Search Results (1,848)

Background/Objectives: Progressive supranuclear palsy (PSP) and Parkinson’s disease (PD) share several clinical manifestations, which may complicate differential diagnosis. The aim of this study was to evaluate whether automated volumetric measurements of the pallidum and ventral diencephalon can differentiate PSP from PD. Methods: Thirty-two patients were included, comprising 20 patients with PD and 12 patients with PSP. All participants underwent 3-T brain magnetic resonance imaging. Automated segmentation and volumetric analysis were performed using the vol2Brain pipeline. Absolute and normalized volumes of the pallidum and ventral diencephalon were compared between groups. Receiver operating characteristic (ROC) analysis was used to assess diagnostic performance. Results: Patients with PSP demonstrated lower pallidal and ventral diencephalic volumes than patients with PD. Differences were observed for both absolute and normalized volumetric measurements (all p ≤ 0.002). Total pallidal volume showed the highest diagnostic performance, with an area under the ROC curve (AUC) of 0.958, sensitivity of 85%, and specificity of 100%. Total ventral diencephalon volume also differentiated PSP from PD, yielding an AUC of 0.829, seNsitivity of 70%, and specificity of 92%. Conclusions: Pallidal and ventral diencephalic volumes differed between patients with PSP and PD. Automated measurements of pallidal and ventral diencephalic volumes may complement conventional MRI findings in patients with parkinsonian syndromes. Further studies are needed to validate these findings in larger cohorts. Full article

Preterm infants are at high risk of neurodevelopmental disorders (NDDs), yet current brain assessments using 2D cranial ultrasound (cUS) remain subjective and limited. 3D cUS offers richer anatomical details but is underutilized due to its complexity and lack of automated tools. We developed and evaluated Deep Learning (DL) methods for total brain (TB) segmentation from 3D cUS of preterm neonates, emphasizing accuracy, reproducibility, and clinical deployability. We compared 2D UNet models, a contextual UNet, and self-configuring 2D/3D nnUNet variants, evaluating accuracy, efficiency, and GPU/CPU feasibility. Clinical relevance was examined through longitudinal total brain volume (TBV) trajectories and their association with 2-year neurodevelopmental outcomes. The nnUNet ensemble achieved the best performance (Dice: 0.97, Volumetric Difference Error: 2.32%), while contextual UNet offered a favorable accuracy-efficiency trade-off on low-resource hardware. Longitudinal analysis showed significantly slower TBV growth in infants with adverse outcomes (p = 0.007), supporting the prognostic value of automated volumetric measurements. We developed a clinician-facing web application to illustrate clinical integration. This work demonstrates the feasibility of DL-based TB segmentation from 3D cUS and provides a scalable reproducible framework supporting early quantitative assessment of brain development, laying the groundwork for future lightweight, privacy-aware AI solutions for Neonatal Intensive Care Unit integration. Full article

Accurate volumetric quantification of brain tumor subregions in magnetic resonance imaging (MRI) is sensitive to local prediction bias, because boundary over-expansion, missed enhancing foci, and class confusion at subregion transitions can propagate to case-level volume errors. To address this issue, this study proposes CPLR-Net, a coarse prior-guided logit-space residual refinement framework with functional symmetry between global layout preservation and local error compensation. A Coarse Prior Segmentor first generates coarse logits that encode the global semantic layout of the tumor. These logits are then concatenated with multi-modal MRI and fed into an efficient multi-scale attention (EMA)-enhanced Residual Refiner, where three-dimensional cross-spatial feature recalibration and gated logit residual updating are used to correct ambiguous boundaries and subregion transitions while maintaining consistency with the coarse semantic prior. Experiments on the Brain Tumor Segmentation (BraTS) 2020 and 2021 datasets show that CPLR-Net achieves favorable overall performance in the Dice similarity coefficient, 95th percentile Hausdorff distance, and absolute volume difference compared with representative three-dimensional segmentation methods, supporting accurate segmentation and volumetric quantification. The results suggest that coarse-prior-constrained residual refinement in logit space provides an effective strategy for reliable key-subregion segmentation and quantitative assessment in brain tumor MRI follow-up scenarios. Full article

