Background and objectives Skull-involving meningiomas remain ill-defined, resulting in heterogeneous classifications and terminology. Existing schemes mainly describe bone-dura relationships and often overlook the relative burden of the intracranial soft-tissue component. This study aimed to develop and apply a deterministic, atlas-normalized, MRI-based radiological framework for the standardized description of imaging-defined osteomeningiomas and to explore associations between compartmental tumor distribution, radiological phenotype, and clinical presentation. Methods We retrospectively reviewed adults with skull-involving meningiomas at our tertiary neurosurgical center between 2000 and 2024. Tumors were segmented on contrast-enhanced T1-weighted MRI, normalized to MNI152 space, and classified using a deterministic voxel-based radiological framework across osseous, juxta-osseous/dural, and intradural compartments. Tumors were classified as primary osteomeningioma (POM; isolated osseous compartment involvement) or secondary osteomeningioma (SOM; osseous plus adjacent juxta-osseous/dural compartment involvement), with subtypes SOM-I (no intradural extension), SOM-IIA (all three compartments involved, with an osseous component equal to or greater than the intradural component), and SOM-IIB (all three compartments involved, with an intradural component greater than the osseous component). These imaging-defined categories were intended as radiological descriptors of compartmental tumor distribution rather than distinctions between microscopic osseous invasion, reactive hyperostosis, or osseous metaplasia. All analyses were performed at the tumor level, with a predefined sensitivity analysis restricted to one index tumor per patient. Exploratory multivariable logistic regression models were fitted for brain edema, epileptic seizure, raised intracranial pressure, and exophthalmos. Results A total of 168 tumors from 149 patients were analyzed. Distribution was POM in 6 cases (3.6%), SOM-I in 37 cases (22.0%), SOM-IIA in 57 cases (33.9%), and SOM-IIB in 68 cases (40.5%). Convexity predominated in POM but was less common in other subtypes. SOM-IIB had the largest intracranial soft tissue component (29.9 ± 30.2 cm3) and the highest rate of brain edema, whereas SOM-IIA had the largest osseous component (24.8 ± 25.9 cm3). Epileptic seizures and signs of raised intracranial pressure were most frequent in SOM-IIB, exophthalmos in SOM-I, and subcutaneous mass in POM. In exploratory adjusted analyses, SOM-IIB remained associated with brain edema, epileptic seizure, and raised intracranial pressure, whereas SOM-I remained associated with exophthalmos. Conclusions This voxel-based, atlas-normalized MRI framework provides a radiological standardization for the description of skull-involving meningiomas. Rather than establishing histological proof of bone or dural invasion, it standardizes compartmental tumor burden across osseous, juxta-osseous/dural, and intradural spaces. In exploratory analyses, the proposed imaging-defined subtypes were associated with distinct clinicoradiological presentation patterns, which warrant further pathological, multimodal, and external validation.

Catatonia is a severe psychomotor syndrome that occurs across psychiatric diagnoses and is increasingly conceptualized as reflecting neurodevelopmental vulnerability. The anterior cingulate cortex (ACC) plays a central role in motor initiation and cognitive-affective integration and displays substantial interindividual variability in its sulcal morphology, which is established prenatally and remains stable across life. In this MRI study, we examined whether ACC sulcal patterns represent a structural trait marker of catatonia. We analyzed high-resolution T1-weighted images from a hospital-based cohort comprising patients with catatonia (N = 109), psychiatric patients without catatonia (N = 323), and healthy controls (N = 91). The presence of the paracingulate sulcus (PCS) in each hemisphere was determined through blinded visual inspection, and regression analyses tested associations with diagnostic group, adjusting for age, sex, scanner type, intracranial volume, and benzodiazepine and antipsychotic exposure. Patients with catatonia exhibited a significantly reduced prevalence of the left PCS and diminished hemispheric asymmetry compared with both non-catatonic patients and healthy controls. These effects were independent of whether catatonia occurred within psychotic or mood disorders. PCS size did not differ across groups, and sulcal pattern did not correlate with catatonia severity among affected individuals. The findings demonstrate that ACC sulcal deviations are specifically associated with catatonia across diagnostic categories, supporting a neurodevelopmental etiology and reinforcing ACC involvement in its pathophysiology. Early-determined sulcal morphology may represent a trait-level marker contributing to vulnerability for catatonia, with implications for early identification, risk stratification, and targeted intervention strategies.

