Active Learning in Brain Tumor Segmentation with Uncertainty Sampling and Annotation Redundancy Restriction.

Deep learning models have demonstrated great potential in medical imaging but are limited by the expensive, large volume of annotations required. To address this, we compared different active learning strategies by training models on subsets of the most informative images using real-world clinical datasets for brain tumor segmentation and proposing a framework that minimizes the data needed while maintaining performance. Then, 638 multi-institutional brain tumor magnetic resonance imaging scans were used to train three-dimensional U-net models and compare active learning strategies. Uncertainty estimation techniques including Bayesian estimation with dropout, bootstrapping, and margins sampling were compared to random query. Strategies to avoid annotating similar images were also considered. We determined the minimum data necessary to achieve performance equivalent to the model trained on the full dataset (α = 0.05). Bayesian approximation with dropout at training and testing showed results equivalent to that of the full data model (target) with around 30% of the training data needed by random query to achieve target performance (p = 0.018). Annotation redundancy restriction techniques can reduce the training data needed by random query to achieve target performance by 20%. We investigated various active learning strategies to minimize the annotation burden for three-dimensional brain tumor segmentation. Dropout uncertainty estimation achieved target performance with the least annotated data.

Umbrella review and network meta-analysis of diagnostic imaging test accuracy studies in Differentiating between brain tumor progression versus pseudoprogression and radionecrosis.

PURPOSE: In this study we gathered and analyzed the available evidence regarding 17 different imaging modalities and performed network meta-analysis to find the most effective modality for the differentiation between brain tumor recurrence and post-treatment radiation effects. METHODS: We conducted a comprehensive systematic search on PubMed and Embase. The quality of eligible studies was assessed using the Assessment of Multiple Systematic Reviews-2 (AMSTAR-2) instrument. For each meta-analysis, we recalculated the effect size, sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio from the individual study data provided in the original meta-analysis using a random-effects model. Imaging technique comparisons were then assessed using NMA. Ranking was assessed using the multidimensional scaling approach and by visually assessing surface under the cumulative ranking curves. RESULTS: We identified 32 eligible studies. High confidence in the results was found in only one of them, with a substantial heterogeneity and small study effect in 21% and 9% of included meta-analysis respectively. Comparisons between MRS Cho/NAA, Cho/Cr, DWI, and DSC were most studied. Our analysis showed MRS (Cho/NAA) and 18F-DOPA PET displayed the highest sensitivity and negative likelihood ratios. 18-FET PET was ranked highest among the 17 studied techniques with statistical significance. APT MRI was the only non-nuclear imaging modality to rank higher than DSC, with statistical insignificance, however. CONCLUSION: The evidence regarding which imaging modality is best for the differentiation between radiation necrosis and post-treatment radiation effects is still inconclusive. Using NMA, our analysis ranked FET PET to be the best for such a task based on the available evidence. APT MRI showed promising results as a non-nuclear alternative.

A Fully Automated Deep-Learning Model for Predicting the Molecular Subtypes of Posterior Fossa Ependymomas Using T2-Weighted Images.

PURPOSE: We aimed to develop and validate a deep learning (DL) model to automatically segment posterior fossa ependymoma (PF-EPN) and predict its molecular subtypes [Group A (PFA) and Group B (PFB)] from preoperative MR images. EXPERIMENTAL DESIGN: We retrospectively identified 227 PF-EPNs (development and internal test sets) with available preoperative T2-weighted (T2w) MR images and molecular status to develop and test a 3D nnU-Net (referred to as T2-nnU-Net) for tumor segmentation and molecular subtype prediction. The network was externally tested using an external independent set [n = 40; subset-1 (n = 31) and subset-2 (n =9)] and prospectively enrolled cases [prospective validation set (n = 27)]. The Dice similarity coefficient was used to evaluate the segmentation performance. Receiver operating characteristic analysis for molecular subtype prediction was performed. RESULTS: For tumor segmentation, the T2-nnU-Net achieved a Dice score of 0.94 ± 0.02 in the internal test set. For molecular subtype prediction, the T2-nnU-Net achieved an AUC of 0.93 and accuracy of 0.89 in the internal test set, an AUC of 0.99 and accuracy of 0.93 in the external test set. In the prospective validation set, the model achieved an AUC of 0.93 and an accuracy of 0.89. The predictive performance of T2-nnU-Net was superior or comparable to that of demographic and multiple radiologic features (AUCs ranging from 0.87 to 0.95). CONCLUSIONS: A fully automated DL model was developed and validated to accurately segment PF-EPNs and predict molecular subtypes using only T2w MR images, which could help in clinical decision-making.

Trial Readiness of Cavernous Malformations With Symptomatic Hemorrhage, Part II: Biomarkers and Trial Modeling.

