[C-11]Methionine ([C-11]MET) PET provides high sensitivity to localize corticotroph pituitary neuroendocrine tumors (PitNETs) in de novo Cushing's disease but remains inconclusive in up to 20% of cases. Current PET protocols rely on late acquisitions (20-40 min post-injection). Since other endocrine tumors present earlier peak tracer uptake, corticotroph PitNETs might also exhibit early differential amino acid uptake, compared to the normal pituitary gland. This exploratory study based on data derived from a previously registered prospective multicenter cohort study (ClinicalTrials.gov identifier: NCT03346954) included 15 patients with pathologically confirmed corticotroph PitNETs accurately localized using MRI. Kinetic analysis showed rapid early uptake in PitNETs and normal pituitary gland, followed by a significant decline over time, with consistently higher uptake in PitNETs (p < 0.05), but no statistically significant difference in temporal uptake pattern (p = 0.09). Early uptake slope and peak uptake were significantly higher in PitNETs compared to normal gland (p < 0.01), with moderate-to-high discriminative performance (area under the curve 0.78; 95%CI 0.60-0.94 and 0.86; 95%CI 0.69-0.98, respectively). Time-to-peak showed no statistically significant discriminative value. These findings suggest that kinetic parameters of [& sup1;& sup1;C]MET PET, particularly early slope and peak uptake, may provide complementary information to static PET for PitNET characterization.

Following right brain damage, some patients experience anosognosia, which is a deficit of self-awareness affecting motor, somatosensory, or visual functions on the left side of the body. While domain dissociations in anosognosia have been studied extensively, the time course of recovery remains poorly understood. This study examines the longitudinal recovery time course of anosognosia in two patients presenting hemiplegia, hemianesthesia, and visual field defect after right brain damage.Anosognosia was assessed using the Bisiach four-point scale, a self-assessment of daily life activities, and interview reports. At onset, both patients exhibited complete anosognosia for left motor, somatosensory and visual field deficit. Anosognosia for hemiplegia recovered first, and was followed by recovery of anosognosia for hemianesthesia, while anosognosia for visual field defect persisted longer.This study provides evidence for differences in the time course of recovery of anosognosia for motor and sensory deficits. These findings question underlying mechanisms of awareness deficits: domain-specific or unitary models.

[11C]Methionine ([11C]MET) PET provides high sensitivity to localize corticotroph pituitary neuroendocrine tumors (PitNETs) in de novo Cushing’s disease but remains inconclusive in up to 20% of cases. Current PET protocols rely on late acquisitions (20–40 min post-injection). Since other endocrine tumors present earlier peak tracer uptake, corticotroph PitNETs might also exhibit early differential amino acid uptake, compared to the normal pituitary gland. This exploratory study based on data derived from a previously registered prospective multicenter cohort study (ClinicalTrials.gov identifier: NCT03346954) included 15 patients with pathologically confirmed corticotroph PitNETs accurately localized using MRI. Kinetic analysis showed rapid early uptake in PitNETs and normal pituitary gland, followed by a significant decline over time, with consistently higher uptake in PitNETs (p < 0.05), but no statistically significant difference in temporal uptake pattern (p = 0.09). Early uptake slope and peak uptake were significantly higher in PitNETs compared to normal gland (p < 0.01), with moderate-to-high discriminative performance (area under the curve 0.78; 95%CI 0.60–0.94 and 0.86; 95%CI 0.69–0.98, respectively). Time-to-peak showed no statistically significant discriminative value. These findings suggest that kinetic parameters of [¹¹C]MET PET, particularly early slope and peak uptake, may provide complementary information to static PET for PitNET characterization.