A bioengineered version of recombinant tissue plasminogen activator, i.e., tenecteplase (TNK-tPA), was designed to have a longer half-life, resistance to plasminogen activator inhibitor-1, and fibrin-targeted plasminogen activation. By comparison to tPA, clinical trials suggest that TNK-tPA may be less susceptible to the effects of alpha2-antiplasmin (α2AP), the primary inhibitor of thrombus dissolution. However, preclinical studies are limited, and whether α2AP affects TNK-tPA’s fibrinolytic activity or efficacy in experimental ischemic stroke is unknown. We examined the effects of TNK-tPA and α2AP on the dissolution of human plasma clots (in vitro) and experimental ischemic brain injury from stroke induced by transient middle cerebral artery ischemia. TNK-tPA induced a dose-dependent increase in plasma clot dissolution; inhibition of α2AP with a specific monoclonal antibody caused a synergistic increase in TNK-tPA-mediated clot dissolution. In experimental ischemic stroke with ischemia and reperfusion, TNK-tPA treatment of α2AP^−/− mice significantly reduced ischemic infarct volume, brain swelling, brain hemorrhage, and neurobehavioral disability vs. TNK-tPA-treated α2AP^+/+ (C57BL/6 background) mice with normal α2AP levels (p < 0.05 to p < 0.0001). α2AP impairs the dissolution of human clots in vitro by TNK-tPA and significantly exacerbates ischemic brain injury, swelling, hemorrhage, and neurobehavioral disability after experimental stroke, even with reperfusion. Targeting α2AP may improve the efficacy of TNK-tPA and reduce hemorrhagic complications. Full article

Background: A growing body of literature associates increases in electronic screen time with a vast array of psychological consequences amongst adolescents, but little is known about the neurological underpinnings of this relationship. Methods: This longitudinal study examines structural and diffusion brain MRI scans from two timepoints collected from the Adolescent Brain Cognitive Development (ABCD) Study—a large, multi-site study with thousands of participants. By assessing both gray matter density (GMD) and gray matter measurements of diffusion microstructure in the adolescent brain, we describe how the developmental trajectory of the brain changes with screen-based media consumption at the subcellular level. Gray matter microstructure was measured across 13 bilateral regions functionally implicated with screen time use and associated with either the control or reward system. Results: After controlling for age, sex, total brain volume, scanning site, sibling relationships, physical activity, and socioeconomic status, this study finds significant positive correlations between increased screen time and axonal signal across six of the 13 regions selected, while also finding significantly decreased intracellular signaling in eight regions, primarily in regions functionally associated with cognitive control. Comparing these associations to normal developmental trajectories suggests adolescent age-related brain development may be accelerated by increased screen time in brain areas associated with reward processing, while age-related brain development may be decelerated in regions of the control system. Highlighting the sensitivity of microstructural analysis, no significant relationships with increased screen time were found using GMD or fractional anisotropy. Conclusions: This work suggests that increased screen usage during adolescent development has a complex association with brain tissue that cannot be completely described by traditional quantifications of tissue microstructure. Full article

Background/Objectives: Spinal muscular atrophy (SMA) types 2 and 3 are increasingly recognised as potentially multisystem disorders that may involve the central nervous system. However, the extent and pattern of brain structural alterations in adults remain insufficiently characterised. This study aimed to assess brain morphology in adult patients with SMA using Computational Anatomy Toolbox 12 (CAT12)-derived region-based morphometry and surface-based morphometry. Methods: The study included 27 right-handed adult patients with SMA types 2 and 3 and 27 age- and sex-matched healthy controls. MRI examinations were performed before initiation of disease-modifying therapy or during the early loading phase of intrathecal nusinersen treatment, with a maximum exposure of two months and no more than three doses before MRI. The groups did not differ significantly in age, sex distribution, or total intracranial volume. Regional volumetric measures and cortical surface-based parameters were extracted from predefined atlas-based regions of interest. Between-group differences and associations with selected clinical and genetic variables were analysed. p-values were corrected for multiple comparisons using the Benjamini–Hochberg false discovery rate procedure. Results: CAT12-derived region-based morphometry showed a significant reduction in grey matter volume in the left thalamus in patients with SMA compared with healthy controls. Surface-based morphometry revealed increased sulcal depth in the left orbital sulcus as the only cortical finding that remained significant after correction for multiple comparisons. Additional differences in cortical thickness and sulcal depth were observed only at nominal or exploratory thresholds and did not survive correction for multiple comparisons. Conclusions: Adults with SMA types 2 and 3 showed selected brain structural alterations, particularly reduced grey matter volume in the left thalamus and increased sulcal depth in the left orbital sulcus. These findings are consistent with the concept that SMA may extend beyond lower motor neuron degeneration. Larger longitudinal multimodal studies are warranted to validate these observations and clarify their clinical significance. Full article