BACKGROUND: Quantitative susceptibility mapping (QSM) and dynamic contrast-enhanced quantitative perfusion (DCEQP) magnetic resonance imaging sequences assessing iron deposition and vascular permeability were previously correlated with new hemorrhage in cerebral cavernous malformations. We assessed their prospective changes in a multisite trial-readiness project. METHODS: Patients with cavernous malformation and symptomatic hemorrhage (SH) in the prior year, without prior or planned lesion resection or irradiation were enrolled. Mean QSM and DCEQP of the SH lesion were acquired at baseline and at 1- and 2-year follow-ups. Sensitivity and specificity of biomarker changes were analyzed in relation to predefined criteria for recurrent SH or asymptomatic change. Sample size calculations for hypothesized therapeutic effects were conducted. RESULTS: We logged 143 QSM and 130 DCEQP paired annual assessments. Annual QSM change was greater in cases with SH than in cases without SH (P=0.019). Annual QSM increase by ≥6% occurred in 7 of 7 cases (100%) with recurrent SH and in 7 of 10 cases (70%) with asymptomatic change during the same epoch and 3.82× more frequently than clinical events. DCEQP change had lower sensitivity for SH and asymptomatic change than QSM change and greater variance. A trial with the smallest sample size would detect a 30% difference in QSM annual change during 2 years of follow-up in 34 or 42 subjects (1 and 2 tailed, respectively); power, 0.8, α=0.05. CONCLUSIONS: Assessment of QSM change is feasible and sensitive to recurrent bleeding in cavernous malformations. Evaluation of an intervention on QSM percent change may be used as a time-averaged difference between 2 arms using a repeated measures analysis. DCEQP change is associated with lesser sensitivity and higher variability than QSM. These results are the basis of an application for certification by the US Food and Drug Administration of QSM as a biomarker of drug effect on bleeding in cavernous malformations. REGISTRATION: URL: https://www.clinicaltrials.gov; Unique identifier: NCT03652181.

Trial Readiness of Cavernous Malformations With Symptomatic Hemorrhage, Part I: Event Rates and Clinical Outcome.

BACKGROUND: Cerebral cavernous malformation with symptomatic hemorrhage (SH) are targets for novel therapies. A multisite trial-readiness project (https://www.clinicaltrials.gov; Unique identifier: NCT03652181) aimed to identify clinical, imaging, and functional changes in these patients. METHODS: We enrolled adult cerebral cavernous malformation patients from 5 high-volume centers with SH within the prior year and no planned surgery. In addition to clinical and imaging review, we assessed baseline, 1- and 2-year National Institutes of Health Stroke Scale, modified Rankin Scale, European Quality of Life 5D-3 L, and patient-reported outcome-measurement information system, Version 2.0. SH and asymptomatic change rates were adjudicated. Changes in functional scores were assessed as a marker for hemorrhage. RESULTS: One hundred twenty-three, 102, and 69 patients completed baseline, 1- and 2-year clinical assessments, respectively. There were 21 SH during 178.3 patient years of follow-up (11.8% per patient year). At baseline, 62.6% and 95.1% of patients had a modified Rankin Scale score of 1 and National Institutes of Health Stroke Scale score of 0 to 4, respectively, which improved to 75.4% (P=0.03) and 100% (P=0.06) at 2 years. At baseline, 74.8% had at least one abnormal patient-reported outcome-measurement information system, Version 2.0 domain compared with 61.2% at 2 years (P=0.004). The most common abnormal European Quality of Life 5D-3 L domains were pain (48.7%), anxiety (41.5%), and participation in usual activities (41.4%). Patients with prospective SH were more likely than those without SH to display functional decline in sleep, fatigue, and social function patient-reported outcome-measurement information system, Version 2.0 domains at 2 years. Other score changes did not differ significantly between groups at 2 years. The sensitivity of scores as an SH marker remained poor at the time interval assessed. CONCLUSIONS: We report SH rate, functional, and patient-reported outcomes in trial-eligible cerebral cavernous malformation with SH patients. Functional outcomes and patient-reported outcomes generally improved over 2 years. No score change was highly sensitive or specific for SH and could not be used as a primary end point in a trial.

CEST2022: Amide proton transfer-weighted MRI improves the diagnostic performance of multiparametric non-contrast-enhanced MRI techniques in patients with post-treatment high-grade gliomas.

