Data-driven normative values based on generative manifold learning for quantitative MRI.

In medicine, abnormalities in quantitative metrics such as the volume reduction of one brain region of an individual versus a control group are often provided as deviations from so-called normal values. These normative reference values are traditionally calculated based on the quantitative values from a control group, which can be adjusted for relevant clinical co-variables, such as age or sex. However, these average normative values do not take into account the globality of the available quantitative information. For example, quantitative analysis of T1-weighted magnetic resonance images based on anatomical structure segmentation frequently includes over 100 cerebral structures in the quantitative reports, and these tend to be analyzed separately. In this study, we propose a global approach to personalized normative values for each brain structure using an unsupervised Artificial Intelligence technique known as generative manifold learning. We test the potential benefit of these personalized normative values in comparison with the more traditional average normative values on a population of patients with drug-resistant epilepsy operated for focal cortical dysplasia, as well as on a supplementary healthy group and on patients with Alzheimer's disease.

Improving rehabilitation of deaf patients by advanced imaging before cochlear implantation.

INTRODUCTION: Cochlear implants have advanced the management of severe to profound deafness. However, there is a strong disparity in hearing performance after implantation from one patient to another. Moreover, there are several advanced kinds of imaging assessment before cochlear implantation. Microstructural white fiber degeneration can be studied with Diffusion weighted MRI (DWI) or tractography of the central auditory pathways. Functional MRI (fMRI) allows us to evaluate brain function, and CT or MRI segmentation to better detect inner ear anomalies. OBJECTIVE: This literature review aims to evaluate how helpful pre-implantation anatomic imaging can be to predict hearing rehabilitation outcomes in deaf patients. These techniques include DWI and fMRI of the central auditory pathways, and automated labyrinth segmentation by CT scan, cone beam CT and MRI. DESIGN: This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Studies were selected by searching in PubMed and by checking the reference lists of relevant articles. Inclusion criteria were adults over 18, with unilateral or bilateral hearing loss, who had DWI acquisition or fMRI or CT/ Cone Beam CT/ MRI image segmentation. RESULTS: After reviewing 172 articles, we finally included 51. Studies on DWI showed changes in the central auditory pathways affecting the white matter, extending to the primary and non-primary auditory cortices, even in sudden and mild hearing impairment. Hearing loss patients show a reorganization of brain activity in various areas, such as the auditory and visual cortices, as well as regions involved in language and emotions, according to fMRI studies. Deep Learning's automatic segmentation produces the best CT segmentation in just a few seconds. MRI segmentation is mainly used to evaluate fluid space of the inner ear and determine the presence of an endolymphatic hydrops. CONCLUSION: Before cochlear implantation, a DWI with tractography can evaluate the central auditory pathways up to the primary and non-primary auditory cortices. This data is then used to generate predictions on the auditory rehabilitation of patients. A CT segmentation with systematic 3D reconstruction allow a better evaluation of cochlear malformations and predictable difficulties during surgery.

Prediction of therapeutic intensity level from automatic multiclass segmentation of traumatic brain injury lesions on CT-scans.

