OBJECTIVE:Brain-predicted age difference (BrainAGE) is a neuroimaging biomarker reflecting brain health, with potential implications for post-stroke recovery. However, training robust BrainAGE models requires large, diverse datasets, often restricted by privacy and regulatory concerns. This study evaluates the performance of federated learning (FL) for BrainAGE estimation in ischemic stroke patients treated with mechanical thrombectomy, and investigates its association with clinical phenotypes and functional outcome. METHODS:We used pre-treatment FLAIR brain images from 1674 stroke patients across 16 hospital centers. We implemented standard machine learning and deep learning models for BrainAGE estimates under three data management strategies: centralized learning (pooled data), FL (local training at each site), and single-site learning. We reported prediction errors and examined associations between BrainAGE and vascular risk factors (e.g., diabetes mellitus, hypertension, smoking), as well as functional outcome at three months post-stroke. Logistic regression evaluated BrainAGE's predictive value for this outcome, adjusting for age, sex, vascular risk factors, stroke severity, time between MRI and arterial puncture, prior intravenous thrombolysis, and recanalization outcome. RESULTS:While centralized learning yielded the most accurate predictions, FL consistently outperformed single-site models. BrainAGE was significantly higher in patients with diabetes mellitus across all models. Comparisons between patients with good and poor functional outcome, and multivariate predictions of these outcome showed the significance of the association between BrainAGE and post-stroke recovery. CONCLUSION:FL enables accurate age predictions without data centralization. The strong association between BrainAGE, vascular risk factors, and post-stroke recovery highlights its potential for personalized prognostic modeling in stroke care.

Neurodegenerative diseases such as Alzheimer's disease (AD) and frontotemporal dementia (FTD) exhibit substantial biological and clinical heterogeneity, complicating diagnosis, subtype characterization, and prediction of disease progression. We introduce PatientSpace, a multimodal graph-based latent representation framework designed to model neurodegenerative disease heterogeneity using T1-weighted MRI and FDG-PET. PatientSpace is built upon a structured variational autoencoder that integrates multimodal neuroimaging features while organizing patients within a latent space constrained by age, diagnosis, and a consistency regularization term encouraging similarity between neuroimaging phenotypes. This design enables the construction of an interpretable patient graph in which neighborhood relationships reflect biological similarity. Applied to cohorts of cognitively normal individuals, AD, and FTD patients, PatientSpace revealed multiple disease clusters associated with distinct neuroimaging patterns and clinical severity. Diagnostic classification achieved performance comparable to state-of-the-art deep learning models, while graph-based neighborhood inference enabled prediction of structural volumes, metabolic activity, and cognitive severity. Projection of mild cognitive impairment (MCI) subjects from an independent cohort further showed that cluster membership was associated with differential risks of dementia conversion and distinct longitudinal trajectories. Together, these results demonstrate that PatientSpace provides an interpretable framework linking multimodal neuroimaging representations to disease subtypes, patient-level characterization, and progression modeling in neurodegenerative disorders.