New or enlarged lesions in malignant gliomas after surgery and chemoradiation can be associated with tumor recurrence or treatment effect. Due to similar radiographic characteristics, conventional-and even some advanced MRI techniques-are limited in distinguishing these two pathologies. Amide proton transfer-weighted (APTw) MRI, a protein-based molecular imaging technique that does not require the administration of any exogenous contrast agent, was recently introduced into the clinical setting. In this study, we evaluated and compared the diagnostic performances of APTw MRI with several non-contrast-enhanced MRI sequences, such as diffusion-weighted imaging, susceptibility-weighted imaging, and pseudo-continuous arterial spin labeling. Thirty-nine scans from 28 glioma patients were obtained on a 3 T MRI scanner. A histogram analysis approach was employed to extract parameters from each tumor area. Statistically significant parameters (P < 0.05) were selected to train multivariate logistic regression models to evaluate the performance of MRI sequences. Multiple histogram parameters, particularly from APTw and pseudo-continuous arterial spin labeling images, demonstrated significant differences between treatment effect and recurrent tumor. The regression model trained on the combination of all significant histogram parameters achieved the best result (area under the curve = 0.89). We found that APTw images added value to other advanced MR images for the differentiation of treatment effect and tumor recurrence.

Cavernous Angioma Symptomatic Hemorrhage (CASH) Trial Readiness II: Imaging Biomarkers and Trial Modeling.

Background: Quantitative susceptibility mapping (QSM) and dynamic contrast enhanced quantitative perfusion (DCEQP) MRI sequences assessing iron deposition and vascular permeability were previously correlated with new hemorrhage in cavernous angiomas. We assessed their prospective changes in cavernous angiomas with symptomatic hemorrhage (CASH) in a multisite trial readiness project ( clinicaltrials.gov NCT03652181 ). Methods: Patients with CASH in the prior year, without prior or planned lesion resection or irradiation were enrolled. Mean QSM and DCEQP of CASH lesion were acquired at baseline, and at 1- and 2-year follow-ups. Sensitivity and specificity of biomarker changes were analyzed in relation to predefined lesional symptomatic hemorrhage (SH) or asymptomatic change (AC). Sample size calculations for hypothesized therapeutic effects were conducted. Results: We logged 143 QSM and 130 DCEQP paired annual assessments. Annual QSM change was greater in cases with SH than in cases without SH (p= 0.019). Annual QSM increase by ≥ 6% occurred in 7 of 7 cases (100%) with recurrent SH and in 7 of 10 cases (70%) with AC during the same epoch, and 3.82 times more frequently than clinical events. DCEQP change had lower sensitivity for SH and AC than QSM change, and greater variance. A trial with smallest sample size would detect a 30% difference in QSM annual change in 34 or 42 subjects (one and two-tailed, respectively), power 0.8, alpha 0.05. Conclusions: Assessment of QSM change is feasible and sensitive to recurrent bleeding in CASH. Evaluation of an intervention on QSM percent change may be used as a time-averaged difference between 2 arms using a repeated measures analysis. DCEQP change is associated with lesser sensitivity and higher variability than QSM. These results are the basis of an application for certification by the U.S. F.D.A. of QSM as a biomarker of drug effect in CASH.

Brain Tumor Segmentation for Multi-Modal MRI with Missing Information.

Deep convolutional neural networks (DCNNs) have shown promise in brain tumor segmentation from multi-modal MRI sequences, accommodating heterogeneity in tumor shape and appearance. The fusion of multiple MRI sequences allows networks to explore complementary tumor information for segmentation. However, developing a network that maintains clinical relevance in situations where certain MRI sequence(s) might be unavailable or unusual poses a significant challenge. While one solution is to train multiple models with different MRI sequence combinations, it is impractical to train every model from all possible sequence combinations. In this paper, we propose a DCNN-based brain tumor segmentation framework incorporating a novel sequence dropout technique in which networks are trained to be robust to missing MRI sequences while employing all other available sequences. Experiments were performed on the RSNA-ASNR-MICCAI BraTS 2021 Challenge dataset. When all MRI sequences were available, there were no significant differences in performance of the model with and without dropout for enhanced tumor (ET), tumor (TC), and whole tumor (WT) (p-values 1.000, 1.000, 0.799, respectively), demonstrating that the addition of dropout improves robustness without hindering overall performance. When key sequences were unavailable, the network with sequence dropout performed significantly better. For example, when tested on only T1, T2, and FLAIR sequences together, DSC for ET, TC, and WT increased from 0.143 to 0.486, 0.431 to 0.680, and 0.854 to 0.901, respectively. Sequence dropout represents a relatively simple yet effective approach for brain tumor segmentation with missing MRI sequences.

The Ribbon Sign as a Radiological Indicator of Intramedullary Spinal Cord Subependymomas.