Purpose: We aimed to compare the volume of the lateral geniculate nucleus (LGN) in patients with different stages of retinitis pigmentosa (RP) with regard to age, sex and symmetry of the LGN. Methods: The investigated cohort included 13 patients with moderate (median Snellen visual acuity 0.75) and 18 patients with advanced (median Snellen visual acuity 0.06) RP-related visual field loss. The volumes of the left and right LGNs were manually measured using ITK-SNAP software after an examination of the brain with a 7 Tesla MRI. A linear mixed statistical model was used to assess LGN volume regarding age and gender of moderate and advanced RP patients and symmetry of both LGNs. Results: The mixed-effects linear model did not reveal a significant effect of disease group on LGN volume after adjusting for age and sex (F(1.27) = 0.01, p = 0.91). A significant effect of the LGN side was demonstrated, with the volume of the right LGN being significantly greater than that of the left (F(1.29) = 29.45, p < 0.001) in both disease groups, left–right. The interaction between disease group and LGN side was not statistically significant (F(1.29) = 0.45, p = 0.51). There is a tendency for LGN to decrease with age (F(1.27) = 3.84, p = 0.060), and there is no gender predilection (F(1.27) = 0.11, p = 0.74) in RP patients. There was correlation found between left LGN volume and visual acuity (ρ = 0.64) and central retinal thickness (ρ = 0.71) in the moderate group). Conclusions: No significant differences in LGN volume were found between patients with moderate and advanced RP. Furthermore, the volume of the right LGN was larger than the volume of the left LGN in RP patients, and this asymmetry is not gender-dependent. Correlation was found between the left LGN volume and visual acuity and central retinal thickness in moderate group. Our findings may have clinical implications for future RP management. Full article

Pathological states associated with metabolic stress, such as traumatic brain injury (TBI), hypoxia, ischemic stroke, and migraine, are considered elevated risk factors for developing Alzheimer’s disease (AD). However, the mechanism underlying the effect of these conditions on the progression of AD remains largely unknown. Here, we determine how metabolic stress associated with spreading depolarization (SD)—a hallmark of stroke, hypoxia, TBI, and migraine—modulates amyloid β (Aβ42) aggregation kinetics through dynamic changes in extracellular space (ECS). To achieve this, we used ThT fluorescence to determine how the formation of different Aβ42 aggregate species depends on Aβ42 concentrations. Based on this input, we build a multiscale computational framework that integrates volume regulation, including its dependence on neuronal ion homeostasis, and Aβ42 aggregation kinetics. Our model predicts that neuronal swelling during SD accelerates aggregation, where the impact of metabolic stress is highly dependent on the timing relative to aggregation progression and the initial monomer concentration. At low monomer concentrations, early SD events promote off-pathway oligomer formation, while at higher concentrations they rapidly drive fibril formation to saturation. In the absence of mature fibrils, recurrent metabolic stress events further amplify oligomer accumulation, whereas pre-existing fibril nuclei suppress oligomer formation at the expense of fibril nucleation and growth. Increasing the intensity of metabolic stress prolongs ECS shrinkage and enhances oligomer formation. These findings reveal a mechanistic link between SD-induced microenvironmental changes and Aβ aggregation dynamics, providing a quantitative framework for understanding how acute brain injury and metabolic stress may contribute to early AD pathogenesis. Full article