Background/Objectives: Intramedullary tumors are uncommon spinal cord lesions that account for a small proportion of central nervous system neoplasms but are associated with a high risk of neurological morbidity. Accurate preoperative characterization is essential because therapeutic strategies, surgical planning, and functional prognosis depend strongly on tumor biology and growth behavior within the confined spinal cord environment. This study aims to characterize the radiological phenotype of intramedullary tumors and to identify imaging patterns that may assist in lesion characterization and diagnostic stratification. Methods: A retrospective analysis of preoperative MRI findings in patients with histopathologically confirmed intramedullary tumors was performed. Preoperative MRI examinations were systematically analyzed to describe imaging features according to tumor histology using conventional sequences (T1-weighted, T2-weighted, and contrast-enhanced imaging). Results: Distinct radiological phenotypes were observed across a wide spectrum of lesions. Glial tumors, including subependymoma, ependymoma, pilocytic astrocytoma, diffuse midline glioma H3K27M, glioblastoma, high-grade astrocytoma with piloid features, ganglioglioma, and diffuse leptomeningeal glioneural tumors, demonstrated variable combinations of cord expansion, margin definition, enhancement patterns, and tract involvement, reflecting differences between expansile and infiltrative growth. Secondary tumors such as metastases frequently exhibited aggressive imaging features, including extensive edema and intense or heterogeneous enhancement. Vascular lesions, including hemangioblastoma and cavernoma, showed characteristic vascular signatures, such as nodular enhancement with flow voids or susceptibility-related signal changes. Developmental lesions, such as epidermoid cysts, neurenteric cysts, and lipoma, displayed distinctive signal characteristics, especially on diffusion and T1, that aided differentiation from neoplastic processes. Conclusions: In conclusion, the structured radiological interpretation functions proposed herein are not only useful for diagnostic purposes, but could also be useful for risk stratification and therapeutic guidance.

PURPOSE:The purpose of this study was to determine the capabilities of dynamic susceptibility contrast (DSC)‑derived microvascular and oxygen metabolism metrics to distinguish radiation necrosis (RN) from tumor progression (TP) in irradiated brain metastases. MATERIALS AND METHODS:Fifty‑eight patients who completed cranial irradiation and underwent DSC perfusion MRI between August 2014 and August 2024 were retrospectively included. There were 31 men and 27 women, with a median age of 60.5 years (first quartile [Q1], 52.3; third quartile [Q3], 68.8). Perfusion, microvascular, and metabolic maps were generated with commercially available software. Lesion‑to‑white‑matter ratios were computed for all DSC-derived microvascular and oxygenation metrics including relative cerebral blood volume (rCBV) and oxygen extraction fraction (rOEF). Reference diagnoses were histopathology (n = 11) or multidisciplinary follow‑up (n = 47). Logistic regression analysis was used to identify metrics associated with RN versus TP, and receiver operating characteristic curve analysis was used to estimate diagnostic performance. For prognosis, overall survival was analyzed using Cox proportional hazards models. RESULTS:A total of 58 brain lesions were studied, including 34 TPs and 24 RNs. Patients with RN had longer overall survival than those with TP (median not reached vs. 22 months; P = 0.01). Among all metrics, only rCBV and rOEF differed significantly. TP showed higher median rCBV (1.8; Q1, 1.2; Q3, 2.8) than RN (1.1; Q1, 0.6; Q3, 1.9) (P = 0.02). RN exhibited greater median rOEF (1.9; Q1, 1.4; Q3, 2.1) than TP (1.5; Q1, 1.3; Q3, 1.8, P = 0.03). rCBV achieved an area under the receiver operating characteristic curve (AUC) of 0.69 (95 % confidence interval [CI]: 0.54-0.83), rOEF an AUC of 0.66 (95 % CI: 0.52-0.81), and their combination and AUC of 0.74 (95 % CI: 0.60-0.87) without significant differences (P ≥ 0.19). After adjusting for rCBV in multivariable analysis, rOEF remained significantly associated with RN (odds ratio, 0.23; 95 % CI: 0.06-0.72; P = 0.02). A greater rOEF was also associated with a longer overall survival in Cox analysis (adjusted hazard ratio, 0.72; 95 % CI: 0.55-0.95. P = 0.02). CONCLUSION:Elevated rCBV is in favor of the diagnosis of TP whereas increased rOEF is in favor of the diagnosis of RN in patients with irradiated brain metastases. Although combining metrics did not confer significant diagnostic advantages, rOEF shows an independent association with longer overall survival.