OBJECTIVE: Intramedullary spinal cord (IMSC) subependymomas are rare World Health Organization grade 1 ependymal tumors. The potential presence of functional neural tissue within the tumor and poorly demarcated planes presents a risk to resection. Anticipating a subependymoma on preoperative imaging can inform surgical decision-making and improve patient counseling. Here, we present our experience recognizing IMSC subependymomas on preoperative magnetic resonance imaging (MRI) based on a distinctive characteristic termed the "ribbon sign." METHODS: We retrospectively reviewed preoperative MRIs of patients presenting with IMSC tumors at a large tertiary academic institution between April 2005 and January 2022. The diagnosis was confirmed histologically. The "ribbon sign" was defined as a ribbon-like structure of T2 isointense spinal cord tissue interwoven between regions of T2 hyperintense tumor. The ribbon sign was confirmed by an expert neuroradiologist. RESULTS: MRIs from 151 patients were reviewed, including 10 patients with IMSC subependymomas. The ribbon sign was demonstrated on 9 (90%) patients with histologically proven subependymomas. Other tumor types did not display the ribbon sign. CONCLUSION: The ribbon sign is a potentially distinctive imaging feature of IMSC subependymomas and indicates the presence of spinal cord tissue between eccentrically located tumors. Recognition of the ribbon sign should prompt clinicians to consider a diagnosis of subependymoma, aiding the neurosurgeon in planning the surgical approach and adjusting the surgical outcome expectation. Consequently, the risks and benefits of gross-versus subtotal resection for palliative debulking should be carefully considered and discussed with patients.

Deep Learning for Noninvasive Assessment of H3 K27M Mutation Status in Diffuse Midline Gliomas Using MR Imaging.

BACKGROUND: Determination of H3 K27M mutation in diffuse midline glioma (DMG) is key for prognostic assessment and stratifying patient subgroups for clinical trials. MRI can noninvasively depict morphological and metabolic characteristics of H3 K27M mutant DMG. PURPOSE: This study aimed to develop a deep learning (DL) approach to noninvasively predict H3 K27M mutation in DMG using T2-weighted images. STUDY TYPE: Retrospective and prospective. POPULATION: For diffuse midline brain gliomas, 341 patients from Center-1 (27 ± 19 years, 184 males), 42 patients from Center-2 (33 ± 19 years, 27 males) and 35 patients (37 ± 18 years, 24 males). For diffuse spinal cord gliomas, 133 patients from Center-1 (30 ± 15 years, 80 males). FIELD STRENGTH/SEQUENCE: 5T and 3T, T2-weighted turbo spin echo imaging. ASSESSMENT: Conventional radiological features were independently reviewed by two neuroradiologists. H3 K27M status was determined by histopathological examination. The Dice coefficient was used to evaluate segmentation performance. Classification performance was evaluated using accuracy, sensitivity, specificity, and area under the curve. STATISTICAL TESTS: Pearson's Chi-squared test, Fisher's exact test, two-sample Student's t-test and Mann-Whitney U test. A two-sided P value <0.05 was considered statistically significant. RESULTS: In the testing cohort, Dice coefficients of tumor segmentation using DL were 0.87 for diffuse midline brain and 0.81 for spinal cord gliomas. In the internal prospective testing dataset, the predictive accuracies, sensitivities, and specificities of H3 K27M mutation status were 92.1%, 98.2%, 82.9% in diffuse midline brain gliomas and 85.4%, 88.9%, 82.6% in spinal cord gliomas. Furthermore, this study showed that the performance generalizes to external institutions, with predictive accuracies of 85.7%-90.5%, sensitivities of 90.9%-96.0%, and specificities of 82.4%-83.3%. DATA CONCLUSION: In this study, an automatic DL framework was developed and validated for accurately predicting H3 K27M mutation using T2-weighted images, which could contribute to the noninvasive determination of H3 K27M status for clinical decision-making. EVIDENCE LEVEL: 2 Technical Efficacy: Stage 2.

Ablation of apparent diffusion coefficient hyperintensity clusters in mesial temporal lobe epilepsy improves seizure outcomes after laser interstitial thermal therapy.