A large MRI voxel-based atlas mapped intracranial meningioma distributions, linking location to symptoms, management, and malignancy grade, and provided a standardized spatial reference to anticipate presentation and guide clinical decision-making.

The optimal oncological management of adult patients with IDH-mutant CNS WHO grade 3 diffuse gliomas remains debated, particularly given the potential neurotoxic effects of adjuvant therapies on cognition and health-related quality-of-life. We conducted a retrospective single-center cohort study including 87 consecutive patients with IDH-mutant CNS WHO grade 3 glioma who underwent awake craniotomy. Patients were stratified into a surgery only group (n = 23) or an adjuvant oncological group (radiotherapy with or without chemotherapy; n = 64). Neurocognitive outcomes, seizure control, and return-to-work rates were assessed. Voxel-wise lesion-symptom mapping was performed to identify associations between radiation dose distribution and postoperative neurocognitive performance. Patients who did not receive postoperative adjuvant therapy showed better language (100 Grade 3 glioma patients without adjuvant radiochemotherapy showed better neurocognitive and return-to-work outcomes, with no loss of seizure control. Voxel-wise mapping linked cognitive decline to irradiation of specific white matter pathways. Standard adjuvant radiochemotherapy in IDH-mutant grade 3 gliomas may impact neurocognitive outcomes, creating an onco-functional paradox in which functional connectivity meticulously preserved during awake surgery is subsequently compromised by postoperative adjuvant radiochemotherapy. Using voxel-wise lesion-symptom mapping, we show that postoperative impairments in language, executive and social function, attention, and memory correspond to radiation exposure of distinct white matter pathways. Patients who did not receive adjuvant radiochemotherapy demonstrated better neurocognitive performance and faster return-to-work rates. These findings support a more individualized postoperative strategy in patients, aiming to preserve health-related quality-of-life while maintaining safe oncological management. This study provides a practical framework for tailoring adjuvant treatment decisions in modern neuro-oncology.

Introduction Electroconvulsive therapy (ECT) is the most effective intervention for depression, yet no validated biomarkers reliably predict which patients will remit. Large-scale structural and functional dysconnectivity is well-established in major depressive disorder, motivating the use of graph-theoretical metrics to capture network architecture and identify treatment-sensitive markers. Methods In a prospective longitudinal study, 41 adult patients with a major depressive episode undergoing ECT and 24 healthy controls underwent anatomical, diffusion and resting-state functional MRI at baseline (V1). Patients were reassessed after five ECT sessions (V2, n = 31) and two weeks following treatment completion (V3, n = 29). Structural networks were constructed from multi-shell diffusion tractography and functional networks from rs-fMRI. Longitudinal change of global and local efficiency was assessed using linear mixed-effects models, and baseline predictors of remission (MADRS ≤10 at V3) were evaluated using ANOVAs and penalized binomial regression. Results Graph-theory metrics remained stable across timepoints, indicating no large-scale network reorganization during ECT. At baseline, patients showed reduced structural local efficiency relative to controls. Future remitters exhibited lower structural local and global efficiency in fractional anisotropy and neurite density index (NDI)-weighted networks at baseline compared with non-remitters, whereas non-remitters showed reduced functional local efficiency at baseline relative to controls. Penalized regression identified baseline NDI-weighted global efficiency as a significant predictor of remission. Conclusion Baseline structural and functional network profiles distinguished ECT remitters from non-remitters. These findings suggest that non-remission is characterized by reduced functional integration with relatively preserved structural connectivity, while remission is associated with preserved functional organization but lower baseline structural efficiency.