The prediction of the therapeutic intensity level (TIL) for severe traumatic brain injury (TBI) patients at the early phase of intensive care unit (ICU) remains challenging. Computed tomography images are still manually quantified and then underexploited. In this study, we develop an artificial intelligence-based tool to segment brain lesions on admission CT-scan and predict TIL within the first week in the ICU. A cohort of 29 head injured patients (87 CT-scans; Dataset1) was used to localize (using a structural atlas), segment (manually or automatically with or without transfer learning) 4 or 7 types of lesions and use these metrics to train classifiers, evaluated with AUC on a nested cross-validation, to predict requirements for TIL sum of 11 points or more during the 8 first days in ICU. The validation of the performances of both segmentation and classification tasks was done with Dice and accuracy scores on a sub-dataset of Dataset1 (internal validation) and an external dataset of 12 TBI patients (12 CT-scans; Dataset2). Automatic 4-class segmentation (without transfer learning) was not able to correctly predict the apparition of a day of extreme TIL (AUC = 60 ± 23%). In contrast, manual quantification of volumes of 7 lesions and their spatial location provided a significantly better prediction power (AUC = 89 ± 17%). Transfer learning significantly improved the automatic 4-class segmentation (DICE scores 0.63 vs 0.34) and trained more efficiently a 7-class convolutional neural network (DICE = 0.64). Both validations showed that segmentations based on transfer learning were able to predict extreme TIL with better or equivalent accuracy (83%) as those made with manual segmentations. Our automatic characterization (volume, type and spatial location) of initial brain lesions observed on CT-scan, publicly available on a dedicated computing platform, could predict requirements for high TIL during the first 8 days after severe TBI. Transfer learning strategies may improve the accuracy of CNN-based segmentation models.Trial registrations Radiomic-TBI cohort; NCT04058379, first posted: 15 august 2019; Radioxy-TC cohort; Health Data Hub index F20220207212747, first posted: 7 February 2022.

Individualised profiling of white matter organisation in moderate-to-severe traumatic brain injury patients.

BACKGROUND AND PURPOSE: Approximately 65% of moderate-to-severe traumatic brain injury (m-sTBI) patients present with poor long-term behavioural outcomes, which can significantly impair activities of daily living. Numerous diffusion-weighted MRI studies have linked these poor outcomes to decreased white matter integrity of several commissural tracts, association fibres and projection fibres in the brain. However, most studies have focused on group-based analyses, which are unable to deal with the substantial between-patient heterogeneity in m-sTBI. As a result, there is increasing interest and need in conducting individualised neuroimaging analyses. MATERIALS AND METHODS: Here, we generated a detailed subject-specific characterisation of microstructural organisation of white matter tracts in 5 chronic patients with m-sTBI (29 - 49y, 2 females), presented as a proof-of-concept. We developed an imaging analysis framework using fixel-based analysis and TractLearn to determine whether the values of fibre density of white matter tracts at the individual patient level deviate from the healthy control group (n = 12, 8F, Mage = 35.7y, age range 25 - 64y). RESULTS: Our individualised analysis revealed unique white matter profiles, confirming the heterogenous nature of m-sTBI and the need of individualised profiles to properly characterise the extent of injury. Future studies incorporating clinical data, as well as utilising larger reference samples and examining the test-retest reliability of the fixel-wise metrics are warranted. CONCLUSIONS: Individualised profiles may assist clinicians in tracking recovery and planning personalised training programs for chronic m-sTBI patients, which is necessary to achieve optimal behavioural outcomes and improved quality of life.

Optimized 3D-FLAIR sequences to shorten the delay between intravenous administration of gadolinium and MRI acquisition in patients with Menière's disease.

