Optimizing Acute Stroke Segmentation on MRI Using Deep Learning: Self-Configuring Neural Networks Provide High Performance Using Only DWI Sequences.

Segmentation of infarcts is clinically important in ischemic stroke management and prognostication. It is unclear what role the combination of DWI, ADC, and FLAIR MRI sequences provide for deep learning in infarct segmentation. Recent technologies in model self-configuration have promised greater performance and generalizability through automated optimization. We assessed the utility of DWI, ADC, and FLAIR sequences on ischemic stroke segmentation, compared self-configuring nnU-Net models to conventional U-Net models without manual optimization, and evaluated the generalizability of results on an external clinical dataset. 3D self-configuring nnU-Net models and standard 3D U-Net models with MONAI were trained on 200 infarcts using DWI, ADC, and FLAIR sequences separately and in all combinations. Segmentation results were compared between models using paired t-test comparison on a hold-out test set of 50 cases. The highest performing model was externally validated on a clinical dataset of 50 MRIs. nnU-Net with DWI sequences attained a Dice score of 0.810 ± 0.155. There was no statistically significant difference when DWI sequences were supplemented with ADC and FLAIR images (Dice score of 0.813 ± 0.150; p = 0.15). nnU-Net models significantly outperformed standard U-Net models for all sequence combinations (p < 0.001). On the external dataset, Dice scores measured 0.704 ± 0.199 for positive cases with false positives with intracranial hemorrhage. Highly optimized neural networks such as nnU-Net provide excellent stroke segmentation even when only provided DWI images, without significant improvement from other sequences. This differs from-and significantly outperforms-standard U-Net architectures. Results translated well to the external clinical environment and provide the groundwork for optimized acute stroke segmentation on MRI.

Identification of metabolites associated with preserved muscle volume after aneurysmal subarachnoid hemorrhage due to high protein supplementation and neuromuscular electrical stimulation.

The INSPIRE randomized clinical trial demonstrated that a high protein diet (HPRO) combined with neuromuscular electrical stimulation (NMES) attenuates muscle atrophy and may improve outcomes after aneurysmal subarachnoid hemorrhage We sought to identify specific metabolites mediating these effects. Blood samples were collected from subjects on admission prior to randomization to either standard of care (SOC; N = 12) or HPRO + NMES (N = 12) and at 7 days. Untargeted metabolomics were performed for each plasma sample. Sparse partial least squared discriminant analysis identified metabolites differentiating each group. Correlation coefficients were calculated between each metabolite and total protein per day and muscle volume. Multivariable models determined associations between metabolites and muscle volume. Unique metabolites (18) were identified differentiating SOC from HPRO + NMES. Of these, 9 had significant positive correlations with protein intake. In multivariable models, N-acetylleucine was significantly associated with preserved temporalis [OR 1.08 (95% CI 1.01, 1.16)] and quadricep [OR 1.08 (95% CI 1.02, 1.15)] muscle volume. Quinolinate was also significantly associated with preserved temporalis [OR 1.05 (95% CI 1.01, 1.09)] and quadricep [OR 1.04 (95% CI 1.00, 1.07)] muscle volume. N-acetylserine and ß-hydroxyisovaleroylcarnitine were associated with preserved temporalis or quadricep volume. Metabolites defining HPRO + NMES had strong correlations with protein intake and were associated with preserved muscle volume.

Advances in chronic subdural hematoma and membrane imaging.

Chronic subdural hematoma (cSDH) is projected to become the most common cranial neurosurgical disease by 2030. Despite medical and surgical management, recurrence rates remain high. Recently, middle meningeal artery embolization (MMAE) has emerged as a promising treatment; however, determinants of disease recurrence are not well understood, and developing novel radiographic biomarkers to assess hematomas and cSDH membranes remains an active area of research. In this narrative review, we summarize the current state-of-the-art for subdural hematoma and membrane imaging and discuss the potential role of MR and dual-energy CT imaging in predicting cSDH recurrence, surgical planning, and selecting patients for embolization treatment.

Acute Development of Traumatic Intracranial Aneurysms After Civilian Gunshot Wounds to the Head.