Background: Danshen–Gegen is a classic herb pair in traditional Chinese medicine, which has been used to treat cardiovascular and cerebrovascular diseases. Ischemic stroke (IS) is a prevalent cerebrovascular condition; ferroptosis is one of the contributing factors driving the progression of IS. This study aims to determine the underlying mechanism and examine if Danshen–Gegen (DG) extract may prevent cerebral ischemia–reperfusion injury by preventing ferroptosis. Methods: The comprehensive compositional characterization of DG was analyzed by ultra-high-performance liquid chromatography coupled with hybrid quadrupole-orbitrap high-resolution mass spectrometry (UPLC-Q-orbitrap MS). The experiments were conducted in middle cerebral artery occlusion/reperfusion (MCAO/R) rats and oxygen-glucose deprivation/re-oxygenation (OGD/R) cells. The neuroprotective effects of DG on IS were assessed by examining rat survival rates, infarct volume, behavioral scores, and cerebral water content. Then, we tested the accumulation of Fe²⁺ and lipid peroxidation products such as reactive oxygen species (ROS), glutathione (GSH), malondialdehyde (MDA), myeloperoxidase (MPO), and 4-hydroxynonenal (4-HNE) in rats and cells. The expression of nuclear factor erythroid-derived 2-like 2 (Nrf2), Solute Carrier Family 7 Member 11 (-xCT), Glutathione peroxidase 4 (GPX4), Cyclooxygenase-2 (COX-2), Transferrin Receptor 1 (TFR1), and Long-chain-fatty-acid–CoA ligase 4 (ACSL4) was also assessed in vivo and in vitro. Results: UPLC-Q-orbitrap MS analysis was performed to characterize the chemical profile of DG, and a total of 33 chemical constituents were successfully identified. DG significantly alleviated the ischemic damage to brain tissue, reduced infarct volume, and improved neurological dysfunction. The content of Fe²⁺ and lipid peroxidation products was markedly decreased. Furthermore, DG could restore the expression of Nrf2, -xCT, and GPX4 with the inhibition of COX-2, TFR1, and ACSL4, thus achieving a suppressive effect on ferroptosis. Conclusions: The regulatory influence of DG via the Nrf2/System xc-/GPX4 axis may play a crucial role in alleviating ferroptosis and enhancing recovery from cerebral ischemia injury. Full article

Background: Intracranial hemorrhage (ICH) is a neurological emergency associated with high mortality and morbidity, requiring rapid diagnosis and early intervention. Artificial intelligence (AI)-based imaging systems have recently gained attention as supportive tools for accelerating emergency brain computed tomography (CT) interpretation. This study aimed to evaluate the diagnostic accuracy and time efficiency of an AI system for detecting ICH on emergency brain CT scans. Methods: This retrospective study included 375 patients who underwent non-contrast brain CT in the emergency department. AI-based image analysis results were compared with final radiologist reports accepted as the reference standard. Sensitivity, specificity, predictive values, accuracy, area under the curve (AUC), and Cohen’s kappa were calculated. Interpretation times, subgroup analyses according to presentation type and work shifts, and hemorrhage volume correlations between AI and the ABC/2 method were also evaluated. Results: Intracranial hemorrhage was detected in 235 patients (62.7%). The AI system achieved a sensitivity of 94.0%, specificity of 87.9%, positive predictive value of 92.9%, negative predictive value of 89.8%, and overall accuracy of 91.7%. The AUC was 0.909, with a Cohen’s kappa coefficient of 0.823. Mean interpretation time was significantly shorter for AI compared with radiologists (11.9 ± 3.0 vs. 70.9 ± 25.4 min, p < 0.001). AI-derived hemorrhage volume measurements strongly correlated with the ABC/2 method (rho = 0.887, p < 0.001). Conclusions: The AI system demonstrated high diagnostic accuracy and a significant reduction in interpretation time for detecting intracranial hemorrhage on emergency brain CT, supporting its potential role as a decision-support tool in emergency radiology workflows. Full article