OBJECTIVES: The aim of this study was to shorten the 4-h delay between the intravenous administration of gadolinium and MRI acquisition for hydrops evaluation using an optimized 3D-FLAIR sequence in patients with Menière's disease. METHODS: This was a single-center prospective study including 29 patients (58 ears), recruited between November 2020 and February 2021. All patients underwent a 3-T MRI with an optimized 3D-FLAIR sequence without contrast then at 1 h, 2 h, and 4 h after intravenous administration of gadobutrol. The signal intensity ratio was quantitatively assessed with the region of interest method. We also evaluated the volume of endolymphatic structures (saccule, utricle) then the presence of endolymphatic hydrops and blood-labyrinthine barrier impairment at each acquisition time. RESULTS: For all ears, the signal intensity ratio was significantly non-inferior at 2 h compared to 4 h, with a mean geometric signal intensity ratio at 0.83 (95% CI: 0.76 to 0.90, one-sided p < .001 for non-inferiority at -30% margin). Mean volume equivalence of saccule and utricle between 2 and 4 h was proven at a ± 0.20 standardized deviation equivalence margin. Intra-rater agreements (Cohen's kappa) were all greater than 0.90 for all endolymphatic hydrops location and blood-labyrinthine-barrier impairment between the 2- and 4-h assessments. CONCLUSIONS: We demonstrated that using an optimized 3D-FLAIR sequence we could shorten the acquisition from 4 to 2 h with a high reliability for the diagnosis of endolymphatic hydrops and blood-labyrinthine-barrier impairment. CLINICAL TRIAL REGISTRATION: Clinical trial no: 38RC15.173 KEY POINTS: • Magnetic resonance imaging with delayed 3D-FLAIR sequences allows the diagnosis of endolymphatic hydrops in patients with definite Menière's disease. • An optimized 3D-FLAIR sequence with a long TR of 16000 ms and a constant flip angle allows for reducing the delay between intravenous injection of gadobutrol and MRI acquisition from 4 to 2 h to diagnose endolymphatic hydrops. • Reducing this delay between intravenous injection and MRI acquisition could have implications for clinical practice for both patients and imaging departments.

MRI characteristics of intralabyrinthine schwannoma on post-contrast 4 h-delayed 3D-FLAIR imaging.

PURPOSE: The purpose of this study was to describe the MRI characteristics of intralabyrinthine schwannoma (ILS) on post contrast three-dimensional (3D) fluid-attenuation-inversion-recovery (FLAIR) images obtained four hours after intravenous administration of a gadolinium-based contrast agent (4h-3D-FLAIR). MATERIALS AND METHODS: This IRB-approved retrospective multi-center study included patients presenting with typical ILS from January 2016 to October 2020. All medical charts were systematically collected. All MRI examinations, including 4h-3D-FLAIR images, were reviewed by two board-certified neuroradiologists. Main outcome measures were location, signal intensity and associated anomalies of ILS. RESULTS: Twenty-seven out of 8730 patients (0.31%) referred for the investigation of a cochleovestibular disorder had a final diagnosis of ILS. There were 13 men and 14 women with a mean age of 52 ± 17 (SD) years (age range: 20-86 years). The most common clinical presentation was unilateral progressive sensorineural hearing loss (16/27; 59%). All ILS were unilateral and 15 (15/27; 55%) were intracochlear. All ILS presented as a hypointense filling defect within the labyrinth on T2-weighted images that enhanced on post-contrast T1-weighted images. On 4h-3D-FLAIR images, all ILS presented as a hypointense filling defect, associated with diffuse perilymphatic hyperintensity. Two patients (2/27; 7%) presented with ipsilateral endolymphatic hydrops. CONCLUSION: ILS displays consistent features on post-contrast 4h-3D-FLAIR images. ILS should not be confused with endolymphatic hydrops and requires a systematic analysis of the corresponding T2-weighted images.

Increased signal intensity with delayed post contrast 3D-FLAIR MRI sequence using constant flip angle and long repetition time for inner ear evaluation.