In previous studies, the incidence of traumatic intracranial aneurysms (TICAs) after civilian gunshot wound to the head (cGSWH) was ∼3%. Given the use of delayed vessel imaging, we hypothesize that a significant fraction of TICAs is missed on initial non-contrasted scans. This study was designed to characterize acute TICAs using admission computed tomographic angiography (aCTA) in cGSWH. Over the period from 2017 to 2022, 341 patients were admitted to R. Adams Cowley Shock Trauma Center with cGSWH; 136 subjects had aCTA ∼3 (standard deviation [SD] 3.5) h post-injury. Demographics, clinical findings, imaging techniques, endovascular/surgical interventions, and outcomes were analyzed. Mean age was 34.7 (SD 13.1), male:female ratio was 120:16. Average admission Glasgow Coma Scale (GCS) score was 6 (SD 3.9). Entry site was frontal in 41, temporal in 55, parietal in 18, occipital in 6, suboccipital in 9, temporo-parietal in 1, and frontobasal-temporal in 6. Projectiles crossed multiple dural compartments in 76 (55%) patients. 35 TICAs were diagnosed in 28 subject: 24 were located along the middle cerebral artery (MCA), 6 in the anterior cerebral artery (ACA), 3 in the internal carotid artery (ICA), 1 in the posterior cerebral artery (PCA), and 1 in the middle meningeal artery (MMA). Eleven TICAs resolved spontaneously in nine patients. Eight aneurysms were treated by endovascular means, two via combined endovascular/open approaches. Forty-nine patients died, 10 of whom had 15 TICAs. Eighty patients developed intracerebral hematoma s (ICHs). Regression models showed that the presence of an ICH was the main predictor of TICA in cGSWH. Larger ICHs (average 22.3 cc vs. 9.4 cc in patients with and without aneurysms, respectively) in patients with cGSWH suggest hidden TICAs. Nearly 30% of patients had spontaneous resolution within 1 week. When CTA was performed acutely, TICAs were 10 times more frequent in cGSWH than in previous literature, and those patients were more likely to proceed to surgery. Almost one third of patients in this series died from the devastating effects of cGSWH.

Iodine-based dual-energy CT predicts early neurological decline from cerebral edema after large hemispheric infarction.

Background & Purpose: Ischemia affecting two thirds of the MCA territory predicts development of malignant cerebral edema. However, early infarcts are hard to diagnose on conventional head CT. We hypothesize that high-energy (190keV) virtual monochromatic images (VMI) from dual-energy CT (DECT) imaging enables earlier detection of secondary injury from malignant cerebral edema (MCE). Methods: Consecutive LHI patients with NIHSS ≥ 15 and DECT within 10 hours of reperfusion from May 2020 to March 2022 were included. We excluded patients with parenchymal hematoma-type 2 transformation. Retrospective analysis of clinical and novel variables included VMI Alberta Stroke Program Early CT Score (ASPECTS), total iodine content, and VMI infarct volume. Primary outcome was early neurological decline (END). Secondary outcomes included hemorrhagic transformation, decompressive craniectomy (DC), and medical treatment of MCE. Fisher's exact test and Wilcoxon test were used for univariate analysis. Logistic regression was used to develop prediction models for categorical outcomes. Results: Eighty-four LHI patients with a median age of 67.5 [IQR 57,78] years and NIHSS 22 [IQR 18,25] were included. Twenty-nine patients had END. VMI ASPECTS, total iodine content, and VMI infarct volume were associated with END. VMI ASPECTS, VMI infarct volume, and total iodine content were predictors of END after adjusting for age, sex, initial NIHSS, and tPA administration, with a AUROC of 0.691 [0.572,0.810], 0.877 [0.800, 0.954], and 0.845 [0.750, 0.940]. By including all three predictors, the model achieved AUROC of 0.903 [0.84,0.97] and was cross validated by leave one out method with AUROC of 0.827. Conclusion: DECT with high-energy VMI and iodine quantification is superior to conventional CT ASPECTS and is a novel predictor for early neurological decline due to malignant cerebral edema after large hemispheric infarction.

High-Protein Supplementation and Neuromuscular Electric Stimulation after Aneurysmal Subarachnoid Hemorrhage Increases Systemic Amino Acid and Oxidative Metabolism: A Plasma Metabolomics Approach.