Objectif Evaluer si l’analyse IRM assiste par intelligence artificielle (IA) via Pixyl.Neuro.MS® améliore la détection des lésions et influence les décisions thérapeutiques dans le suivi de patients atteints de sclérose en plaques (SEP). Matériaux et méthodes L’étude a été menée rétrospectivement chez 83 patients atteints de SEP (âge moyen 49 ans, principalement des femmes) ayant bénéficié d’IRM successives. Pixyl.Neuro.MS® a été utilisé pour comparer les IRM FLAIR antérieures et récentes. Cet outil segmente et caractérise automatiquement les lésions selon leur évolution temporelle, facilitant l’analyse des nouvelles lésions et des lésions en expansion (EL). Pour chaque patient, un rapport radiologique assisté par IA a été généré et comparé au rapport conventionnel, et son impact potentiel sur les décisions thérapeutiques a été évalué. Résultats L'analyse assistée par IA a donné des résultats nettement supérieurs à ceux obtenus avec l'interprétation radiologique standard. De nouvelles lésions ont été identifiées chez 30,1 % des patients grâce à l'analyse assistée par IA, contre 14,6 % avec l'interprétation conventionnelle (p < 0,001). Des EL ont été détectées chez 33,7 % des patients grâce à l'évaluation assistée par IA (p < 0,001), alors qu'aucune n'a été identifiée avec l'interprétation standard. En moyenne, les radiologues aidés par l'IA ont identifié davantage de nouvelles lésions par patient (0,82 contre 0,46) et ont obtenu un nombre plus élevé de lésions vraies positives. L'intégration des résultats assistés par l'IA aux données cliniques a conduit à une modification du traitement chez 10,8 % des patients, soulignant l'impact clinique potentiel de cette approche (Figure 1). Conclusion L'IA pourrait jouer un rôle clé dans l'amélioration de la détection des nouvelles lésions et des EL chez les patients SEP. Pixyl.Neuro.MS® a renforcé l'interprétation radiologique et indirectement contribué à la révision des décisions thérapeutiques chez un sous-groupe de patients.

OBJECTIVE:Skull base tumors are difficult to approach because of their deep location and their entrapment by cranial nerves (CNs), arteries, and veins. CN trajectories can be challenging to describe by conventional MRI when they are deformed by skull base tumors. High-resolution (HR) T2-weighted MRI (T2) has been able to depict normal CN cisternal paths; however, tractography has demonstrated value in CN reconstruction when the nerves are displaced by skull base tumors. In the present study, the authors aimed to compare HR T2 to tractography in the detection of CNs displaced by skull base tumors. METHODS:From a case series of various complex skull base tumors managed between July 2015 and December 2023 in a single department, HR T2 scans were acquired, as were diffusion images with dedicated postprocessing, including distortion correction, region of interest design, probabilistic fiber tracking, and three-plane visualization. The positions of CNs displaced by skull base tumors were then compared between HR T2 and tractography. RESULTS:A total of 132 patients were included in the study. They presented with various skull base tumors: vestibular schwannomas (n = 47), cerebellopontine angle (CPA) meningiomas (n = 46), CPA epidermoid cysts (n = 12), cavernous sinus schwannomas (n = 8), cavernous sinus meningiomas (n = 3), and 16 less frequent histological types. A total of 442 CNs were identified as being displaced by skull base tumors. Of these nerves, 236 (53.4%) were identifiable using HR T2, and 358 (81.0%) were successfully reconstructed using tractography (p < 0.001, McNemar test), although not significantly for the abducens nerve, lower nerves, and hypoglossal nerve. Interestingly, the identification rate of the abducens nerve was higher on HR T2 than tractography (43.5% vs 34.8%). CONCLUSIONS:The present study revealed that tractography is more effective in predicting the position of most nerves displaced by skull base tumors, whereas HR T2 can identify the smallest CNs such as the abducens nerve.