PURPOSE: The purpose of this study was to compare the degree of perilymphatic enhancement between 4 hour post-contrast constant flip angle three-dimensional fluid attenuated inversion recovery (3D-FLAIR) images obtained with short repetition time (TR) and those obtained with long TR. MATERIALS AND METHODS: This single-center, prospective study included patients who underwent MRI of the inner ear with heavily T2-weighted sequence, 3D-FLAIR sequence with a "short" TR of 10,000 ms (s3D-FLAIR) and with a "long" TR of 16,000 ms (l3D-FLAIR). Signal intensity ratio (SIR) and contrast-to-noise ratio (CNR) obtained with s3D-FLAIR and l3D-FLAIR were quantitatively assessed using region of interest (ROI) method and compared. The morphology of the endolymphatic space on both sequences was also evaluated. RESULTS: From March 2020 to July 2020, 20 consecutive patients were enrolled (9 women and 11 men; mean age, 52.1 ± 14.5 [SD] years; age range: 29-75 years). On l3D-FLAIR images, mean SIR (21.1 ± 8.8 [SD]; range: 7.6-46.1) was significantly greater than that on s3D-FLAIR images (15.7 ± 6.7 [SD]; range: 5.9-33.4) (P < 0.01). On l3D-FLAIR images, mean CNR (17 ± 8.5 [SD]; range: 2-40) was significantly greater than that on s3D-FLAIR images (12 ± 6.3 [SD]; range: 3.2-29.8) (P < 0.01). Kappa value for inter-rater agreement for endolymphatic hydrops, vestibular atelectasis and perilymphatic fistula were 0.93 (95% CI: 0.74-1), 1 (95% CI: 0.85-1) and 1 (95% CI: 0.85-1) respectively. CONCLUSION: This study demonstrates that the sensitivity of 3D-FLAIR sequences to low concentration gadolinium in the perilymphatic space is improved by elongation of the TR, with SIR and CNR increased by +34.4% and +41.3% respectively.

Benign recurrent vestibulopathy: MRI and vestibular tests results in a series of 128 cases.

PURPOSE: Benign recurrent vestibulopathy is a recent entity, close to probable Meniere's disease and vestibular-migraine. So far, no study has systematically investigated the presence of endolymphatic hydrops of the lateral semicircular canal in benign recurrent vestibulopathy using magnetic resonance imaging. The aim of this study was to determine magnetic resonance imaging data and vestibular test results in patients with benign recurrent vestibulopathy. METHODS: 128 patients with benign recurrent vestibulopathy included since 2010 were retrospectively analyzed. Patients had magnetic resonance imaging with a delayed acquisition, audiogram, head shaking test, caloric-test, skull-vibration-induced-nystagmus-test, video-head- impulse-test, and vestibular evoked myogenic potential. Endolymphatic hydrops presence was classified into four categories: cochlear, saccular, lateral semicircular canal and association with saccule. RESULTS: In benign reccurent vestibulopathy, 23% of cases showed an endolymphatic hydrops on magnetic resonance imaging, more frequently located in the lateral semicircular canal (59%) and related to the disease duration. The most often impaired tests were caloric-test (49%) with fluctuations of hypofunction in 67% and skull-vibration-induced-nystagmus-test (61%). No correlation between the caloric-test and the presence and location of the endolymphatic hydrops was observed. CONCLUSION: In our series of benign reccurent vestibulopathy, a rare endolymphatic hydrops was most often observed for the lateral semicircular canal and correlated with the seniority of the pathology. Hydrops identified at the magnetic resonance imaging was not correlated with the caloric-test results. Skull-vibration-induced-nystagmus-test and caloric-test were the most often modified vestibular tests.

Tractography dissection variability: What happens when 42 groups dissect 14 white matter bundles on the same dataset?

White matter bundle segmentation using diffusion MRI fiber tractography has become the method of choice to identify white matter fiber pathways in vivo in human brains. However, like other analyses of complex data, there is considerable variability in segmentation protocols and techniques. This can result in different reconstructions of the same intended white matter pathways, which directly affects tractography results, quantification, and interpretation. In this study, we aim to evaluate and quantify the variability that arises from different protocols for bundle segmentation. Through an open call to users of fiber tractography, including anatomists, clinicians, and algorithm developers, 42 independent teams were given processed sets of human whole-brain streamlines and asked to segment 14 white matter fascicles on six subjects. In total, we received 57 different bundle segmentation protocols, which enabled detailed volume-based and streamline-based analyses of agreement and disagreement among protocols for each fiber pathway. Results show that even when given the exact same sets of underlying streamlines, the variability across protocols for bundle segmentation is greater than all other sources of variability in the virtual dissection process, including variability within protocols and variability across subjects. In order to foster the use of tractography bundle dissection in routine clinical settings, and as a fundamental analytical tool, future endeavors must aim to resolve and reduce this heterogeneity. Although external validation is needed to verify the anatomical accuracy of bundle dissections, reducing heterogeneity is a step towards reproducible research and may be achieved through the use of standard nomenclature and definitions of white matter bundles and well-chosen constraints and decisions in the dissection process.