Brain age, as distinct from chronological age, may reveal post-stroke recovery mechanisms, but longitudinal studies tracking brain age are lacking. We explored longitudinal change of brain age post-stroke and its relation to upper limb sensorimotor outcome. T1-weighted MRI at baseline (∼3 weeks) and follow-up (3-7 months) post-stroke was used to estimate brain age. Difference to chronological age was calculated as brain age gap (BAG). Grey and white matter changes and lesion location related to increased brain ageing were investigated, controlling for lesion volume. Association between BAG change and upper limb sensorimotor outcome was studied using linear mixed effects regression. Totally, 114 stroke patients with arm/hand hemiparesis were pooled from three studies. BAG significantly increased from baseline to follow-up, a period of ∼6 months, by a mean of 3.62 years (t = -7.31; P < 0.001). Voxel-based morphometry showed that high BAG change was related to reduced grey and white matter volume ipsilesionally, extending beyond the stroke lesion. Voxel-based lesion symptom mapping showed that lesion to thalamocortical projections, internal capsule and corona radiata related to accelerated brain ageing. BAG change was significantly associated with motor outcomes in the sub-acute to chronic phase, as expressed by Fugl-Meyer assessment (β = -5.62, SE = 2.81, t = -2.00, P = 0.05), maximum grip strength (β = -0.14, SE = 0.04, t = -3.36, P = 0.001) and dexterity assessment (β = -0.09, SE = 0.04, t = -2.17, P = 0.03). We demonstrate increased brain ageing within the first few months post-stroke. This secondary neurodegeneration was negatively related to motor outcome. Brain age may be a valid whole-brain probe of individual secondary post-stroke degeneration, relevant for predicting recovery and identifying targets of neural plasticity.

DICOM is an industry-standard for medical imaging data targeted at interoperability across systems. This enables transfer, storage and processing of imaging data regardless of the manufacturer. Pragmatically, manufacturers often store detailed acquisition parameters in private rather than public DICOM tags. In parallel, the DICOM standard itself has gradually evolved by introducing new public tags and properties to better capture emerging imaging technologies. Accurately extracting these details is essential for reproducible neuroimaging research. To address this need, we created a series of DICOM datasets illustrating how various manufacturers encode acquisition details that are critical for modern processing and analysis. These minimal test cases, covering CT and MR modalities, highlight manufacturer-specific conventions, including the use of public tags, private tags, and proprietary data structures. For each DICOM dataset, we provide corresponding NIfTI-formatted images with metadata JSON files following the BIDS standard, using consistent terminology to mitigate variations in how manufacturers encode acquisition details. Our repository provides validation datasets for any tool that is intended to extract acquisition details from medical imaging data.

Introduction Dans l'Accident Vasculaire Cérébral ischémique (AVCi), le seuil IRM de Coefficient Apparent de Diffusion initial (ADC0, de 620×10-6mm2/s) pour délimiter le cœur ischémique a été établi par intersection entre l'infarctus final et l'ADC0, sans correction des déformations liée à l’évolution de l'infarctus. Nous proposons une nouvelle méthodologie pour évaluer ce seuil chez des patients recanalisés précocement. Matérieletméthodes Les patients avec AVCi traités dans notre centre (GHU Paris) par recanalisation précoce et complète (<90minutes post IRM-initiale, IRM0, mTICI2c ou 3), avec IRM de suivi à 24heures (IRM24h) ont été rétrospectivement inclus. Pour la correction d’’dème (CO), l'IRM0 et l'IRM24h ont été co-registrées par une méthode non-linéaire. L'infarctus final (Infarct24h) était délimité par contourage manuel de l'hypersignal sur l'IRM24h, sans et avec CO. Le Dice Score Coefficient (DSC), reflétant le meilleur chevauchement entre l'Infarct24h et l'ADC0, était calculé en faisant varier le seuil d'ADC de 400x10-6mm2/s à 1200x10-6mm2/s (par intervalles de 1x10-6mm2/s). L'ADC individuel optimal (optADC) correspondait au DSC le plus élevé, calculé sans ou avec CO. Les optADC, DSC et volumes ischémiques initiaux obtenus après application du seuil optADC moyen étaient comparés par test de Wilcoxon apparié. Résultats Chez 56 patients inclus, l'optADC moyen et le DSC moyen avant CO étaient respectivement de 593±41×10-6mm2/s et 0.52. L'application d'une CO permettait d'obtenir un optADC significativement plus bas (585±41×10-6mm2/s, p<0.001) et un DSC plus élevé (0.55, p<0.001). Les volumes ischémiques initiaux déterminés avec l'optADC moyen étaient respectivement de 15.4±24mL et 14.2±23mL sans et après CO (p<0.001). Avec ou sans CO, l'optADC était positivement corrélé au NIHSS initial (p=0.002) et négativement à l'ASPECT initial (p=0.001). Conclusion Notre méthode originale conduit à reconsidérer le seuil d'ADC actuellement reconnu pour délimiter le cœur ischémique et met en lumière l'intérêt de corriger l’œdème sur l'IRM24h.