Background/Objectives: Disability in relapsing–remitting multiple sclerosis (RRMS) reflects multiple pathological processes that may not be fully captured by individual MRI markers. Because an EDSS score ≥ 3 represents a clinically meaningful threshold of moderate-to-higher disability, this study evaluated the cross-sectional ability of brain volumetric measures and IVIM-DWI parameters to identify patients with RRMS who had EDSS ≥ 3. Methods: This retrospective cross-sectional study included 189 patients with RRMS who had complete EDSS, lesion count, IVIM-DWI, and brain volumetric MRI data. Patients were stratified into EDSS < 3 and EDSS ≥ 3 groups. Single-marker ROC analyses were performed for lesion count, IVIM-DWI parameters, absolute volumetric measures, and fractional volumetric measures. Combined ROC models were constructed to assess the discriminatory performance of integrating lesion burden, volumetric MRI, and IVIM-DWI metrics. Internal model stability was evaluated using 5-fold cross-validation. Results: Patients with EDSS ≥ 3 had higher lesion count, higher ADC and D values, lower gray matter, white matter, and brain parenchymal volumes, higher CSF volume, higher CSF fraction, and lower brain parenchymal fraction (BPF). In single-marker ROC analysis, BPF and CSF fraction showed the strongest discriminatory performance (AUC = 0.705), followed by lesion count (AUC = 0.696), whereas absolute BV showed minimal discriminatory value (AUC = 0.503). The core lesion count + BPF model achieved an AUC of 0.757 and a cross-validated AUC of 0.737. Adding perfusion fraction (f) produced the numerically highest cross-validated performance, with an AUC of 0.769 and a cross-validated AUC of 0.746, although the incremental improvement over lesion count + BPF was modest and not statistically significant. Conclusions: Fractional volumetric measures showed stronger discriminatory performance than absolute BV for EDSS ≥ 3 stratification in RRMS. The lesion count + BPF + f model achieved the highest cross-validated performance among the evaluated MRI models; however, overall discrimination was moderate, and the incremental improvement associated with f was not statistically significant. These exploratory findings require evaluation in larger longitudinal cohorts with external validation. Full article

Background: Febrile seizures (FSs) are the most common neurological emergency in early childhood; however, the biological basis of disease heterogeneity remains poorly understood. Although growing evidence suggests that gut–brain axis dysregulation contributes to seizure susceptibility, it remains unclear whether gut microbiota-associated metabolic disturbances are linked to clinical phenotypes, particularly simple FS (SFS) and complex FS (CFS). Methods: An integrated multiomics study was conducted in clinically characterized pediatric cohorts, comprising 50 children with FS and 50 healthy controls, and their gut microbiota was profiled via 16S rRNA sequencing. As some pediatric serum specimens did not meet the minimum volume requirement of the analytical platform, serum amino acid profiling was performed in a subset of samples using an equal-volume pooling strategy. In brief, two individual serum samples from the same study group were combined into one composite sample, yielding 25 pooled samples in the FS group and 25 in the control group. Subsequently, untargeted fecal metabolomics was performed in an expanded cohort of 53 healthy controls, 50 children with SFS, and 42 children with CFS. Additionally, the central metabolic profiles of the CFS and SFS groups were compared using untargeted cerebrospinal fluid metabolomics. Given the variation in sample sizes across omics platforms, each dataset was analyzed within its corresponding eligible subset, and cross-omics integration was interpreted primarily at the pathway and phenotype levels. Results: Children with FS exhibited reduced gut microbial diversity and altered microbial composition, characterized by the enrichment of Streptococcus, Enterococcus, and Escherichia–Shigella, along with the depletion of beneficial taxa, including Faecalibacterium, Lachnoclostridium, and Parasutterella. Functional prediction indicated significant changes in amino acid-related pathways, especially arginine and proline metabolism, amino acid metabolism, and glutathione metabolism. Serum profiling showed elevated levels of phenylalanine, kynurenine, and γ-aminobutyric acid, along with reduced levels of tryptophan, threonine, lysine, glutamine, taurine, citrulline, 3-methylhistidine, α-aminobutyric acid, hydroxyproline, and phosphoethanolamine. Correlation analysis identified Lachnoclostridium and Parasutterella as key taxa associated with neuroactive metabolites. Additionally, fecal metabolomics revealed that both SFS and CFS samples exhibited significant metabolic divergence from the controls, with arginine biosynthesis emerging as a shared altered pathway and L-arginine reduced in both phenotypes. Notably, cerebrospinal fluid metabolomics demonstrated clear metabolic separation between CFS and SFS, signifying phenotype-specific central metabolic signatures. Conclusions: FS is related to gut microbiota dysbiosis, systemic amino acid remodeling, and phenotype-associated metabolic stratification. Arginine metabolism may represent a shared mechanistic hub across FS phenotypes, while central metabolic divergence may contribute to the biological distinction between SFS and CFS. These findings establish a multiomics framework for understanding FS pathogenesis and identifying potential biomarkers and therapeutic targets. Full article