Language is essential to social communication. Its complexity and hierarchical organization, from low-level operations to high-order integrative processes, may provide valuable diagnostic insights in neurodegeneration beyond classical aphasia syndromes. However, systematic investigations in these conditions, particularly in behavioural variant frontotemporal degeneration (bvFTD), remain scarce. To refine differential diagnosis, an exhaustive characterization of language functions is required. We systematically compared multi-level language functioning across bvFTD, Alzheimer's disease (AD) and primary psychiatric disorders (PPD), within an integrative neurolinguistic framework distinguishing lexical, syntactic and discursive levels, together with a cross-modal transposition/transcoding dimension. A total of 85 patients (including 34 with bvFTD, 30 with AD, 21 with PPD) and 40 matched healthy controls underwent an extensive language assessment using the GREMOTs battery. Composite, quantitative as well as qualitative indices were computed for each linguistic level. Structural MRI data were analysed using voxel-based morphometry (P < 0.05 corrected for multiple comparisons). All clinical groups exhibited lexical impairments relative to controls, with bvFTD presenting the most severe and widespread deficits across fluency, naming and comprehension (partial eta-squared, ηp 2, ranging from 0.18 to 0.49). AD and PPD showed milder lexical inefficiencies (ηp 2 = 0.11-0.39 and ηp 2 = 0.07-0.29, respectively). Syntactic processing was also more impaired in bvFTD (ηp 2 = 0.03-0.27) than in AD and PPD (ηp 2 = 0.02-0.07 and ηp 2 = 0.00-0.15, respectively). At the discourse-level, bvFTD displayed key deficits (ηp 2 = 0.05-0.29), with pervasive pragmatic breakdowns whereas AD and PPD showed milder integrative deficits with preserved global coherence (ηp 2 = 0.00-0.17 and ηp 2 = 0.02-0.13). Transcoding and transposing tasks revealed minor deficits, mainly in bvFTD (ηp 2 = 0.01-0.25). A logistic regression identified that a subset of 12/23 tasks accurately classified 85.9% of bvFTD cases (sensitivity: 57.6%; specificity: 95.6%). The analysis of the types of responses (including errors) allowed to provide a more comprehensive group profiling. In bvFTD, the decrease of language performance related to widespread frontotemporal and posterior (including cerebellar) atrophy, whereas AD showed more restricted frontal and temporal involvement. PPD displayed smaller fronto-temporal, insular and precuneal associations. In conclusion, these findings delineate a graded, multi-level linguistic profile across neurodegenerative and psychiatric conditions. bvFTD is mainly characterized by pervasive lexical and discursive-pragmatic impairments, alongside syntactic difficulties, while AD and PPD primarily show lexical inefficiencies with preserved syntax. Convergent neural evidence supports a distributed network model of language integrating frontal-insular control and temporal semantic systems. Embedding such multi-level assessments into clinical practice could enhance diagnostic precision and provide valid cognitive endpoints for future trials.

Accurate delineation of acute ischemic stroke lesions in MRI is a key component of stroke diagnosis and management. In recent years, deep learning models have been successfully applied to the automatic segmentation of such lesions. While most proposed architectures are based on the U-Net framework, they primarily differ in their choice of loss functions and in the use of deep supervision, residual connections, and attention mechanisms. Moreover, many implementations are not publicly available, and the optimal configuration for acute ischemic stroke (AIS) lesion segmentation remains unclear. In this work, we introduce ISLA (Ischemic Stroke Lesion Analyzer), a new deep learning model for AIS lesion segmentation from diffusion MRI, trained on three multicenter databases totaling more than 1500 AIS participants. Through systematic optimization of the loss function, convolutional architecture, deep supervision, and attention mechanisms, we developed a robust segmentation framework. We further investigated unsupervised domain adaptation to improve generalization to an external clinical dataset. ISLA outperformed two state-of-the-art approaches for AIS lesion segmentation on an external test set. Codes and trained models will be made publicly available to facilitate reuse and reproducibility.

Auditory–verbal hallucinations (AVHs) are among the most disabling symptoms of schizophrenia and often persist despite the use of adequate antipsychotic treatment. Conventional low-frequency repetitive transcranial magnetic stimulation (rTMS) targeting the T3P3 scalp site has demonstrated limited efficacy, likely due to interindividual variability in AVH-related brain networks. In this multicenter, randomized, double-blind phase 3 trial, 70 patients with drug-resistant AVHs received active 1-Hz rTMS targeted either with an individualized fMRI-based symptom-capture procedure or by using conventional T3P3 localization. fMRI-guided rTMS yielded a greater reduction in Auditory Hallucination Rating Scale (AHRS) scores at one month (mean difference, −5.43; 95% CI, −8.92 to −1.94), and the effects were sustained at three and six months. The number-needed-to-treat for neuroguided rTMS was 3.5. Clinical response was associated with greater E-field overlap with AVH-related networks. These findings demonstrate that fMRI-guided neuronavigation increases rTMS efficacy, thus supporting its use to optimize the treatment of drug-resistant AVHs in schizophrenia.

Multi‑site magnetic resonance imaging (MRI) studies enable studying brain structure across diverse populations, but scanner‑related variability remains a critical barrier to pooled analyses. Here, we introduce NeuroHarm‑Kit, the first open‑source, end‑to‑end toolbox that unifies state‑of‑the‑art deep‑learning harmonization models (STGAN, HACA3, MURD, DISARM++, IGUANe) for 3D T1‑weighted scans. NeuroHarm‑Kit provides standardized preprocessing and pretrained weights, thereby enabling reproducible, head‑to‑head comparisons. Applied to traveling‑subject and healthy aging cohorts, NeuroHarm‑Kit demonstrates that different methods excel in distinct metrics: some optimize intensity‑distribution alignment, others preserve anatomical fidelity or biological information. No single approach uniformly outperforms across all criteria, underscoring the need to tailor harmonization choices to specific research objectives. The toolbox provides researchers with a ready-to-use framework to compare harmonization strategies and select the most appropriate method for their data and study design. With its modular design, NeuroHarm‑Kit can be extended to additional modalities and new methods. By lowering methodological barriers and promoting transparent benchmarking, this toolbox aims to accelerate innovation and reproducibility in multi‑site neuroimaging harmonization. GitHub project is available at https://github.com/barna-hache/NeuroHarm-kit.