Contribution of CT-Scan Analysis by Artificial Intelligence to the Clinical Care of TBI Patients.

The gold standard to diagnose intracerebral lesions after traumatic brain injury (TBI) is computed tomography (CT) scan, and due to its accessibility and improved quality of images, the global burden of CT scan for TBI patients is increasing. The recent developments of automated determination of traumatic brain lesions and medical-decision process using artificial intelligence (AI) represent opportunities to help clinicians in screening more patients, identifying the nature and volume of lesions and estimating the patient outcome. This short review will summarize what is ongoing with the use of AI and CT scan for patients with TBI.

Tracking white-matter brain modifications in chronic non-bothersome acoustic trauma tinnitus.

Subjective tinnitus is a symptom characterized by the perception of sound with no external acoustic source, most often accompanied by co-morbidities. To date, the specific role of white matter abnormalities related to tinnitus reaches no consensus in the literature. The goal of this study was to explore the structural connectivity related to tinnitus percept per se, thus focusing on a specific population presenting chronic non-bothersome tinnitus of similar etiology (noise induced) without co-morbidities. We acquired diffusion-weighted images with high angular resolution in a homogeneous group of mildly impacted tinnitus participants (n = 19) and their matched controls (n = 19). We focused the study on two subsets of fiber bundles of interest: on one hand, we extracted the acoustic radiation and further included any intersecting fiber bundles; on the other hand, we explored the tracts related to the limbic system. We modeled the diffusion signal using constrained spherical deconvolution. We conducted a deep-learning based tractography segmentation and mapped Apparent Fiber Density (AFD) on the bundles of interest. C, as well as Fractional Anisotropy (FA) and FOD peak amplitude for comparison. Between group statistical comparison was performed along the 27 tracts of interest controlling for confounding hearing loss, tinnitus severity, and duration since onset. We tested a potential correlation with hearing loss, tinnitus duration and tinnitus handicap score along these tracts. In the tinnitus group, we observed increased AFD related to chronic tinnitus percept after acoustic trauma in two main white matter regions. First, in the right hemisphere, in the isthmus between inferior temporal and inferior frontal cortices, in the uncinate fasciculus (UF), and in the inferior fronto-occipital bundle (IFO). Second, in the left hemisphere, underneath the superior parietal region in the thalamo parietal tract and parieto-occipital pontine tract. Between-group differences in the acoustic radiations were not significant with AFD but were with FA. Furthermore, significant correlations with hearing loss were found in the left hemisphere in the inferior longitudinal fasciculus and in the fronto-pontine tract. No additional correlation was found with tinnitus duration nor with tinnitus handicap, as reflected by THI scores. The regions that displayed tinnitus related increased AFD also displayed increased FA. The isthmus of the UF and IFO in the right hemisphere appear to be involved with a number of neuropsychiatric and traumatic disorders confirming the involvement of the limbic system even in chronic non-bothersome tinnitus subjects, potentially suggesting a common pathway between these pathologies. White matter changes underneath the superior parietal cortex found here in tinnitus participants supports the implication of an auditory-somatosensory pathway in tinnitus perception.

TractLearn: A geodesic learning framework for quantitative analysis of brain bundles.