BACKGROUND AND OBJECTIVES:The widely used apparent diffusion coefficient (ADC) threshold of 620 × 10-6 mm2/s to automatically delineate the ischemic core on diffusion-weighted imaging (DWI) was established in recanalized patients after IV thrombolysis. However, recanalization was assessed 3-6 hours after treatment, which may entail substantial core growth and in turn ADC threshold overestimation, while follow-up imaging was performed at day 30, implying potential infarct shrinkage. Other studies typically used follow-up MRI scans on days 2-5 but did not correct for vasogenic edema, implying potential infarct overestimation. We re-evaluated the ADC threshold in patients with very early recanalization after endovascular therapy (EVT), using 24-hour follow-up MRI and applying edema correction (EC). METHODS:Consecutive patients with modified Treatment In Cerebral Infarction 2c-3 recanalization within 90 minutes after baseline MRI and who underwent follow-up MRI ≈24 hours were identified from our center's EVT registry (2012-2021). Basilar occlusions and small DWI lesions (<3 mL) were excluded. The baseline ischemic lesion and the final infarct (Infarct24h) were manually delineated on initial DWI (DWI0) and 24-hour DWI. EC was performed by applying nonlinear coregistration of follow-up MRI onto MRI0. The intersection between DWI0 and Infarct24h was overlaid onto the baseline ADC map. A receiver operating characteristic analysis compared "core" with "noncore" voxels (obtained by varying the ADC threshold) across patients, and the Youden index was computed to determine the optimized ADC threshold (OptADC) without and with EC. RESULTS:Among 1,024 patients, 56 were eligible and included (median age 73 years; 52% female; median [interquartile range (IQR)] NIH Stroke Scale score 12 [8-19], median MRI0-to-recanalization delay 70.5 [52-81] minutes). The OptADC was 611 × 10-6 mm2/s (area under the curve = 0.704; sensitivity = 61.6%; specificity = 70.4%; Youden index = 0.320) and 612 × 10-6 mm2/s without and with EC, respectively. The median (IQR) individual-OptADC was 621.5 (585.5-672.8) × 10-6 mm2/s. The baseline core volume was significantly (p < 0.001), but only marginally, smaller using the optimized vs the reference ADC threshold (16.4 ± 25 mL and 17.0 ± 26 mL, respectively; volume difference larger than 3 mL in 1 patient only). DISCUSSION:We revisited the ADC core threshold using a stringent methodology, including EVT-induced ultra-early documented complete recanalization, ≈24-hour follow-up MRI, and EC. The resulting OptADC was marginally smaller than the reference threshold and, accordingly, modestly influenced the measured baseline core volume. The methodological refinements implemented here, including EC, seem essential for future research exploring core ADC.

PURPOSE:The purpose of this study was to validate a deep learning algorithm that generates T2*-weighted images from diffusion-weighted (DW) images and to compare its performance with that of true T2*-weighted images for hemorrhage detection on MRI in patients with acute stroke. MATERIALS AND METHODS:This single-center, retrospective study included DW- and T2*-weighted images obtained less than 48 hours after symptom onset in consecutive patients admitted for acute stroke. Datasets were divided into training (60 %), validation (20 %), and test (20 %) sets, with stratification by stroke type (hemorrhagic/ischemic). A generative adversarial network was trained to produce generative T2*-weighted images using DW images. Concordance between true T2*-weighted images and generative T2*-weighted images for hemorrhage detection was independently graded by two readers into three categories (parenchymal hematoma, hemorrhagic infarct or no hemorrhage), and discordances were resolved by consensus reading. Sensitivity, specificity and accuracy of generative T2*-weighted images were estimated using true T2*-weighted images as the standard of reference. RESULTS:A total of 1491 MRI sets from 939 patients (487 women, 452 men) with a median age of 71 years (first quartile, 57; third quartile, 81; range: 21-101) were included. In the test set (n = 300), there were no differences between true T2*-weighted images and generative T2*-weighted images for intraobserver reproducibility (κ = 0.97 [95 % CI: 0.95-0.99] vs. 0.95 [95 % CI: 0.92-0.97]; P = 0.27) and interobserver reproducibility (κ = 0.93 [95 % CI: 0.90-0.97] vs. 0.92 [95 % CI: 0.88-0.96]; P = 0.64). After consensus reading, concordance between true T2*-weighted images and generative T2*-weighted images was excellent (κ = 0.92; 95 % CI: 0.91-0.96). Generative T2*-weighted images achieved 90 % sensitivity (73/81; 95 % CI: 81-96), 97 % specificity (213/219; 95 % CI: 94-99) and 95 % accuracy (286/300; 95 % CI: 92-97) for the diagnosis of any cerebral hemorrhage (hemorrhagic infarct or parenchymal hemorrhage). CONCLUSION:Generative T2*-weighted images and true T2*-weighted images have non-different diagnostic performances for hemorrhage detection in patients with acute stroke and may be used to shorten MRI protocols.