Patient access to radiology reports has heightened the need for patient-friendly communication. Automated generation of patient-centered summaries using large language models (LLMs) is a promising solution. However, their use on real-life reports is limited by privacy concerns. Here, we evaluate the safety and effectiveness of on-premise, privacy-preserving LLMs for generating lay summaries of real French brain MRI reports for emergency presentations of headache. In this retrospective study, we sampled 105 brain MRI reports (January–December 2022) for radiologist evaluation and a subset of 30 reports for non-physician evaluation. Three open-weights models (Llama 3.3 70B, Athene V2, Mistral Small) generated French lay summaries via a single standardized prompt. Radiologists’ mean ratings across models were high for exactness (4.10, 95% CI: 4.04–4.16), exhaustiveness (4.34, 95% CI: 4.29–4.39), didacticness (3.83, 95% CI: 3.79–3.88), and readiness for clinical use (3.84, 95% CI: 3.79–3.89). Non-physicians reported higher perceived understanding with summaries, from 2.85 (95% CI: 2.67–3.04) to 4.27 (95% CI: 4.15–4.38, p < 0.001). The correct identification rate for reports increased from 75.2% to 83.6% (p < 0.001). The ability to identify causal findings also improved, from 80.6% to 84.8% (p < 0.001) overall. Overall error rate in LLM-generated lay summaries was 19.7% (62/315), warranting expert oversight.

Over a third of minor stroke patients experience post-stroke cognitive impairment (PSCI), but no validated tools exist to identify at-risk patients early. This study investigated whether disconnection features derived from infarcts and white matter hyperintensities (WMH) could serve as markers for short- and long-term cognitive decline in first-ever minor ischemic stroke patients. First-ever minor ischemic stroke patients (NIHSS ≤ 7) were prospectively followed at 72-h, 6 months, and 36 months post-stroke with cognitive tests and brain MRI. Infarct and WMH volumes were semi-automatically assessed on DWI and FLAIR sequences. Bayesian tract-based disconnection models estimated remote pathological effects of infarcts and WMH. Associations between disconnection features and cognitive outcomes were analyzed using canonical correlation analyses, adjusted for age, education, and multiple comparisons. Among 105 patients (31% female, mean age 63 ± 12 years), infarct volume averaged 10.28 ± 17.10 cm 3 and predominantly involved the middle cerebral artery territory (83%). WMH burden was higher in frontal periventricular white matter. Infarct-based features did not significantly relate to PCSI. However, a WMH-derived disconnection factor, involving commissural and frontal tracts, and the right superior longitudinal fasciculus, was significantly associated with PSCI at 6 months (OR = 9.96, p value = 0.02) and 36 months (OR = 12.27, p value = 0.006), particularly in executive/attention, language, and visuospatial domains. This factor, unrelated to WMH volume, outperformed demographic and clinical predictors of PSCI. WMH-induced disconnection may be associated with short- and long-term PSCI in minor stroke. Routine MR-derived features could identify at-risk patients for rehabilitation trials.

