Standardized evaluation of the extent of resection in glioblastoma with automated early post-operative segmentation.

Standard treatment of patients with glioblastoma includes surgical resection of the tumor. The extent of resection (EOR) achieved during surgery significantly impacts prognosis and is used to stratify patients in clinical trials. In this study, we developed a U-Net-based deep-learning model to segment contrast-enhancing tumor on post-operative MRI exams taken within 72 h of resection surgery and used these segmentations to classify the EOR as either maximal or submaximal. The model was trained on 122 multiparametric MRI scans from our institution and achieved a mean Dice score of 0.52 ± 0.03 on an external dataset (n = 248), a performance -on par with the interrater agreement between expert annotators as reported in literature. We obtained an EOR classification precision/recall of 0.72/0.78 on the internal test dataset (n = 462) and 0.90/0.87 on the external dataset. Furthermore, Kaplan-Meier curves were used to compare the overall survival between patients with maximal and submaximal resection in the internal test dataset, as determined by either clinicians or the model. There was no significant difference between the survival predictions using the model's and clinical EOR classification. We find that the proposed segmentation model is capable of reliably classifying the EOR of glioblastoma tumors on early post-operative MRI scans. Moreover, we show that stratification of patients based on the model's predictions offers at least the same prognostic value as when done by clinicians.

Constructing personalized characterizations of structural brain aberrations in patients with dementia using explainable artificial intelligence.

Deep learning approaches for clinical predictions based on magnetic resonance imaging data have shown great promise as a translational technology for diagnosis and prognosis in neurological disorders, but its clinical impact has been limited. This is partially attributed to the opaqueness of deep learning models, causing insufficient understanding of what underlies their decisions. To overcome this, we trained convolutional neural networks on structural brain scans to differentiate dementia patients from healthy controls, and applied layerwise relevance propagation to procure individual-level explanations of the model predictions. Through extensive validations we demonstrate that deviations recognized by the model corroborate existing knowledge of structural brain aberrations in dementia. By employing the explainable dementia classifier in a longitudinal dataset of patients with mild cognitive impairment, we show that the spatially rich explanations complement the model prediction when forecasting transition to dementia and help characterize the biological manifestation of disease in the individual brain. Overall, our work exemplifies the clinical potential of explainable artificial intelligence in precision medicine.

Radiological features of brain hemorrhage through automated segmentation from computed tomography in stroke and traumatic brain injury.

Introduction: Radiological assessment is necessary to diagnose spontaneous intracerebral hemorrhage (ICH) and traumatic brain injury intracranial hemorrhage (TBI-bleed). Artificial intelligence (AI) deep learning tools provide a means for decision support. This study evaluates the hemorrhage segmentations produced from three-dimensional deep learning AI model that was developed using non-contrast computed tomography (CT) imaging data external to the current study. Methods: Non-contrast CT imaging data from 1263 patients were accessed across seven data sources (referred to as sites) in Norway and Sweden. Patients were included based on ICH, TBI-bleed, or mild TBI diagnosis. Initial non-contrast CT images were available for all participants. Hemorrhage location frequency maps were generated. The number of estimated haematoma clusters was correlated with the total haematoma volume. Ground truth expert annotations were available for one ICH site; hence, a comparison was made with the estimated haematoma volumes. Segmentation volume estimates were used in a receiver operator characteristics (ROC) analysis for all samples (i.e., bleed detected) and then specifically for one site with few TBI-bleed cases. Results: The hemorrhage frequency maps showed spatial patterns of estimated lesions consistent with ICH or TBI-bleed presentations. There was a positive correlation between the estimated number of clusters and total haematoma volume for each site (correlation range: 0.45-0.74; each p-value < 0.01) and evidence of ICH between-site differences. Relative to hand-drawn annotations for one ICH site, the VIOLA-AI segmentation mask achieved a median Dice Similarity Coefficient of 0.82 (interquartile range: 0.78 and 0.83), resulting in a small overestimate in the haematoma volume by a median of 0.47 mL (interquartile range: 0.04 and 1.75 mL). The bleed detection ROC analysis for the whole sample gave a high area-under-the-curve (AUC) of 0.92 (with sensitivity and specificity of 83.28% and 95.41%); however, when considering only the mild head injury site, the TBI-bleed detection gave an AUC of 0.70. Discussion: An open-source segmentation tool was used to visualize hemorrhage locations across multiple data sources and revealed quantitative hemorrhage site differences. The automated total hemorrhage volume estimate correlated with a per-participant hemorrhage cluster count. ROC results were moderate-to-high. The VIOLA-AI tool had promising results and might be useful for various types of intracranial hemorrhage.