Purpose: In acute ischemic stroke, the typically employed 620×10 -6 mm 2 /s ADC threshold to automatically delineate the core on diffusion-weighted imaging (DWI) was established by overlaying the manually defined core onto the baseline ADC (ADC 0 ) in recanalized patients following intravenous thrombolysis. However, recanalization was assessed 3-6hrs after treatment, which may entail substantial core growth, and follow-up imaging was at 90 days, implying infarct underestimation due to atrophy. Other studies typically used day 2-5 follow-up MRI (FU-MRI) but did not correct for vasogenic edema, implying infarct overestimation. We re-evaluated the ADC core threshold in patients with early recanalization following EVT, and applied edema correction on 24h post-EVT MRI. Methods: From our center’s EVT registry (2012-2021), all consecutive mothership patients with mTICI2c-3 recanalization within 90min after baseline MRI (MRI 0 ) and who underwent FU-MRI at 24h post-EVT were included. Basilar occlusions and small DWI lesion (<3mL) were excluded. Edema correction was performed by applying non-linear coregistration of FU-MRI onto MRI 0 using the ANTs library. The baseline ischemic lesion was manually delineated on initial DWI (DWI 0 ), and the final infarct ROI (Infarct 24h ) was manually delineated on non-corrected and edema-corrected FU-DWI. The intersection between DWI 0 and Infarct 24h (the manual core) was then overlayed onto the ADC 0 , after exclusion of CSF and spurious ADC voxels. A ROC analysis then compared all ‘core’ voxels to ‘non-core’ voxels across patients, and the Youden Index was computed to determine the optimal ADC (OptADC) core threshold. Results: Across 56 eligible patients (median age: 73yrs; female 52%; median [IQR] NIHSS: 12 [8-19]), the mean ADC within the manual core was 609.8 (± 66.8)×10 -6 mm 2 /s and the OptADC was 584×10 -6 mm 2 /s without edema correction. The corresponding values with edema correction were 598.7 (± 69.1)×10 -6 mm 2 /s and 636×10 -6 mm 2 /s, respectively (both p<0.001 vs. non-corrected data). Conclusion: We report here the first ADC core threshold values derived from EVT-induced ultra-early complete recanalization using 24h follow-up MRI, with and without edema correction. Our strict selection criteria likely partly explain the substantially lower uncorrected ADC threshold found as compared to the classic 620×10 -6 mm 2 /s threshold. We also document the effect of edema correction on ADC threshold calculation when an early follow-up MRI is used.

BACKGROUND:Neuromodulation of deep brain regions has shown promise for treatment-resistant depression (TRD). However, it currently requires neurosurgical electrode implantation, posing significant risks and limiting widespread use while TRD affects around 100 million people worldwide. Low-intensity transcranial ultrasound stimulation (TUS) could allow precise and non-invasive deep neuromodulation, provided that the challenge of the defocusing effects of the skull is tackled. OBJECTIVE/HYPOTHESIS:Here, we present the development of a portable and neuronavigated TUS prototype based on the use of patient-specific metamaterials (metalens) that correct for skull-induced aberrations. We then present the first application of metalens-based Transcranial Ultrasound Stimulation (mTUS) in TRD. The primary objective was to assess the safety and efficacy of mTUS targeting on individual level specific white matter tracts of the subcallosal cingulate involved in TRD. METHODS:The safety and precision of this device was addressed through a series of numerical simulations and experimental measurements on ex vivo human skulls. Five participants with TRD were included in this open-label study (ClinicalTrials.gov identifier: NCT06085950) and underwent an intensive 5-day course of mTUS with a total of 25 sessions of 5 min each. RESULTS:No serious adverse events occurred during the study. By day 5 of treatment, depression severity was reduced by an average of 60.9 % (range: [30 %-83.9 %]), and four out of five patients qualified as responders, with two of them in remission. CONCLUSIONS:This study provides first-in-human evidence of the potential of mTUS as a precise, safe and effective non-invasive neuromodulation technique for neuropsychiatric disorders involving deep brain regions, offering a safer and more accessible alternative to invasive approaches.