Objectives: To explore the brain imaging correlates of the association between chess skill level and visual spatial working memory (VSWM) in school-aged children. Methods: This study analyzed a final sample of 20 school-aged children (aged 10–15 years, mean age of 13.48 ± 1.23 years; including 9 experts and 11 novices) selected from an initial pool of 40 recruited participants from primary and secondary schools and chess clubs in Beijing. They were divided into an experimental group (expert-level school-aged children from a Beijing chess club) and a control group (novice-level children from a Beijing primary and secondary school who had not systematically participated in chess courses but understood basic chess knowledge). The N-back task was used to assess behavioral differences in VSWM. Voxel-based morphometry (VBM) and surface-based morphometry (SBM) techniques were employed to analyze differences in gray matter volume and cortical thickness in the brain structure of expert- and novice-level children. Differences in resting-state functional imaging were also analyzed. Results: The results showed the following: (1) The expert-level children demonstrated significantly higher accuracy and shorter reaction times in VSWM tasks compared to novices. (2) Differences in gray matter volume and cortical thickness were observed between the expert and novice groups. (3) Expert-level children showed significantly higher resting-state ALFF/fALFF values in the dorsolateral superior frontal gyrus, cingulate gyrus, and orbitofrontal cortex compared to novices. Conclusions: In conclusion, expert-level school-aged children exhibit distinctive regional structural and intrinsic functional profiles. These non-causal neural variations co-vary with selective cognitive baselines characterizing specialized low-load visual spatial processing rather than a generalized working memory expansion, reflecting brain signatures associated with varying tiers of chess expertise. Full article

The neuronal ceroid lipofuscinoses are the most common group of human paediatric genetic neurodegenerative disorders and have also been reported in multiple animal species. This study explores sheep with the CLN6 disease subtype, which occurs in humans either as a late-infantile or as an adult-onset disease. This study characterised morphometric changes that occur in the brains of 15-month-old Merino sheep with CLN6 disease compared to healthy, adult Merino control brains using ultra-high field magnetic resonance imaging (MRI). The formalin-fixed brains of seven affected and three wild-type control sheep were scanned in a 9.4 T Bruker MRI scanner with a T1-weighted gradient echo. Bioimaging technology ‘Amira-Avizo’ was used to segment each region of interest to create a 3D reconstruction of each ovine brain. Volumetric and statistical analysis of each region of interest found that the thalamus, corpus callosum, occipital cortex, hippocampal region and striatum of affected sheep all experienced a significant loss of volume; 80%, 77%, 73%, 50%, 46% respectively, compared to the control brains. The left side of the affected brain showed a significant reduction of 44%, while the volume of the lateral ventricle non-significantly increased by 43%. However, the cerebellum and arbour vitae of affected sheep lost 15% and 8% of their volume, respectively, which was not significant. These findings are overall consistent with previous gross pathology, histopathology and traditional in vivo MRI findings of ovine CLN6 disease models, but suggest that postmortem ultra-high field MRI of formalin-fixed brains can be a complementary approach to in vivo studies. Volumetric analysis of affected brain regions, including the corpus callosum, thalamus, occipital cortex, striatum, and hippocampal region, can provide valuable biomarkers for assessing the effectiveness of therapeutic interventions in NCL. Full article

Background/Objectives: In emergency settings, it is often infeasible to place patients in an anatomical position for CT scanning. Consequently, emergency brain CT scans of patients with traumatic brain injury frequently demonstrate considerable anatomical misalignment. These inconsistencies compromise the performance of deep learning-based 3D hematoma segmentation. This study aims to enhance segmentation robustness by proposing a registration-based data augmentation strategy utilizing anatomical landmarks. Methods: We propose a framework termed cross-subject registration-based augmentation (CSRA), which uses anatomical landmarks to rigidly register patient CT volumes to selected reference CT volumes and generate anatomically aligned CT-label pairs for training. Results: A total of 339 patients who underwent brain CT imaging were enrolled from a Level 1 trauma center. CSRA-3s achieved the highest mean Dice and IoU and the lowest mean HD95 among the evaluated augmentation strategies. Patient-level paired analysis showed that the most robust statistically supported benefit was a significant reduction in HD95 compared with a conventional geometric augmentation baseline, indicating improved boundary agreement. Conclusions: The proposed augmentation strategy mitigated the effect of anatomic position discrepancies on segmentation performance, particularly in terms of boundary agreement, without modifying existing model architectures. CSRA may serve as a model-agnostic training-time augmentation strategy for improving segmentation robustness in anatomically inconsistent emergency CT imaging, although multicenter external validation is required before clinical deployment. Full article