing method and a real-valued network with acceleration rates of 5 and 8 (p < 0.001). Meanwhile, there was no significant difference between fully sampled and ComplexNet approaches in terms of overall image quality and artifacts (p > 0.05) at both acceleration rates. Furthermore, ComplexNet showed comparable diagnostic performance to the fully sampled SWI for visualizing a wide range of pathology, including hemorrhage, cerebral microbleeds, and brain tumor. CONCLUSIONS: ComplexNet can effectively accelerate SWI while providing superior performance in terms of overall image quality and visualization of pathology for routine clinical brain imaging. KEY POINTS: • The complex-valued convolutional neural network (ComplexNet) allowed fast and high-quality reconstruction of highly accelerated SWI data, with an average reconstruction time of 19 ms per section. • ComplexNet achieved significantly improved quantitative image metrics compared to a conventional compressed sensing method and a real-valued network with acceleration rates of 5 and 8 (p < 0.001). • ComplexNet showed comparable diagnostic performance to the fully sampled SWI for visualizing a wide range of pathology, including hemorrhage, cerebral microbleeds, and brain tumor.