OBJECTIVE: Laser interstitial thermal therapy (LiTT) is a minimally invasive surgical procedure for intractable mesial temporal epilepsy (mTLE). LiTT is safe and effective, but seizure outcomes are highly variable due to patient variability, suboptimal targeting, and incomplete ablation of the epileptogenic zone. Apparent diffusion coefficient (ADC) is a magnetic resonance imaging (MRI) sequence that can identify potential epileptogenic foci in the mesial temporal lobe to improve ablation and seizure outcomes. The objective of this study was to investigate whether ablation of tissue clusters with high ADC values in the mesial temporal structures is associated with seizure outcome in mTLE after LiTT. METHODS: Twenty-seven patients with mTLE who underwent LiTT at our institution were analyzed. One-year seizure outcome was categorized as complete seizure freedom (International League Against Epilepsy [ILAE] Class I) and residual seizures (ILAE Class II-VI). Volumes of hippocampus and amygdala were segmented from the preoperative T1 MRI sequence. Spatially distinct hyperintensity clusters were identified in the preoperative ADC map. Proportion of cluster volume and number ablated were associated with seizure outcomes. RESULTS: The mean age at surgery was 37.5 years and the mean follow-up duration was 1.9 years. Proportions of hippocampal cluster volume (p = .013) and number (p = .03) ablated were significantly higher in patients with seizure freedom. For amygdala clusters, the proportion of cluster number ablated was significantly associated with seizure outcome (p = .026). In the combined amygdalohippocampal complex, ablation of amygdalohippocampal clusters reliably predicted seizure outcome by their volume ablated (area under the curve [AUC] = 0.7670, p = .02). SIGNIFICANCE: Seizure outcome after LiTT in patients with mTLE was associated significantly with the extent of cluster ablation in the amygdalohippocampal complex. The results suggest that preoperative ADC analysis may help identify high-yield pathological tissue clusters that represent epileptogenic foci. ADC-based cluster analysis can potentially assist ablation targeting and improve seizure outcome after LiTT in mTLE.

Clinical application of brain perfusion imaging in detecting stroke mimics: A review.

Stroke mimics constitute a significant proportion of patients with suspected acute ischemic stroke. These conditions may resemble acute ischemic stroke and demonstrate abnormalities on perfusion imaging sequences. The most common stroke mimics include seizure/epilepsy, migraine with aura, brain tumors, functional disorders, infectious encephalopathies, Wernicke's encephalopathy, and metabolic abnormalities. Brain perfusion imaging techniques, particularly computed tomography perfusion and magnetic resonance perfusion, are being widely used in routine clinical practice for treatment selection in patients presenting with large vessel occlusion. At the same time, the utilization of these imaging modalities enables the opportunity to better diagnose patients with stroke mimics in a time-sensitive setting, leading to appropriate management, decision-making, and resource allocation. In this review, we describe patterns of perfusion abnormalities that could discriminate patients with stroke mimics from those with acute ischemic stroke and provide specific case examples to illustrate these perfusion abnormalities. In addition, we discuss the challenges associated with interpretation of perfusion images in stroke-related pathologies. In general, perfusion imaging can provide additional information in some cases-when used in combination with conventional magnetic resonance imaging and computed tomography-and might help in detecting stroke mimics among patients who present with acute onset focal neurological symptoms.

The Role of Preoperative Magnetic Resonance Imaging in Assessing Neurovascular Compression Before Microvascular Decompression in Trigeminal Neuralgia.

BACKGROUND: Preoperative magnetic resonance imaging (MRI) is a standard component of the preoperative clinical workup for patients before microvascular decompression (MVD). However, its ability to accurately exclude neurovascular compression of the trigeminal nerve is not well understood. METHODS: We retrospectively reviewed 1020 patients with available preoperative MRI data before microvascular decompression. General patient demographics and clinical characteristics were collected for each case. We recorded both evidence of neurovascular conflict on preoperative MRI radiology notes and intraoperative compression from operative notes. Sensitivity, specificity, positive predictive value, and negative predictive value were determined for general MRI, high-resolution MRI, and non-high resolution. RESULTS: Overall, preoperative MRI before MVD showed a sensitivity of 75.8%, specificity of 65.8%, positive predictive value of 92.4%, and negative predictive value of 33.3% in predicting neurovascular compression of the trigeminal nerve. In particular, MRI was unable to identify 21.0% cases of sole arterial compression, 42.5% cases of sole venous compression, and combined arterial and venous compression in 18.5% of cases. A total of 958 patients (93.9%) underwent high-resolution preoperative MRI with skull base sequences. This imaging showed a sensitivity of 75.6%, specificity of 66.9%, positive predictive value of 92.5% and a negative predictive value of 33.4% in predicting trigeminal nerve neurovascular compression. Non-high-resolution MRI showed a sensitivity of 78.8%, specificity of 50.0%, positive predictive value of 89.1%, and negative predictive value of 31.3%. The negative predictive values of general, high-resolution, and non-high-resolution MRIs were all <50%. CONCLUSIONS: Preoperative MRI may offer a high predictive value for neurovascular conflict and should be part of the standard preoperative care workup for patients with trigeminal neuralgia. However, lack of neurovascular conflict on preoperative imaging is not sufficient to exclude patients from undergoing MVD.

Arterial Spin Labeling Imaging Characteristics of Anti-leucine-rich Glioma-Inactivated 1 Encephalitis: A Qualitative and Quantitative Analysis.