Introduction Electroconvulsive therapy (ECT) induces an increase in hippocampal volume presumed to reflect neurogenesis in severely depressed patients. We hypothesized that Neurite Orientation Dispersion and Density Imaging (NODDI) provides in vivo evidence of hippocampal neurogenesis following ECT. Methods This prospective longitudinal study included 43 depressed patients treated by ECT. Three sequential evaluations (V1: baseline, V2: at 2 weeks into ECT, V3: 14 days within completing ECT) included a 3T MR-scan with 3D T1-weighted and multi-shell diffusion (b = 200/1500/2500 s/mm2, 30/45/60 directions) sequences and clinical assessment with depression scales. Q-ball, Diffusion Tensor and NODDI models provided the following metrics: axial (AD), radial (RD) and mean diffusivity (MD), fractional anisotropy (FA) and generalized FA (GFA), neurite density index (NDI), isotropic fraction (Fiso), neurite orientation and dispersion index (ODI). FreeSurfer was used to extract whole hippocampal and subfields volumes from T1-weighted images. A linear mixed-effect model assessed the changes over time in hippocampal volumes and mean diffusion metrics, and their relationship with clinical response was analyzed with ANOVA. Bonferroni corrections were applied. Results 107 MRI were obtained at V1 (n = 43), V2 (n = 34) and V3 (n = 30) from 43 patients. Mean (± SD) interval between V1-V3 was 70 ± 25 days. Diffusion metrics in the hippocampus were: at V2, a decrease in left GFA, right AD, bilateral Fiso, and a bilateral ODI increase. Additionally, at V3, we observed a left MD decrease, bilateral AD decrease, right NDI increase, and bilateral ODI increase. Notably, NDI and Fiso changes were localized to the dentate gyrus but not to the hippocampal tail. ECT-responders showed a significant right hippocampus volume increase at 2 weeks into ECT. Conclusion After ECT, the observed increase in hippocampal volume is accompanied by bilateral changes in NODDI parameters, consistent with hippocampal neuroplasticity.

BACKGROUND:Correlating the human connectome with clinical responses elicited during intraoperative brain mapping helps understanding of the intrinsic organization of the human brain. Methods for locating eloquent sites on neuroimaging are not standardized. In the present study, we standardized a methodology for locating subcortical eloquent sites identified during intraoperative mapping for awake brain tumor resection on a reference brain template. METHODS:Subcortical eloquent sites were tagged by co-registration of intraoperative photographs with early postoperative MRI ( < 48 h). Neuroimaging data were normalized into MNI152 space. To assess whether the location of subcortical eloquent sites on the MNI template was concordant with the expected brain connectivity, we compared each subcortical eloquent site with the Human Connectome Project 1065 probabilistic tractography atlas. RESULTS:We analyze 290 subcortical eloquent sites identified during 69/90 awake surgeries. 2/290 (0.7%) subcortical eloquent sites identified intraoperatively do not intersect with a fiber tract according to the reference atlas. Among the other 288 that successfully intersect with, at least, one white matter tract, 255/288 (88.5%) have a clinical response elicited intraoperatively that is congruent with the intersected white matter tract. In the remaining 33/288 (11.5%) functional incongruent and the 2/290 (0.7%) anatomical incongruent subcortical sites, the minimal mean distance between the eloquent site and a congruent with matter tract is 3.6 ± 4.4 mm (range 1.0-23.9, median 3.6, interquartile range 2.5-5.4). CONCLUSIONS:We propose a standardized methodology to locate with accuracy on a reference brain template subcortical eloquent sites identified intraoperatively during functional brain mapping using direct electrical stimulations under awake condition.