BACKGROUND:Frailty is a prevalent condition among older adults with neurocognitive disorders. OBJECTIVES:To ascertain whether frailty contributes to the severity of cognitive impairment and neuropsychiatric symptoms, and its association with cerebral pathology measured in vivo by fluid and imaging biomarkers. DESIGN:We conducted cross-sectional and longitudinal analyses based on CLEM Study, a multicentre memory-clinic cohort that recruited participants between 2014 and 2018. SETTING:CLEM Study occurred in eight memory centres in France (Lyon, Paris, Strasbourg, Poitiers, Tours, Grenoble) and Monaco. PARTICIPANTS:A total of 168 participants (mean age 80.5 ± 4.8 years) with mild to moderate dementia due to at least one aetiological diagnosis between Alzheimer's disease, dementia with Lewy bodies or vascular dementia were included in the study. MEASUREMENTS:The participants were evaluated at baseline and followed up for two years. The concept of frailty was operationalised using a 45-item Frailty Index. Cognition was assessed using the ADAS-cog scale, while neuropsychiatric symptoms were evaluated with the Neuropsychiatric Inventory. The cerebral pathological score, a proxy for brain pathologies, was a composite score based on the presence of several in vivo biomarkers: presynaptic dopaminergic denervation on 123I-FP-CIT SPECT (DaTscan®), vascular lesions on MRI, elevated blood-based pTau181, neurofilaments light-chain or glial fibrillary acid protein. Linear and mixed regression analyses were conducted to model the relationships between cognitive or neuropsychiatric symptoms, frailty and cerebral pathologic score, adjusted for age, sex and education. RESULTS:The findings indicate an impact of both frailty (β = 0.28, 95 % CI [0.14-0.43], p < 0.001) and cerebral pathological score (β = 0.30, 95 % CI [0.13-0.47], p = 0.002) on cognitive impairment. However, only frailty was associated with neuropsychiatric symptoms (β = 0.28, 95 % CI [0.14-0.43], p < 0.001), particularly with apathy (β = 0.40, 95 % CI [0.26-0.53], p < 0.001). We found an association between cerebral pathological score and longitudinal cognitive decline (β = 0.36, 95 % CI [0.19-0.53], p < 0.001) in exploratory analyses with available longitudinal data at 24 months (n = 74). CONCLUSIONS:Neurocognitive disorders are complex entities, where cognitive and neuropsychiatric symptoms are not fully influenced by the same factors. When cognitive symptoms seem more driven by cerebral pathology than frailty, neuropsychiatric symptoms appear to be more influenced by general state of frailty. Measuring and treating frailty might be a key factor in dealing with neuropsychiatric symptoms and their consequences.

The assessment of new Multiple Sclerosis (MS) lesions is a time-consuming and difficult task that may lead to an underestimation of patient disease activity. Efficient methods to detect new T2/FLAIR lesions are therefore critical to assist clinicians in this process. In this context, we proposed and organized in 2021 the MSSeg2 challenge aiming at comparing methods segmenting new MS lesions. For this purpose, we built a high-quality dataset of 100 pairs of FLAIR MRI with precise delineation of new MS lesions with a size superior to 3 mm[Formula: see text]. From these 100 pairs of MS patient images from various scanners and French clinical centers, 40 were shared to the challengers before they submitted their methods. 30 methods from 24 international teams were submitted and evaluated on the FLI-IAM dedicated platform on the 60 remaining images. Overall, we observed that even at lesion scale, expert annotations were variable (40% of lesions were annotated by two or fewer experts out of the four). The best expert exhibited a mean F1 score of 0.679 (SD = 0.345) while the best method exhibited a mean F1 score of 0.698 (SD = 0.295) on the 35 patients with new lesions. Moreover, we did not observe evidence of differences between the top-ranked methods and the best expert performances as evaluated by the F1 score (9 methods exhibited no clear evidence against no difference of mean F1). Similarly, we did not observe evidence of difference in performances between the top-ranked methods and the best expert in classifying patients into a 3 categories clinically relevant scale (0 lesion, 1 or 2 lesions and >2 lesions; 21 methods exhibited no clear evidence against no difference of performances with a best classification accuracy of 85 % for both the best expert and the best method). While results from current automated methods still remain perfectible, our results highlight their potential usefulness in detecting new FLAIR MS lesions.