Background: The INSPIRE randomized clinical trial demonstrated that a high protein diet (HPRO) combined with neuromuscular electrical stimulation (NMES) attenuates muscle atrophy and may improve functional outcomes after aSAH. Using an untargeted metabolomics approach, we sought to identify specific metabolites mediating these effects. Methods: Blood samples were collected from subjects on admission prior to randomization to either standard of care (SOC; N=12) or HPRO+NMES (N=12) and at 7 days as part of the INSPIRE protocol. Untargeted metabolomics were performed for each plasma sample. Paired fold changes were calculated for each metabolite among subjects in the HPRO+NMES group at baseline and 7 days after intervention. Changes in metabolites from baseline to 7 days were compared for the HPRO+NMES and SOC groups. Sparse partial least squared discriminant analysis (sPLS-DA) identified metabolites discriminating each group. Pearson's correlation coefficients were calculated between each metabolite and total protein per day, nitrogen balance, and muscle volume Multivariable models were developed to determine associations between each metabolite and muscle volume. Results: A total of 18 unique metabolites were identified including pre and post treatment and differentiating SOC vs HPRO+NMES. Of these, 9 had significant positive correlations with protein intake: N-acetylserine (ρ=0.61, P=1.56×10-3), N-acetylleucine (ρ=0.58, P=2.97×10-3), ß-hydroxyisovaleroylcarnitine (ρ=0.53, P=8.35×10-3), tiglyl carnitine (ρ=0.48, P=0.0168), N-acetylisoleucine (ρ=0.48, P=0.0183), N-acetylthreonine (ρ=0.47, P=0.0218), N-acetylkynurenine (ρ=0.45, P=0.0263), N-acetylvaline (ρ=0.44, P=0.0306), and urea (ρ=0.43, P=0.0381). In multivariable regression models, N-acetylleucine was significantly associated with preserved temporalis [OR 1.08 (95%CI 1.01, 1.16)] and quadricep [OR 1.08 (95%CI 1.02, 1.15)] muscle volume. Quinolinate was also significantly associated with preserved temporalis [OR 1.05 (95%CI 1.01, 1.09)] and quadricep [OR 1.04 (95%CI 1.00, 1.07)] muscle volume. N-acetylserine, N-acetylcitrulline, and b-hydroxyisovaleroylcarnitine were also associated with preserved temporalis or quadricep volume. Conclusions: Metabolites defining the HPRO+NMES intervention mainly consisted of amino acid derivatives. These metabolites had strong correlations with protein intake and were associated with preserved muscle volume.

Quantitative DECT of Iodine in Chronic Subdural Hematoma as Surrogate of Membrane Exudation: A Pilot Feasibility Study.

OBJECTIVE: We explore the feasibility to estimate the exudation from chronic subdural hematoma (CSDH) membranes, by using dual-energy computed tomography (DECT) quantification of iodine leak and test if the derived quantitative variables and membrane morphology correlates with hematoma volume, internal architecture (homogeneous, laminar, separated, and trabecular types), and fractional hyperdense hematoma at presentation. METHODS: In this retrospective study, consecutive CSDH patients with postcontrast DECT head images from January 2020 and June 2021 were analyzed. Predictor variables derived from DECT were correlated with outcome variables followed by mixed-effects regression analysis. RESULTS: The study included 36 patients with 50 observations (mean age, 72.6 years; standard deviation, 11.6 years); 31 were men. Dual-energy CT variables that correlated with hematoma volume were external membrane volume (ρ, 0.37; P = 0.008) and iodine concentration (ρ, -0.29; P = 0.04). Variables that correlated with separated type of hematoma were total iodine leak (median [Q 1 , Q 3 ], 68.3 mg [48.5, 88.9] vs 38.8 mg [15.5, 62.9]; P = 0.001) and iodine leak per unit membrane volume (median [Q 1 , Q 3 ], 16.47 mg/mL [10.19, 20.65] vs 8.68 mg/mL [5.72, 11.41]; P = 0.002). Membrane grade was the only variable that correlated with fractional hyperdense hematoma (ρ, 0.28; P = 0.05). Regression analysis showed total iodine leak as the strongest predictor of separated type hematoma (odds ratio [95% confidence interval], 1.06 per mg [1.01, 1.1]). CONCLUSIONS: Dual-energy CT demonstrates iodine leak from CSDH membranes. The variables derived from DECT correlated with hematoma volume, internal architecture, and fractional hyperdense hematoma.

Accelerating voxelwise annotation of cross-sectional imaging through AI collaborative labeling with quality assurance and bias mitigation.