Background and Significance: Autoimmune encephalitis (AE) is a rare group of diseases that can present with stroke-like symptoms. Anti-leucine-rich glioma inactivated 1 (LGI1) encephalitis is an AE subtype that is infrequently associated with neoplasms and highly responsive to prompt immunotherapy treatment. Therefore, accurate diagnosis of LGI1 AE is essential in timely patient management. Neuroimaging plays a critical role in evaluating stroke and stroke mimics such as AE. Arterial Spin Labeling (ASL) is an MRI perfusion modality that measures cerebral blood flow (CBF) and is increasingly used in everyday clinical practice for stroke and stroke mimic assessment as a non-contrast sequence. Our goal in this preliminary study is to demonstrate the added value of ASL in detecting LGI1 AE for prompt diagnosis and treatment. Methods: In this retrospective single center study, we identified six patients with seropositive LGI1 AE who underwent baseline MRI with single delay 3D pseudocontinuous ASL (pCASL), including five males and one female between ages 28 and 76 years, with mean age of 55 years. Two neuroradiologists qualitatively interpreted the ASL images by visual inspection of CBF using a two-point scale (increased, decreased) when compared to both the ipsilateral and contralateral unaffected temporal and non-temporal cortex. The primary measures on baseline ASL evaluation were a) presence of ASL signal abnormality, b) if present, signal characterization based on the two-point scale, c) territorial vascular distribution, d) localization, and e) laterality. Quantitative assessment was also performed on postprocessed pCASL cerebral blood flow (CBF) maps. The obtained CBF values were then compared between the affected temporal cortex and each of the unaffected ipsilateral parietal, contralateral temporal, and contralateral parietal cortices. Results: On consensus qualitative assessment, all six patients demonstrated ASL hyperperfusion and corresponding FLAIR hyperintensity in the hippocampus and/or amygdala in a non-territorial distribution (6/6, 100%). The ASL hyperperfusion was found in the right hippocampus or amygdala in 5/6 (83%) of cases. Four of the six patients underwent initial follow-up imaging where all four showed resolution of the initial ASL hyperperfusion. In the same study on structural imaging, all four patients were also diagnosed with mesial temporal sclerosis (MTS). Quantitative assessment was separately performed and demonstrated markedly increased CBF values in the affected temporal cortex (mean, 111.2 ml/min/100 g) compared to the unaffected ipsilateral parietal cortex (mean, 49 ml/min/100 g), contralateral temporal cortex (mean, 58.2 ml/min/100 g), and contralateral parietal cortex (mean, 52.2 ml/min/100 g). Discussion: In this preliminary study of six patients, we demonstrate an ASL hyperperfusion pattern, with a possible predilection for the right mesial temporal lobe on both qualitative and quantitative assessments in patients with seropositive LGI1. Larger scale studies are necessary to further characterize the strength of these associations.

Direct quantification of epistemic and aleatoric uncertainty in 3D U-net segmentation.

Purpose: To derive a multinomial probability function and quantitative measures of the data and epistemic uncertainty as direct output of a 3D U-Net segmentation network. Approach: A set of T1 brain MRI images were downloaded from the Connectome Project and segmented using FMRIB's FAST algorithm to be used as ground truth. A 3D U-Net neural network was trained with sample sizes of 200, 500, and 898 T1 brain images using a loss function defined as the negative logarithm of the likelihood based on a derivation of the definition of the multinomial probability function. From this definition, the epistemic and aleatoric uncertainty equations were derived and used to quantify maps of the uncertainty along with tissue segmentations. Results: Maps of the tissue segmentation along with the epistemic and aleatoric uncertainty, per voxel, are presented. The uncertainty decreased based on the increasing number of training data used to train the neural network. The neural network trained with 898 volumes resulted in uncertainty maps that were high primarily in the tissue boundary regions. The epistemic and aleatoric uncertainty were averaged over all test data (connectome and tumor separately), and the epistemic uncertainty showed a decreasing trend, as expected, with increasing numbers of data used to train the model. The aleatoric uncertainty showed a similar trend which was also expected as the aleatoric uncertainty is not expected to be as dependent on the number of training data. Conclusion: The derived uncertainty equations from a multinomial probability distribution were able to quantify the aleatoric and epistemic uncertainty per voxel and are applicable for all two-dimensional and three-dimensional neural networks.

Apparent diffusion coefficient of piriform cortex and seizure outcome in mesial temporal lobe epilepsy after MR-guided laser interstitial thermal therapy: a single-institution experience.