Deep learning-based convolutional neural networks have recently proved their efficiency in providing fast segmentation of major brain fascicles structures, based on diffusion-weighted imaging. The quantitative analysis of brain fascicles then relies on metrics either coming from the tractography process itself or from each voxel along the bundle. Statistical detection of abnormal voxels in the context of disease usually relies on univariate and multivariate statistics models, such as the General Linear Model (GLM). Yet in the case of high-dimensional low sample size data, the GLM often implies high standard deviation range in controls due to anatomical variability, despite the commonly used smoothing process. This can lead to difficulties to detect subtle quantitative alterations from a brain bundle at the voxel scale. Here we introduce TractLearn, a unified framework for brain fascicles quantitative analyses by using geodesic learning as a data-driven learning task. TractLearn allows a mapping between the image high-dimensional domain and the reduced latent space of brain fascicles using a Riemannian approach. We illustrate the robustness of this method on a healthy population with test-retest acquisition of multi-shell diffusion MRI data, demonstrating that it is possible to separately study the global effect due to different MRI sessions from the effect of local bundle alterations. We have then tested the efficiency of our algorithm on a sample of 5 age-matched subjects referred with mild traumatic brain injury. Our contributions are to propose: 1/ A manifold approach to capture controls variability as standard reference instead of an atlas approach based on a Euclidean mean. 2/ A tool to detect global variation of voxels' quantitative values, which accounts for voxels' interactions in a structure rather than analyzing voxels independently. 3/ A ready-to-plug algorithm to highlight nonlinear variation of diffusion MRI metrics. With this regard, TractLearn is a ready-to-use algorithm for precision medicine.

Imaging of endolymphatic hydrops: A comprehensive update in primary and secondary hydropic ear disease.

BACKGROUND: Since the first description by Hallpike and Cairns, the excess of endolymphatic fluid, also known as endolymphatic hydrops (EH), has been established as being the main biomarker in patients with Menière's disease. Recently, the concept of primary (PHED) and secondary hydropic ear disease (SHED) has been introduced. PHED corresponded to Menière's disease while SHED was defined as the presence of EH in patients with pre-existing inner ear disease. OBJECTIVE: In this article, we would like to summarize the methodology of hydrops exploration using MRI and the previously published radiological findings in patients with PHED and SHED. RESULTS: Before the emergence of delayed inner ear MRI, the presence of EH was assumed based on clinical symptoms. However, because of the recent technical developments, inner ear MRI became an important tool in clinical settings for identifying EH in vivo, in patients with PHED and SHED. The presence of EH on MRI is related with the degree of sensorineural hearing loss whether in patients with PHED or SHED. By contrast, in PHED or SHED patients without sensorineural hearing loss, MRI showed no sign of EH. CONCLUSIONS: Thanks to the recent technical developments, inner ear MRI became an important tool in clinical settings for identifying EH in vivo, in patients with PHED and SHED.

MRI diagnosis of saccular hydrops: Comparison of heavily-T2 FIESTA-C and 3D-FLAIR sequences with delayed acquisition.

PURPOSE: Currently, 3D-FLAIR sequence performed 4hours after the intravenous administration of a single dose of contrast media is the imaging technique of choice for the diagnosis of saccular hydrops (SH). Recently, the diagnosis of SH has also been reported with heavily-T2 weighted sequences. MATERIALS AND METHODS: In this retrospective imaging study, we performed 3D-FLAIR sequences 4hours after contrast media administration and 3D FIESTA-C sequences before and 4hours after contrast media administration in 30 patients with unilateral definite, probable or possible clinical diagnosis of Menière's disease (MD). Two radiologists, blinded to the clinical data, independently assessed the presence of saccular hydrops. Inter-reader agreement tests were performed. RESULTS: On delayed post-contrast 3D-FLAIR sequence, 15 patients out of 30 referred with a SH that was never seen on the controlateral asymptomatic side. The specificity and the sensitivity to detect MD side were 100% and 50% respectively. On non-enhanced 3D FIESTA-C sequence, 16 patients out of 30 (53%) referred with a saccular hydrops that was observed in 6 patients on the clinical asymptomatic ear. The specificity and the sensitivity to detect MD side were 80% and 33% respectively. On delayed 3D FIESTA-C sequence, 13 patients out of 30 (43%) referred with a saccular hydrops that was seen in 4 patients on the controlateral asymptomatic side. The specificity and the sensitivity to detect MD side were 83% and 27% respectively. CONCLUSION: Delayed post-contrast 3D-FLAIR is highly specific of MD symptoms while 3D FIESTA-C sequences are less sensitive and specific for the diagnosis of SH.