Over the past years, social cognition has been envisaged as a promising domain to distinguish behavioral variant frontotemporal degeneration (bvFTD) from its main differential diagnoses that is primary psychiatric disorders (PPD). The core-processes approach, which has emphasized the importance of emotion recognition and mentalizing, has been particularly useful to better characterize each condition and enhance the earliness of FTD’s diagnosis. However, new findings evidencing conflicting results regarding the ability of social cognition to distinguish bvFTD from PPD have underlined the importance of moving beyond the core processes approach. We reviewed all cases with a suspission of bvFTD in the last 8 years in the Lille memory clinic, at least followed-up for 24 months with a neuropsychological assessment and an MRI and/or PET-scan. We then applied a quantitative comparison approach based on total scores, then a qualitative approach, based on responses and errors types. Then, neuroimaging analyses were run, and biomarkers were analyzed. Data of 56 patients with a probable to certain bvFTD and 47 patients with a primary psychiatric disorders (late major depression, bipolar disorders, schizophrenia…) were analyzed at classical and social cognitive (mini-SEA) tests. Overall, clinical groups were not different on executive functionning, attention, motor & perceptual functions. Minor differences were retrieved in memory and langage processing. Important differences were retrieved in social cognition. Qualitative differences were retrieved in facial emotion recognition (inter & intra valence errors) and mentalizing (type of mental inference, emotional labelling), as well as memory functioning (primacy/recency ratio, intrusions). Anatomical and functional brain networks involved showed a combination of overlaping and distincts areas. Regarding biomarkers, NFLs showed promising results with AUC = 0.88). While the usual approach (considering general or subscores scores) may not be the more efficient way, a more qualitative neuropsychological approach has the potential to provide relevant cognitive markers for the clinical distinction between bvFTD and PPD, particularly regarding social cognition.

Objectifs Les méthodologies d’étude du connectome fonctionnel sont mises en avant. La trajectoire de développement du connectome fonctionnel cérébral aux stades fœtal et infantile est examinée, ainsi que les implications des lésions cérébrales sur le développement. Méthodes La méthodologie et les principaux résultats des études utilisant l’imagerie par résonance magnétique fonctionnelle au repos ont été passés en revue. Des techniques analytiques, telles que la connectivité basée sur des régions d’intérêt, l’analyse en composantes indépendantes et la théorie des graphes pour l’analyse des réseaux sont explorées pour fournir une compréhension globale de l’organisation fonctionnelle du cerveau en développement. Résultats Le développement précoce du cerveau est marqué par des changements significatifs dans la connectivité fonctionnelle. Des études sur les fœtus ont démontré une augmentation de la connectivité inter-hémisphérique avec l’âge gestationnel, tandis que des recherches sur les nouveau-nés ont identifié la maturation progressive de réseaux tels que celui du sensorimoteur et du mode par défaut. Des schémas de connectivité altérés, notamment chez les nourrissons prématurés, ont été révélés. Conclusion Les avancées des technologies de neuroimagerie ont fourni des informations cruciales sur le connectome précoce, offrant ainsi un potentiel de diagnostic précoce et d’interventions ciblées pour les troubles neurodéveloppementaux. Comprendre ces premières étapes de la connectivité cérébrale est essentiel pour influencer les résultats du développement et améliorer les soins de santé pédiatrique.

INTRODUCTION:Early-onset Alzheimer's disease (EOAD) population is a clinically, genetically and pathologically heterogeneous condition. Identifying biomarkers related to disease progression is crucial for advancing clinical trials and improving therapeutic strategies. This study aims to differentiate EOAD patients with varying rates of progression using Brain Age Gap Estimation (BrainAGE)-based clustering algorithm applied to structural magnetic resonance images (MRI). METHODS:A retrospective analysis of a longitudinal cohort consisting of 142 participants who met the criteria for early-onset probable Alzheimer's disease was conducted. Participants were assessed clinically, neuropsychologically and with structural MRI at baseline and annually for 6 years. A Brain Age Gap Estimation (BrainAGE) deep learning model pre-trained on 3,227 3D T1-weighted MRI of healthy subjects was used to extract encoded MRI representations at baseline. Then, k-means clustering was performed on these encoded representations to stratify the population. The resulting clusters were then analyzed for disease severity, cognitive phenotype and brain volumes at baseline and longitudinally. RESULTS:The optimal number of clusters was determined to be 2. Clusters differed significantly in BrainAGE scores (5.44 [± 8] years vs 15.25 [± 5 years], p < 0.001). The high BrainAGE cluster was associated with older age (p = 0.001) and higher proportion of female patients (p = 0.005), as well as greater disease severity based on Mini Mental State Examination (MMSE) scores (19.32 [±4.62] vs 14.14 [±6.93], p < 0.001) and gray matter volume (0.35 [±0.03] vs 0.32 [±0.02], p < 0.001). Longitudinal analyses revealed significant differences in disease progression (MMSE decline of -2.35 [±0.15] pts/year vs -3.02 [±0.25] pts/year, p = 0.02; CDR 1.58 [±0.10] pts/year vs 1.99 [±0.16] pts/year, p = 0.03). CONCLUSION:K-means clustering of BrainAGE encoded representations stratified EOAD patients based on varying rates of disease progression. These findings underscore the potential of using BrainAGE as a biomarker for better understanding and managing EOAD.