White matter hyperintensities (WMHs) are highly prevalent in Alzheimer's disease (AD) and arise from interacting vascular pathologies (including hypertensive small vessel disease and cerebral amyloid angiopathy) alongside inflammatory and neurodegenerative processes. In the era of anti-amyloid monoclonal antibodies, this heterogeneity is increasingly relevant for both treatment efficacy and safety. WMHs may signal mixed AD-vascular pathology that dilutes the cognitive benefit of amyloid-targeting therapies and may also index vulnerability of the neurovascular unit that predisposes to amyloid-related imaging abnormalities (ARIAs), although direct evidence remains limited. In this perspective, we synthesize current knowledge on the origins of WMHs in AD, review advanced magnetic resonance imaging and biomarker approaches that aim to refine lesion characterization in vivo, and discuss how WMHs should be interpreted in memory clinic practice when considering anti-amyloid therapies. We conclude with a research roadmap to integrate WMH phenotyping into precision risk-benefit assessment and ARIA prediction .

BACKGROUND:Magnetic resonance imaging (MRI) is central to the diagnosis and follow-up of multiple sclerosis (MS). Since the 2020 OFSEP recommendations, the 2024 McDonald criteria have introduced major changes, including recognition of the optic nerve as a fifth anatomical topography and integration of two specificity biomarkers - the central vein sign (CVS) and the paramagnetic rim lesion (PRL). OBJECTIVE:To update the French OFSEP MRI recommendations for MS while remaining feasible on every 1.5 T and 3 T scanner in France. METHODS:A multidisciplinary OFSEP-SFNR working group conducted a modified Delphi-like consensus process. RECOMMENDATIONS:The core brain protocol comprises isotropic fat-suppressed 3D FLAIR (1 mm³), DWI with ADC maps, and an optimised 3D susceptibility-sensitive sequence (3D T2* GRE or 3D SWI with filtered phase) for CVS and PRL. A contrast-enhanced 3D spin-echo T1 sequence and a dedicated orbital protocol for suspected optic neuritis are added at diagnosis. The spinal cord protocol includes at least two complementary sequences, with a high-contrast heavily T1-weighted sequence when available. Follow-up uses the same core protocol with strictly reproducible 3D FLAIR; gadolinium is restricted to targeted indications. CONCLUSION:This vendor-neutral framework supports earlier, more specific MS diagnosis and reduced gadolinium use across French centres.

White matter hyperintensities (WMH) are common in older adults and are associated with cognitive disorders. They typically arise from small vessel disease, leading to demyelination and axonal loss. WMH are thus considered markers of cerebrovascular changes. However, other pathophysiological processes can lead to WMH, particularly in Alzheimer’s disease (AD). Understanding the diverse origins of WMH could enhance the diagnosis and treatment of AD patients. We hypothesize that multimodal neuroimaging could help understand the heterogeneity of WMH and pinpoint their specific origin. We included 142 older adults from the community and memory clinic (with an emphasis on patients within the Alzheimer’s continuum), and tested if multimodal neuroimaging signal within regional WMH (including T1w, T2w, 18F-florbetapir [AV45] and 18F-fluorodeoxyglucose [FDG] PET), is associated with amyloid load and cognition. We showed that intra-WMH T1w and T2w signal in the parietal and frontal lobes were linked to amyloid status; intra-WMH T2w signal in all regions negatively correlated with amyloid load, while intra-WMH T1w signals in the parietal lobe positively correlated with amyloid load; finally, intra-WMH T1w signal negatively correlated with cognition while T2w and marginally AV45 signals positively correlated with cognition. This study demonstrates the potential of multimodal neuroimaging to unravel the heterogeneity of WMH, which could enhance their interpretation and improve clinical decision-making.