Diagnostic imaging strategies of acute intracerebral hemorrhage in European academic hospitals-a decision-making analysis.

PURPOSE: To evaluate and compare which factors are relevant to the diagnostic decision-making and imaging workup of intracerebral hemorrhages in large, specialized European centers. METHODS: Expert neuroradiologists from ten large, specialized centers (where endovascular stroke treatment is routinely performed) in nine European countries were selected in cooperation with the European Society of Neuroradiology (ESNR). The experts were asked to describe how and when they would investigate specific causes in a patient who presented with an acute, atraumatic, intracerebral hemorrhage for two given locations: (1) basal ganglia, thalamus, pons or cerebellum; (2) lobar hemorrhage. Answers were collected, and decision trees were compared. RESULTS: Criteria that were considered relevant for decision-making reflect recommendations from current guidelines and were similar in all participating centers. CT Angiography or MR angiography was considered essential by the majority of centers regardless of other factors. Imaging in clinical practice tended to surpass guideline recommendations and was heterogeneous among different centers, e.g., in a scenario suggestive of typical hypertensive hemorrhage, recommendations ranged from no further follow-up imaging to CT angiography and MR angiography. In no case was a consensus above 60% achieved. CONCLUSION: In European clinical practices, existing guidelines for diagnostic imaging strategies in ICH evaluation are followed as a basis but in most cases, additional imaging investigation is undertaken. Significant differences in imaging workup were observed among the centers. Results suggest a high level of awareness and caution regarding potentially underlying pathology other than hypertensive disease.

Longitudinal brain age prediction and cognitive function after stroke.

Advanced age is associated with post-stroke cognitive decline. Machine learning based on brain scans can be used to estimate brain age of patients, and the corresponding difference from chronological age, the brain age gap (BAG), has been investigated in a range of clinical conditions, yet not thoroughly in post-stroke neurocognitive disorder (NCD). We aimed to investigate the association between BAG and post-stroke NCD over time. Lower BAG (younger appearing brain compared to chronological age) was found associated with lower risk of post-stroke NCD up to 36 months after stroke, even among those showing no evidence of impairments 3 months after hospital admission. For patients with no NCD at baseline, survival analysis suggested that higher baseline BAG was associated with higher risk of post-stroke NCD at 18 and 36 months. In conclusion, a younger appearing brain is associated with a lower risk of post-stroke NCD.

Neuroimaging markers of dual impairment in cognition and physical performance following stroke: The Nor-COAST study.

Background: Cognitive decline and decline in physical performance are common after stroke. Concurrent impairments in the two domains are reported to give increased risk of dementia and functional decline. The concept of dual impairment of physical performance and cognition after stroke is poorly investigated. Clinically accessible imaging markers of stroke and pre-existing brain pathology might help identify patients at risk. Objective: The primary aim of this study was to investigate to which extent pre-stroke cerebral pathology was associated with dual impairment in cognition and physical performance at time of stroke. Secondary aims were to examine whether white matter hyperintensities, medial temporal lobe atrophy, and stroke lesion volume and location were associated with dual impairment. Methods: Participants from the Norwegian Cognitive Impairment After Stroke (Nor-COAST) study with available MRI data at baseline were included in this cross-sectional study. Logistic regression analyses were conducted, with impairment status (no impairment, impaired cognition, impaired physical performance, and dual impairment) as the dependent variable and MRI markers as covariates. Pre-existing brain pathologies were classified into neurodegenerative, cerebrovascular, or mixed pathology. In addition, white matter hyperintensities and medial temporal lobe atrophy were included as independent covariates. Stroke volume and location were also ascertained from study-specific MRI scans. Results: Participants' (n = 348) mean (SD) age was 72.3 (11.3) years; 148 (42.5%) were women. Participants with dual impairment (n = 99) were significantly older, had experienced a more severe stroke, and had a higher comorbidity burden and poorer pre-stroke function. Stroke lesion volume (odds ratio 1.03, 95%, confidence interval 1.00 to 1.05, p = 0.035), but not stroke location or pre-existing brain pathology, was associated with dual impairment, after adjusting for age and sex. Conclusion: In this large cohort of stroke survivors having suffered mainly mild to moderate stroke, stroke lesion volume-but not pre-existing brain pathology-was associated with dual impairment early after stroke, confirming the role of stroke severity in functional decline.

Longitudinal Brain Changes After Stroke and the Association With Cognitive Decline.