BACKGROUND AND OBJECTIVES:Cerebrovascular reserve (CVR) is a key physiological mechanism allowing the brain to adapt to fluctuating perfusion, particularly relevant in the management of neurovascular disorders such as idiopathic (iMM) and syndromic moyamoya (sMM). Although 99m Tc-HMPAO SPECT with acetazolamide is commonly used for CVR assessment, it faces limitations including low spatial resolution, artifacts, and variability in interpretation. This study primarily aims to evaluate a novel, semiautomated, and more objective method for interpreting HMPAO SPECT in CVR assessment. As a secondary objective, the method is applied to a cohort of patients who underwent revascularization surgery for iMM or sMM. METHODS:A retrospective analysis was performed on prospectively collected data from a tertiary neuroscience center, including 20 adult patients with iMM (n = 9) or sMM (n = 11). Clinical and imaging data were reviewed. 99m Tc-HMPAO SPECT images were assessed independently by 2 nuclear medicine physicians blinded to clinical details. Images were registered to T1-weighted MRI and overlaid with an arterial territory atlas. Territories classified as healthy by experts were defined as true negatives; all others as altered. Statistical comparisons were made using Student's t-tests with false discovery rate correction. RESULTS:Among the 20 patients (12 females), patients with sMM were older and had more cardiovascular risk factors. The proposed method significantly discriminated between altered and healthy perfusion territories. Compared with expert interpretation, the method demonstrated specificities of 93% (iMM) and 92% (sMM), with negative predictive values of 80% and 75%%, respectively. CONCLUSION:Although 99m Tc-HMPAO remains a validated modality for CVR assessment, its interpretation is operator-dependent. The proposed semiautomated method offers high specificity and greater objectivity, supporting its integration into clinical workflows. Further multicenter validation is warranted.

OBJECTIVE:To assess changes in brain functional connectivity associated with the painful nature of peripheral polyneuropathy. METHODS:Resting-state functional magnetic resonance imaging (rs-fMRI) was performed in 26 patients with painful or painless polyneuropathy. According to previously published results, connectivity was studied regarding the default mode network (DMN), intrathalamic and thalamocortical connections, and, the different brain networks (pain matrices) involved in the "nociceptive", "attentional" and "emotional" aspects of the chronic pain experience. RESULTS:No change in DMN connectivity was found between groups. Thalamocortical connectivity was reduced in patients with painful polyneuropathy, especially for the thalamic cluster connected to the motor cortex, while intra-thalamic (mediolateral) connectivity was increased in patients with painless polyneuropathy. Intra-connectivity was increased within the pain matrices, especially the "nociceptive" matrix, in patients with painful polyneuropathy, while inter-connectivity was increased between the "attentional" and "emotional" pain matrices in patients with painless polyneuropathy. Increased connectivity between the posterior insula and parietal operculum positively correlated with the neuropathic pain symptom score and impact of pain on daily functioning. CONCLUSIONS:Painful polyneuropathy was characterized by increased intra-connectivity within each pain matrix and reduced thalamocortical connectivity of certain thalamic clusters, notably linked to the motor cortex. Conversely, painless polyneuropathy was characterized by increased connectivity within the thalamus and between the different pain matrices. Although various methodological limitations must be acknowledged (small sample size, lack of a control group of healthy subjects or measurement of pain intensity during neuroimaging examination), these results provide new information on the changes in brain connectivity associated with painful polyneuropathies. This study also brings new arguments to explain the efficacy of motor cortex stimulation in the treatment of chronic neuropathic pain.