MRI contribution for the detection of endolymphatic hydrops in patients with superior canal dehiscence syndrome.

PURPOSE: Patients with superior semicircular canal dehiscence syndrome, which can only be treated by surgery, present cochleo-vestibular symptoms related to a third-mobile window but also endolymphatic hydrops. Since cVEMP and oVEMP are disturbed by the presence of the dehiscence, the aim of the study is to assess the value of MRI for the diagnosis of endolymphatic hydrops in patients with superior semicircular canal dehiscence syndrome in comparison with cVEMP and oVEMP. METHODS: In this retrospective cohort study we enrolled 33 ears in 24 patients with superior semicircular dehiscence syndrome who underwent a 4-h delayed intravenous Gd-enhanced 3D-FLAIR MRI and pure tone audiometry, cVEMP and oVEMP. For each patient MRI images were evaluated by two radiologists who used a compartmental endolymphatic hydrops grading system in comparison with cVEMP and oVEMP. RESULTS: Endolymphatic hydrops was found on MRI in 9 out of 33 SCDS ears (27.3%). We found no significant correlation between the presence of endolymphatic hydrops on MRI and cVEMP and oVEMP (p = 0.36 and p = 0.7, respectively). However, there was a significant correlation between the presence of endolymphatic hydrops on MRI and the degree of sensorineural hearing loss, Air Conduction-Pure Tone Average level (p = 0.012) and Bone Conduction-Pure Tone Average level (p = 0.09), respectively. CONCLUSION: We demonstrated that EH might be observed in 27.3% of superior semicircular dehiscence syndrome ears. The role of inner ear MRI is important to detect endolymphatic hydrops, since cVEMP and oVEMP are disturbed by the presence of the dehiscence, because these patients could benefit from a medical treatment. LEVEL OF EVIDENCE: Level 3.

Machine learning assisted DSC-MRI radiomics as a tool for glioma classification by grade and mutation status.

BACKGROUND: Combining MRI techniques with machine learning methodology is rapidly gaining attention as a promising method for staging of brain gliomas. This study assesses the diagnostic value of such a framework applied to dynamic susceptibility contrast (DSC)-MRI in classifying treatment-naïve gliomas from a multi-center patients into WHO grades II-IV and across their isocitrate dehydrogenase (IDH) mutation status. METHODS: Three hundred thirty-three patients from 6 tertiary centres, diagnosed histologically and molecularly with primary gliomas (IDH-mutant = 151 or IDH-wildtype = 182) were retrospectively identified. Raw DSC-MRI data was post-processed for normalised leakage-corrected relative cerebral blood volume (rCBV) maps. Shape, intensity distribution (histogram) and rotational invariant Haralick texture features over the tumour mask were extracted. Differences in extracted features across glioma grades and mutation status were tested using the Wilcoxon two-sample test. A random-forest algorithm was employed (2-fold cross-validation, 250 repeats) to predict grades or mutation status using the extracted features. RESULTS: Shape, distribution and texture features showed significant differences across mutation status. WHO grade II-III differentiation was mostly driven by shape features while texture and intensity feature were more relevant for the III-IV separation. Increased number of features became significant when differentiating grades further apart from one another. Gliomas were correctly stratified by mutation status in 71% and by grade in 53% of the cases (87% of the gliomas grades predicted with distance less than 1). CONCLUSIONS: Despite large heterogeneity in the multi-center dataset, machine learning assisted DSC-MRI radiomics hold potential to address the inherent variability and presents a promising approach for non-invasive glioma molecular subtyping and grading.