BackgroundMultiple sclerosis is an inflammatory demyelinating disease of the CNS. Annual MRI exams are crucial for disease monitoring. Interpreting high T2/FLAIR lesion loads can be laborious. AI aids in lesion detection, and choosing between different solutions can be challenging.AimThis study compares two distinct software, Pixyl.Neuro.MS® and Jazz®, to assess their performance in T2/FLAIR lesion detection between two-time points.MethodsRetrospective analysis included follow-up MRIs from 35 MS patients. Pixyl.Neuro.MS® automatically segments and classifies lesions. Jazz® automates the reading process and image display. Two readers (15 and 4 years of experience) conducted radiological analysis, followed by AI-assisted readings. A number of new lesions (NL) and reading times were recorded, with ground truth (GT) established by consensus. AI-detected lesions were classified as true (TP) and false positives (FP). Statistical analysis used SPSS (p < 0.05).ResultsPixyl.Neuro.MS® readings averaged 2 min 46 s ± 1 min 4 s while using Jazz® 3 min 33 s ± 2 min 24 s. Over 50% of the population had a high lesion load (>20 lesions). Both software significantly improved NL detection (p < 0.01 for both), revealing them in more patients than standard readings. Standard reports found 8 NL in 2 patients, while AI-assisted readings detected at least 17 TP in 7 patients and rejected 61 FP lesions. GT detected 21 lesions in 19 patients.ConclusionBoth AI software have been found to enhance NL detection in MS patients, outperforming standard methods. These tools offer crucial advantages for accurate disease monitoring.

OBJECTIVE:Skull base tumor surgery remains challenging because these tumors are deeply seated and trapped within numerous cranial nerves and vessels. Accurate histopathological analysis of skull base tumors will strongly impact their further management. Yet, currently there is no noninvasive validated method to confirm their diagnosis. In a recent study of MRI tractography, the authors used fiber orientation distribution (FOD) and noted that this diffusion model formed a pattern that could correspond to the histopathological type of skull base tumors. The aim of this study was to propose a new imaging method for skull base tumors that can detect the diagnosis according to the FOD pattern. METHODS:From an 81-case series of skull base tumors, MRI diffusion images were investigated by 3 independent observers. Diffusion patterns were classified as centrifugal, wrapped, and chaotic, corresponding to meningiomas, schwannomas, and epidermoid cysts, respectively. RESULTS:The overall identification rate was 80.7%, with an excellent concordance between the 3 observers (Fleiss's κ coefficient = 0.765, p < 0.0001). The identification rate increased along with the observers' anatomoradiological expertise (72.8%, 77.8%, and 91.4%, for observer 1 [low expertise], 2 [average expertise], and 3 [extensive expertise], respectively) and was higher for schwannomas (88.6%), than meningiomas (72.6%) and epidermoid cysts (66.7%). CONCLUSIONS:This new imaging tool could assist in complex skull base tumor identification and characterization, as well as surgical management.

OBJECTIVE:To develop a multidisciplinary French framework addressing neurosurgical management in the initial phase of traumatic brain injury (TBI) in adults and children. DESIGN:A panel of 29 experts was formed at the request of the French Society of Neurosurgery (SFNC), with the participation of the French Society of Pediatric Neurosurgery (SFNCP), French Society of Private-Practice Neurosurgeons (SFNCL), French-Speaking Neurocritical Care and Neuro-Anesthesiology Society (ANARLF), French Society of Anesthesia, Critical Care and Perioperative Medicine (SFAR), French-Speaking Pediatric Emergency and Intensive Care Group (GFRUP), French Society of Neuroradiology (SFNR), French-Speaking Infectious Diseases Society (SPILF), and the French Society of Physical Medicine and Rehabilitation (SOFMER). METHODS:Questions were formulated using the PICO (Patients, Intervention, Comparison, Outcome) format, grouped into 7 categories: 1. Factors of poor prognosis, 2. Extradural hematoma, 3. Acute subdural hematoma, 4. Skull-base fracture and dural tear, 5. Penetrating traumatic brain injury, 6. Post-traumatic cerebrospinal fluid disorder, and 7. Pediatric specificities. RESULTS:Synthesis by the experts and application of the GRADE® method resulted in the formulation of 45 recommendations. Strong consensus was reached for all recommendations at the first round of rating, CONCLUSION: There was a strong consensus among the experts on important interdisciplinary recommendations to improve the neurosurgical management of patients with TBI.