OBJECTIVE:Neuropsychological assessment of social cognition has traditionally focused on mentalizing and emotions recognition. Recently developed digital measures allow clinicians to capture direct markers of abnormal social interactions, but they have not yet been used to distinguish neurological from psychiatric populations. This study examined prosodic alterations and explored structural neural correlates in behavioral variant frontotemporal degeneration (bvFTD) versus late-onset or atypical psychiatric disorders (LOAPD) and healthy controls (HCs). METHOD:We analyzed audio recordings from 31 patients with bvFTD, 15 patients with LOAPD, and 39 HCs across two speech samples: an anamnestic interview and a narrative task. Fundamental frequency (f₀) metrics were extracted to assess between-group differences and to identify voxel-based morphometry correlates of prosodic alteration. RESULTS:Compared with HCs, patients with bvFTD showed a reduced f₀ range in both the anamnestic interview (p = .025, η² = .09, 95% CI [-23.83, -1.04]) and the narrative task (p = .002, η² = .14, 95% CI [-36.21, -4.74]). In the anamnestic interview, both bvFTD (p = .010, 95% CI [-36.59, -6.23]) and LOAPD (p = .012, 95% CI [-46.30, -8.81]) groups exhibited lower f₀ variability than HCs; no differences were observed during the narrative task. In bvFTD, reduced prosodic measures correlated with atrophy in the left superior frontal gyrus and the right middle and inferior temporal gyri (p < .05, family-wise error corrected). CONCLUSIONS:Patients with bvFTD demonstrate a narrowed pitch span and reduced intonational variability, linked to disruptions in frontotemporal networks integrating emotional and semantic cues into speech. These findings highlight the relevance of prosodic alterations as a target for further research in bvFTD and assessment. (PsycInfo Database Record (c) 2026 APA, all rights reserved).

Catatonia is a well characterized psychomotor syndrome combining motor, behavioural and neurovegetative signs. Benzodiazepines are the first-choice treatment, effective in 70 % of cases. Currently, the factors associated with benzodiazepine resistance remain unknown. We aimed to develop machine learning models using clinical and neuroimaging data to predict benzodiazepine response in catatonic patients. This study examined a cohort of catatonic patients who underwent standardized clinical evaluation, 3 T brain MRI, and benzodiazepine trial. Based on clinical response, patients were classified as benzodiazepine responders or non-responders. Cortical thickness and regional brain volumes were measured. Two machine learning models (linear model and gradient boosting tree model) were developed to identify predictors of treatment response using clinical, demographic, and neuroimaging data. The cohort included 65 catatonic patients, comprising 30 benzodiazepine responders and 35 non-responders. Using clinical data alone, the linear model achieved 63% precision, 51% recall, a specificity of 61%, and 58% AUC, while the gradient boosting tree (GBT) model attained 46% precision, 60% recall, a specificity of 62% and 64% AUC. Incorporating neuroimaging data improved model performance, with the linear model achieving 66% precision, 57% recall, a specificity of 67%, and 70% AUC, and the GBT model attaining 50% precision, 50% recall, a specificity of 62% and 70% AUC. The integration of imaging data with demographic and clinical information significantly enhanced the predictive performance of the models. The duration of the catatonic syndrome, along with the presence of mitgehen (passive obedience) and immobility/stupor, and the volume of the right medial orbito-frontal cortex emerged as important factors in predicting non-response to benzodiazepines.

The characteristics of biomedical signals are not captured by conventional measures like the average amplitude of the signal. The methodologies derived from fractal geometry have been a very useful approach to study the degree of irregularity of a signal. The monofractal analysis of a signal is defined by a single power-law exponent in assuming a scale invariance in time and space. However, temporal and spatial variation in the scale-invariant structure of the biomedical signal often appears. In this case, multifractal analysis is well-suited because it is defined by a multifractal spectrum of power-law exponents. There are several approaches to the implementation of this analysis, and there are numerous ways to present these. In this chapter, we review the use of multifractal analysis for the purpose of characterizing signals in neuroimaging. After describing the tenets of multifractal analysis, we present several approaches to estimating the multifractal spectrum. Finally, we describe the applications of this spectrum on biomedical signals in the characterization of several diseases in neurosciences.