Background: precision-medicine quantitative tools for cross-sectional imaging require painstaking labeling of targets that vary considerably in volume, prohibiting scaling of data annotation efforts and supervised training to large datasets for robust and generalizable clinical performance. A straight-forward time-saving strategy involves manual editing of AI-generated labels, which we call AI-collaborative labeling (AICL). Factors affecting the efficacy and utility of such an approach are unknown. Reduction in time effort is not well documented. Further, edited AI labels may be prone to automation bias. Purpose: In this pilot, using a cohort of CTs with intracavitary hemorrhage, we evaluate both time savings and AICL label quality and propose criteria that must be met for using AICL annotations as a high-throughput, high-quality ground truth. Methods: 57 CT scans of patients with traumatic intracavitary hemorrhage were included. No participant recruited for this study had previously interpreted the scans. nnU-net models trained on small existing datasets for each feature (hemothorax/hemoperitoneum/pelvic hematoma; n = 77-253) were used in inference. Two common scenarios served as baseline comparison- de novo expert manual labeling, and expert edits of trained staff labels. Parameters included time effort and image quality graded by a blinded independent expert using a 9-point scale. The observer also attempted to discriminate AICL and expert labels in a random subset (n = 18). Data were compared with ANOVA and post-hoc paired signed rank tests with Bonferroni correction. Results: AICL reduced time effort 2.8-fold compared to staff label editing, and 8.7-fold compared to expert labeling (corrected p < 0.0006). Mean Likert grades for AICL (8.4, SD:0.6) were significantly higher than for expert labels (7.8, SD:0.9) and edited staff labels (7.7, SD:0.8) (corrected p < 0.0006). The independent observer failed to correctly discriminate AI and human labels. Conclusion: For our use case and annotators, AICL facilitates rapid large-scale curation of high-quality ground truth. The proposed quality control regime can be employed by other investigators prior to embarking on AICL for segmentation tasks in large datasets.

Surgical Decompression of Traumatic Cervical Spinal Cord Injury: A Pilot Study Comparing Real-Time Intraoperative Ultrasound After Laminectomy With Postoperative MRI and CT Myelography.

BACKGROUND: Decompression of the injured spinal cord confers neuroprotection. Compared with timing of surgery, verification of surgical decompression is understudied. OBJECTIVE: To compare the judgment of cervical spinal cord decompression using real-time intraoperative ultrasound (IOUS) following laminectomy with postoperative MRI and CT myelography. METHODS: Fifty-one patients were retrospectively reviewed. Completeness of decompression was evaluated by real-time IOUS and compared with postoperative MRI (47 cases) and CT myelography (4 cases). RESULTS: Five cases (9.8%) underwent additional laminectomy after initial IOUS evaluation to yield a final judgment of adequate decompression using IOUS in all 51 cases (100%). Postoperative MRI/CT myelography showed adequate decompression in 43 cases (84.31%). Six cases had insufficient bony decompression, of which 3 (50%) had cerebrospinal fluid effacement at >1 level. Two cases had severe circumferential intradural swelling despite adequate bony decompression. Between groups with and without adequate decompression on postoperative MRI/CT myelography, there were significant differences for American Spinal Injury Association motor score, American Spinal Injury Association Impairment Scale grade, AO Spine injury morphology, and intramedullary lesion length (IMLL). Multivariate analysis using stepwise variable selection and logistic regression showed that preoperative IMLL was the most significant predictor of inadequate decompression on postoperative imaging (P = .024). CONCLUSION: Patients with severe clinical injury and large IMLL were more likely to have inadequate decompression on postoperative MRI/CT myelography. IOUS can serve as a supplement to postoperative MRI/CT myelography for the assessment of spinal cord decompression. However, further investigation, additional surgeon experience, and anticipation of prolonged swelling after surgery are required.

Dual-energy CT imaging of chronic subdural hematoma membranes: technical note.

This technical note describes a novel dual-energy CT (DECT) protocol with iodine map reconstruction that will enable visualization of chronic subdural hematoma (CSDH) membranes. We describe the technique and discuss the potential implications for surgical management. The cohort included 36 patients with 50 hematomas. Enhancing external membrane was demonstrated in all the 50 hematomas, incomplete internal membrane in 13, and complete internal membrane in 23 hematomas. A spandrel sign at the transition zone that indicates partial or complete formation of internal membrane was demonstrated in 36 hematomas. KEY POINTS: • Iodine maps from 5-min delayed post-contrast DECT provide spectral contrast difference and facilitate segregation of chronic subdural hematoma membranes. • The ability to image the membranes helps in assessing the degree of organization of the hematoma by providing the information about the membrane thickness, volume, complexity of the membranes, and the proportion of the liquefied component within the hematoma before surgical procedures are undertaken. • Membrane visualization helps in the localization of the transition zone and extension of the membranes over the cerebral lobes helping in the determination of craniotomy location and size, during membranectomy.