Background and ObjectivesAbnormal brain MRI is associated with poor outcomes in anti-N-methyl-d-aspartate receptor encephalitis (NMDARE). We aimed to characterize the lesions on brain MRI in NMDARE and to assess the clinical and prognostic associations. MethodsThis retrospective cohort study included patients with NMDARE identified at the French Reference Center for Autoimmune Encephalitis, with at least a one-year follow-up, and with available brain MRI results. In case of brain extralimbic lesion, the image files were reviewed when available. Clinical data were collected from medical records. Multivariable logistic regression analysis was used to study the outcomes at 2-year follow-up; recovery was defined as modified Rankin Scale score <= 1. ResultsAmong the 255 patients included, 37 (14.5%) had limbic hyperintensities and 41 (16.1%) had extralimbic lesions that included multiple sclerosis (MS)-like lesions (14/41, 34.1%); extensive lesions (5/41, 12.2%); and poorly demarcated fluffy lesions, either multifocal (10/41, 24.4%) or involving the cerebral cortex or cerebellum (6/41 each, 14.6%). Extralimbic lesions coexisting with limbic lesions (19/41 patients, 46.3%) were mostly fluffy lesions (11/19, 57.9%). Ten patients had overlapping demyelinating syndromes: 4 with MS, 4 with myelin oligodendrocyte glycoprotein-associated disorder, and 2 with neuromyelitis optica spectrum disorder; all had MS-like (7/10 patients) or extensive (3/10 patients) lesions, and none had fluffy lesions. Extralimbic lesions were associated with symptoms nontypical for NMDARE (23/41, 56.1%, p < 0.001), especially cerebellar ataxia (17/41, 41.5%) and motor impairment (12/41, 29.3%). At 2 years, patients with MS-like or extensive lesions had a lower recovery rate (5/12, 41.7%, and 1/4, 25%, respectively) compared with the patients without extralimbic lesions (124/162, 76.5%; p = 0.014 and p = 0.047, respectively). In multivariable analysis, MS-like lesions, but not hippocampal nor fluffy lesions, were associated with absence of recovery at 2 years (adjusted OR 0.1, 95% CI 0.03-0.42, p = 0.002; extensive lesions [n = 4] not included in the analysis). DiscussionBrain MRI lesions in NMDARE include limbic hyperintensities and 3 patterns of extralimbic lesions, which are associated with nontypical NMDARE symptoms. Moreover, MS-like and extensive lesions, but not fluffy nor hippocampal lesions, are associated with overlapping demyelinating syndromes and poor clinical outcomes at 2 years. These findings can have practical implications on the monitoring of patients with NMDARE.

Introduction L'interprétation des IRM de suivi de patients atteints de Sclérose En Plaques (SEP) est à la fois difficile et chronophage. L'utilisation de l'IA, que ce soit pour faciliter la détection des lésions démyélinisantes typiques en IRM, ou automatiser le processus de lecture, est prometteuse. Cette étude se penche sur deux logiciels, Pixyl.Neuro.MS® et Jazz®, afin d'évaluer leur performance dans l'aide à la détection longitudinale des lésions T2/FLAIR chez les patients atteints de SEP présentant une forte charge lésionnelle. Matériels et méthodes Les IRM de suivi de 35 patients atteints de SEP ont été rétrospectivement examinés. Pixyl.Neuro.MS® (v1.8.7) segmente et effectue la catégorisation longitudinale des lésions (nouvelles versus préexistantes), tandis que Jazz® automatise le processus de lecture et l'affichage des images. Les temps d'interprétation avec les logiciels l'IA ont été mesurés. La vérité terrain sur les nouvelles lésions (NL) est définie par le consensus, entre les deux radiologues, des lectures assistées par l'IA. Les nouvelles lésions détectées à l'aide de chaque logiciel ont été comparées à la vérité terrain, afin de les catégoriser en vraies positives (VP) ou fausses positives (FP). Résultats Les temps nécessaires pour la seconde lecture avec les outils d'IA sont reportés sur la figure 1. Les deux logiciels ont permis aux radiologues d'identifier correctement un nombre plus élevé de nouvelles lésions (p<0,01 pour les deux logiciels) par rapport aux lectures radiologiques non assistées par l'IA (voir fig. 2) réalisées en routine clinique, malgré la présence de faux positifs. Conclusion Les deux logiciels d'IA améliorent de manière significative la détection des nouvelles lésions chez les patients atteints de SEP par rapport aux méthodes standards. Ces outils offrent ainsi un avantage crucial dans le suivi précis de l'évolution de la maladie.