Background: Post-stroke cognitive impairment (PSCI) occurs in up to 50% of stroke survivors. Presence of pre-existing vascular brain injury, in particular the extent of white matter hyperintensities (WMH), is associated with worse cognitive outcome after stroke, but the role of WMH location in this association is unclear. Aims: We determined if WMH in strategic white matter tracts explain cognitive performance after stroke. Methods: Individual patient data from nine ischemic stroke cohorts with magnetic resonance imaging (MRI) were harmonized through the Meta VCI Map consortium. The association between WMH volumes in strategic tracts and domain-specific cognitive functioning (attention and executive functioning, information processing speed, language and verbal memory) was assessed using linear mixed models and lasso regression. We used a hypothesis-driven design, primarily addressing four white matter tracts known to be strategic in memory clinic patients: the left and right anterior thalamic radiation, forceps major, and left inferior fronto-occipital fasciculus. Results: The total study sample consisted of 1568 patients (39.9% female, mean age = 67.3 years). Total WMH volume was strongly related to cognitive performance on all four cognitive domains. WMH volume in the left anterior thalamic radiation was significantly associated with cognitive performance on attention and executive functioning and information processing speed and WMH volume in the forceps major with information processing speed. The multivariable lasso regression showed that these associations were independent of age, sex, education, and total infarct volume and had larger coefficients than total WMH volume. Conclusion: These results show tract-specific relations between WMH volume and cognitive performance after ischemic stroke, independent of total WMH volume. This implies that the concept of strategic lesions in PSCI extends beyond acute infarcts and also involves pre-existing WMH. Data access statement: The Meta VCI Map consortium is dedicated to data sharing, following our guidelines.

With the arrival of disease -modifying drugs, neurodegenerative diseases will require an accurate diagnosis for optimal treatment. Convolutional neural networks are powerful deep learning techniques that can provide great help to physicians in image analysis. The purpose of this study is to introduce and validate a 3D neural network for classification of Alzheimer's disease (AD), frontotemporal dementia (FTD) or cognitively normal (CN) subjects based on brain glucose metabolism. Retrospective [18F]-FDG-PET scans of 199 CE, 192 FTD and 200 CN subjects were collected from our local database, Alzheimer's disease and frontotemporal lobar degeneration neuroimaging initiatives. Training and test sets were created using randomization on a 90 %-10 % basis, and training of a 3D VGG16-like neural network was performed using data augmentation and cross -validation. Performance was compared to clinical interpretation by three specialists in the independent test set. Regions determining classification were identified in an occlusion experiment and Gradient -weighted Class Activation Mapping. Test set subjects were age- and sex -matched across categories. The model achieved an overall 89.8 % accuracy in predicting the class of test scans. Areas under the ROC curves were 93.3 % for AD, 95.3 % for FTD, and 99.9 % for CN. The physicians' consensus showed a 69.5 % accuracy, and there was substantial agreement between them (kappa = 0.61, 95 % CI: 0.49-0.73). To our knowledge, this is the first study to introduce a deep learning model able to discriminate AD and FTD based on [18F]-FDG PET scans, and to isolate CN subjects with excellent accuracy. These initial results are promising and hint at the potential for generalization to data from other centers.

Introduction La démence fronto-temporale (DFT) est une maladie neurodégénérative se manifestant par des troubles du comportement mais les phénotypes prédementielles des DFT sont peu décrits hormis le trouble comportemental léger (MBI). Objectifs Déterminer de manière rétrospective les caractéristiques cliniques et radiologiques des DFT débutantes, la proportion de patients répondant aux critères de MBI au stade prodromal. Méthodes Patients inclus : diagnostic de DFTvc probable, suivi d’au moins deux ans au centre mémoire de Lille entre 2010 et 2021, avec initialement un MMSE supérieur à 20 et des biomarqueurs du liquide cérébrospinal négatifs pour la maladie d’Alzheimer. Patients exclus : phénotype initial langagier ou parkinsonien atypique isolé. Constitution des groupes : phénotype initial « comportemental » ou « non-comportemental ». Comparaison : sur le plan clinique, neuropsychologique et radiologique. Résultats Soixante-dix-sept patients avec diagnostic probable de DFTvc (28 certaines) ont été inclus : 57 patients (74,0 %) avaient un phénotype initial « comportemental » et 20 (26,0 %) un phénotype « non-comportemental ». Ces derniers correspondaient à des formes amnésiques (n=9), dyséxécutives (n=4), mixtes (n=6) ou psychiatriques (n=1). Les critères de MBI étaient remplis par la majorité des patients “comportemental” mais pas par les « non-comportemental » Des différences significatives ont été mise en évidence également en imagerie. Discussion Nous avons mis en évidence dans ce travail que les DFTvc ne débutaient pas tous par des troubles du comportements mais dans un bon nombre de cas par une présentation cognitive isolée dont certaines amnésiques pures, ce qui allonge le délai diagnostique. Il pourrait être intéressant de proposer de nouveaux critères diagnostiques pour caractériser ces formes prodromales de DFTvc, actuellement mal définies. Conclusion Il existe de manière fréquente des présentations initiales non comportementales de DFTvc.