OBJECTIVE: Piriform cortex (PC) is one of the critical structures in the epileptogenesis of mesial temporal lobe epilepsy (mTLE), but its role is poorly understood. The authors examined the utility of apparent diffusion coefficient (ADC; an MR-based marker of tissue pathology) of the PC as a predictor of seizure outcome in patients with mTLE undergoing MR-guided laser interstitial thermal therapy (MRgLITT). METHODS: A total of 33 patients diagnosed with mTLE who underwent MRgLITT at the authors' institution were included in the study. The 6-month postoperative seizure outcomes were classified using the International League Against Epilepsy (ILAE) system as good (complete seizure freedom, ILAE class I) and poor (seizure present, ILAE classes II-VI). The PC and ablation volumes were manually segmented from both the preoperative and intraoperative MRI sequences, respectively. The mean ADC intensities of 1) preablation PC; 2) total ablation volume; 3) ablated portion of PC; and 4) postablation residual PC were calculated and compared between good and poor outcome groups. Additionally, the preoperative PC volumes and proportion of PC volume ablated were examined and compared between the subjects in the two outcome groups. RESULTS: The mean age at surgery was 36.5 ± 3.0 years, and the mean follow-up duration was 1.9 ± 0.2 years. Thirteen patients (39.4%) had a good outcome. The proportion of PC ablated was significantly associated with seizure outcome (10.16 vs 3.30, p < 0.05). After accounting for the variability in diffusion tensor imaging acquisition parameters, patients with good outcome had a significantly higher mean ADC of the preablation PC (0.3770 vs -0.0108, p < 0.05) and the postoperative residual PC (0.4197 vs 0.0309, p < 0.05) regions compared to those with poor outcomes. No significant differences in ADC of the ablated portion of PC were observed (0.2758 vs -0.4628, p = 0.12) after performing multivariate analysis. CONCLUSIONS: A higher proportion of PC ablated was associated with complete seizure freedom. Preoperative and postoperative residual ADC measures of PC were significantly higher in the good seizure outcome group in patients with mTLE who underwent MRgLITT, suggesting that ADC analysis can assist with postablation outcome prediction and patient stratification.

Automated eloquent cortex localization in brain tumor patients using multi-task graph neural networks.

Localizing the eloquent cortex is a crucial part of presurgical planning. While invasive mapping is the gold standard, there is increasing interest in using noninvasive fMRI to shorten and improve the process. However, many surgical patients cannot adequately perform task-based fMRI protocols. Resting-state fMRI has emerged as an alternative modality, but automated eloquent cortex localization remains an open challenge. In this paper, we develop a novel deep learning architecture to simultaneously identify language and primary motor cortex from rs-fMRI connectivity. Our approach uses the representational power of convolutional neural networks alongside the generalization power of multi-task learning to find a shared representation between the eloquent subnetworks. We validate our method on data from the publicly available Human Connectome Project and on a brain tumor dataset acquired at the Johns Hopkins Hospital. We compare our method against feature-based machine learning approaches and a fully-connected deep learning model that does not account for the shared network organization of the data. Our model achieves significantly better performance than competing baselines. We also assess the generalizability and robustness of our method. Our results clearly demonstrate the advantages of our graph convolution architecture combined with multi-task learning and highlight the promise of using rs-fMRI as a presurgical mapping tool.

Identification of the Somatomotor Network from Language Task-based fMRI Compared with Resting-State fMRI in Patients with Brain Lesions.

Background Resting-state functional MRI (rs-fMRI) is a potential alternative to task-based functional MRI (tb-fMRI) for somatomotor network (SMN) identification. Brain networks can also be generated from tb-fMRI by using independent component analysis (ICA). Purpose To investigate whether the SMN can be identified by using ICA from a language task without a motor component, the sentence completion functional MRI (sc-fMRI) task, compared with rs-fMRI. Materials and Methods The sc-fMRI and rs-fMRI scans in patients who underwent presurgical brain mapping between 2012 and 2016 were analyzed, using the same imaging parameters (other than scanning time) on a 3.0-T MRI scanner. ICA was performed on rs-fMRI and sc-fMRI scans with use of a tool to separate data sets into their spatial and temporal components. Two neuroradiologists independently determined the presence of the dorsal SMN (dSMN) and ventral SMN (vSMN) on each study. Groups were compared by using t tests, and logistic regression was performed to identify predictors of the presence of SMNs. Results One hundred patients (mean age, 40.9 years ± 14.8 [standard deviation]; 61 men) were evaluated. The dSMN and vSMN were identified in 86% (86 of 100) and 76% (76 of 100) of rs-fMRI scans and 85% (85 of 100) and 69% (69 of 100) of sc-fMRI scans, respectively. The concordance between rs-fMRI and sc-fMRI for presence of dSMN and vSMN was 75% (75 of 100 patients) and 53% (53 of 100 patients), respectively. In 10 of 14 patients (71%) where rs-fMRI did not show the dSMN, sc-fMRI demonstrated it. This rate was 67% for the vSMN (16 of 24 patients). Conclusion In the majority of patients, independent component analysis of sentence completion task functional MRI scans reliably demonstrated the somatomotor network compared with resting-state functional MRI scans. Identifying target networks with a single sentence completion scan could reduce overall functional MRI scanning times by eliminating the need for separate motor tasks. © RSNA, 2021 Online supplemental material is available for this article. See also the editorial by Field and Birn in this issue.