PURPOSE:Neuroradiological findings associated with neurological presentations in acute SARS-CoV-2 infection are very heterogeneous. We aimed to develop a standardized framework for describing MR neuroimaging patterns in Covid-19, to test this in an international multicentre study and to determine the prevalence of observed MRI patterns and their association with clinical presentation and outcome. METHODS:An international expert consortium developed a framework for assessment of brain MRI patterns in Covid-19 based on published literature and professional experience. We performed a retrospective analysis of the proposed framework, involving brain MRI scans from 458 Covid-19 patients with neurological symptoms, including data from 1 February to 31 May 2020. Two readers at 25 centres across five countries assessed the local MRI studies regarding the presence of one or more predefined MRI patterns. Imaging and clinical data were analysed using Bayesian statistics. RESULTS:Of 458 patients, 58.5% had an abnormal MRI. Overall, 94% of all imaging pathologies seen were captured by our proposed classification. Ischemic strokes were the most frequent pattern overall (25.6%), followed by microhaemorrhages (15.9%). Ischemic infarct patterns were more frequent in non-ICU patients, while the haemorrhagic patterns were more frequent in ICU patients. White matter lesions (10.9%) were more frequent than grey matter lesions (8.1%), and leptomeningeal contrast enhancement was present in 8.3% of patients. Patient outcome was not associated with any MRI patterns. CONCLUSION:Our proposed classification of specific MRI patterns in Covid-19, covered 94% of observed abnormalities, while patient outcome, death or home discharge, was not associated with any MRI patterns.

Deep networks interpretability is fundamental in critical domains like medicine: using easily explainable networks with decisions based on radiological signs and not on spurious confounders would reassure the clinicians. Confidence is reinforced by the integration of intrinsic properties and characteristics of monotonic networks could be used to design such intrinsically explainable networks. As they are considered as too constrained and difficult to train, they are often very shallow and rarely used for image applications. In this work, we propose a procedure to transform any architecture into a trainable monotonic network, identifying the critical importance of weights initialization, and highlight the interest of such networks for explicability and interpretability. By constraining the features and gradients of a healthy vs pathological images classifier, we show, using counterfactual examples, that the network decision is more based on radiological signs of the pathology and outperform state-of-the-art weakly supervised anomaly detection methods.

Robust and continuous in vivo differentiation of spinal tracts along the brain–spinal cord axis is limited. We stitched brain and spinal cord diffusion tensor imaging (DTI) to create continuity between the brain and cervical spinal cord, enabling tractography along the central nervous system, producing an atlas of the spinal cord white matter. This prospective pilot study included four healthy subjects. Brain and cervical spinal cord 3-T DTI acquisitions were performed. Distortions were corrected using the Functional magnetic resonance imaging of the brain Software Library (FSL) software package. A semiautomatic stitching process was achieved using cross-correlation. Once the highest correlation peak was identified, rigid registration allowed accurate image alignment and fusion. Regions of interest were drawn in the brainstem according to atlas-guided projection tracts. Fiber tracking was performed using a deterministic approach with Diffusion Spectrum Imaging (DSI) Studio. The median fiber length from stitched-image tractography (192 mm) was significantly greater than that from both the brain (111.5 mm) and spinal cord (115 mm) fields of view. The white matter fiber atlas described: the corticospinal tract in the medial part of the lateral funiculus; the rubrospinal tract in the lateral funiculus, overlapped with the corticospinal tract; the gracilis and cuneatus tracts in the dorsal columns; the spinothalamic tract in the ventrolateral part of the spinal cord, around the ventral horn; and the spinocerebellar tracts overlapping them, in the lateral funiculus. Stitching brain and spinal cord DTI fields of view provided an in vivo spinal cord white matter atlas in humans. This study provides a detailed and individualized mapping of spinal tracts, serving as a potential tool for neurosurgical planning, particularly in procedures involving intramedullary tumors. It also may enhance the accuracy of prognostic assessments in patients with spinal cord injury, multiple sclerosis, or degenerative myelopathy.