Apparent diffusion coefficient is associated with seizure outcome after magnetic resonance-guided laser interstitial thermal therapy for mesial temporal lobe epilepsy.

BACKGROUND: Magnetic resonance-guided laser interstitial thermal therapy (MRgLiTT) is becoming a first-line surgical therapy for mesial temporal lobe epilepsy (mTLE) due to good seizure control and low complication risk. However, seizure outcomes after MRgLiTT remain highly variable and there is a need to improve patient selection and post-operative prognostication. In this retrospective study, we investigated whether the pre-operative MRI-derived apparent diffusion coefficient (ADC), used as a marker of tissue pathology in the mesial temporal structures could help predict seizure outcome. METHODS: Thirty-five patients who underwent MRgLiTT at our institution between 2014 and 2019 were included in the study. Demographic and clinical data were retrospectively collected. Seizure outcome was defined as good (ILAE Class I-II) and poor (ILAE Class III-VI). Volumetrics were performed on pre-ablation hippocampus and amygdala. Ablation volumes, and the proportion of ablated hippocampus and amygdala calculated via their respective mean voxel-wise ADC intensities were quantified from pre-operative and intra-operative post-ablation MRIs and statistically compared between the two outcome cohorts. Univarate and multivariate regression analysis was performed to identify demographic, clinical, and radiographic predictors of seizure outcome. RESULTS: Mean age at LiTT was 36 years and 14 (40 %) were female. Mean follow-up duration was 1.90 ± 0.17 years. Twenty-seven (77 %) patients had mesial temporal sclerosis. There was no significant difference in the ablation volumes and proportion of ablated volume of hippocampus and amygdala between the two outcome groups. Patients with good seizure outcome had significantly higher normalized ADC intensities in the ablated mesial temporal structures compared to those with poor outcome (0.01 ± 0.08 vs.-0.29 ± 0.06; p = 0.015). CONCLUSIONS: mTLE patients with higher ADC intensities in the ablated regions of the hippocampus and the amygdala are more likely to have good seizure outcome following MRgLiTT. Our results suggest that pre-operative ADC analysis may improve both patient selection and epileptogenic zone targeting during MRgLiTT. Further investigation with large, prospective cohorts is needed to validate the clinical utility of ADC in improving seizure outcome following MRgLiTT.

Can Preoperative Mapping with Functional MRI Reduce Morbidity in Brain Tumor Resection? A Systematic Review and Meta-Analysis of 68 Observational Studies.

Background Preoperative functional MRI (fMRI) is one of several techniques developed to localize critical brain structures and brain tumors. However, the usefulness of fMRI for preoperative surgical planning and its potential effect on neurologic outcomes remain unclear. Purpose To assess the overall postoperative morbidity among patients with brain tumors by using preoperative fMRI versus surgery without this tool or with use of standard (nonfunctional) neuronavigation. Materials and Methods A systematic review and meta-analysis of studies across major databases from 1946 to June 20, 2020, were conducted. Inclusion criteria were original studies that (a) included patients with brain tumors, (b) performed preoperative neuroimaging workup with fMRI, (c) investigated the usefulness of a preoperative or intraoperative functional neuroimaging technique and used that technique to resect cerebral tumors, and (d) reported postoperative clinical measures. Pooled estimates for adverse event rate (ER) effect size (log ER, log odds ratio, or Hedges g) with 95% CIs were computed by using a random-effects model. Results Sixty-eight studies met eligibility criteria (3280 participants; 58.9% men [1555 of 2641]; mean age, 46 years ± 8 [standard deviation]). Functional deterioration after surgical procedure was less likely to occur when fMRI mapping was performed before the operation (odds ratio, 0.25; 95% CI: 0.12, 0.53; P < .001]), and postsurgical Karnofsky performance status scores were higher in patients who underwent fMRI mapping (Hedges g, 0.66; 95% CI: 0.21, 1.11; P = .004]). Craniotomies for tumor resection performed with preoperative fMRI were associated with a pooled adverse ER of 11% (95% CI: 8.4, 13.1), compared with a 21.0% ER (95% CI: 12.2, 33.5) in patients who did not undergo fMRI mapping. Conclusion From the currently available data, the benefit of preoperative functional MRI planning for the resection of brain tumors appears to reduce postsurgical morbidity, especially when used with other advanced imaging techniques, such as diffusion-tensor imaging, intraoperative MRI, or cortical stimulation. © RSNA, 2021 Online supplemental material is available for this article.