Background: Cognitive impairment is common after stroke. So is cortical- and subcortical atrophy, with studies reporting more atrophy in the ipsilesional hemisphere than the contralesional hemisphere. The current study aimed to investigate the longitudinal associations between (I) lateralization of brain atrophy and stroke hemisphere, and (II) cognitive impairment and brain atrophy after stroke. We expected to find that (I) cortical thickness and hippocampal-, thalamic-, and caudate nucleus volumes declined more in the ipsilesional than the contralesional hemisphere up to 36 months after stroke. Furthermore, we predicted that (II) cognitive decline was associated with greater stroke volumes, and with greater cortical thickness and subcortical structural volume atrophy across the 36 months. Methods: Stroke survivors from five Norwegian hospitals were included from the multisite-prospective "Norwegian Cognitive Impairment After Stroke" (Nor-COAST) study. Analyses were run with clinical, neuropsychological and structural magnetic resonance imaging (MRI) data from baseline, 18- and 36 months. Cortical thicknesses and subcortical volumes were obtained via FreeSurfer segmentations and stroke lesion volumes were semi-automatically derived using ITK-SNAP. Cognition was measured using MoCA. Results: Findings from 244 stroke survivors [age = 72.2 (11.3) years, women = 55.7%, stroke severity NIHSS = 4.9 (5.0)] were included at baseline. Of these, 145 (59.4%) had an MRI scan at 18 months and 72 (49.7% of 18 months) at 36 months. Most cortices and subcortices showed a higher ipsi- compared to contralesional atrophy rate, with the effect being more prominent in the right hemisphere. Next, greater degrees of atrophy particularly in the medial temporal lobe after left-sided strokes and larger stroke lesion volumes after right-sided strokes were associated with cognitive decline over time. Conclusion: Atrophy in the ipsilesional hemisphere was greater than in the contralesional hemisphere over time. This effect was found to be more prominent in the right hemisphere, pointing to a possible higher resilience to stroke of the left hemisphere. Lastly, greater atrophy of the cortex and subcortex, as well as larger stroke volume, were associated with worse cognition over time and should be included in risk assessments of cognitive decline after stroke.

Is Frailty Index a better predictor than pre-stroke modified Rankin Scale for neurocognitive outcomes 3-months post-stroke?

BACKGROUND: The prognostic value of frailty measures for post-stroke neurocognitive disorder (NCD) remains to be evaluated. AIMS: The aim of this study was to compare the predictive value of pre-stroke FI with pre-stroke modified Rankin Scale (mRS) for post-stroke cognitive impairment. Further, we explored the added value of including FI in prediction models for cognitive prognosis post-stroke. METHODS: We generated a 36-item Frailty Index (FI), based on the Rockwood FI, to measure frailty based on pre-stroke medical conditions recorded in the Nor-COAST multicentre prospective study baseline assessments. Consecutive participants with a FI score and completed cognitive test battery at three months were included. We generated Odds Ratio (OR) with NCD as the dependent variable. The predictors of primary interest were pre-stroke frailty and mRS. We also measured the predictive values of mRS and FI by the area (AUC) under the receiver operating characteristic curve. RESULTS: 598 participants (43.0% women, mean/SD age = 71.6/11.9, mean/SD education = 12.5/3.8, mean/SD pre-stroke mRS = 0.8/1.0, mean/SD GDS pre-stroke = 1.4/0.8, mean/SD NIHSS day 1 3/4), had a FI mean/SD score = 0.14/0.10. The logistic regression analyses showed that FI (OR 3.09), as well as the mRS (OR 2.21), were strong predictors of major NCD. When FI and mRS were entered as predictors simultaneously, the OR for mRS decreased relatively more than that for FI. AUC for NCD post-stroke was higher for FI than for mRS, both for major NCD (0.762 vs 0.677) and for any NCD (0.681 vs 0.638). CONCLUSIONS: FI is a stronger predictor of post-stroke NCD than pre-stroke mRS and could be a part of the prediction models for cognitive prognosis post-stroke. TRIAL REGISTRATION: ClinicalTrials.gov Identifier: NCT02650531 .

Predicting the Emergence of Major Neurocognitive Disorder Within Three Months After a Stroke.

Background: Neurocognitive disorder (NCD) is common after stroke, with major NCD appearing in about 10% of survivors of a first-ever stroke. We aimed to classify clinical- and imaging factors related to rapid development of major NCD 3 months after a stroke, so as to examine the optimal composition of factors for predicting rapid development of the disorder. We hypothesized that the prediction would mainly be driven by neurodegenerative as opposed to vascular brain changes. Methods: Stroke survivors from five Norwegian hospitals were included from the "Norwegian COgnitive Impairment After STroke" (Nor-COAST) study. A support vector machine (SVM) classifier was trained to distinguish between patients who developed major NCD 3 months after the stroke and those who did not. Potential predictor factors were based on previous literature and included both vascular and neurodegenerative factors from clinical and structural magnetic resonance imaging findings. Cortical thickness was obtained via FreeSurfer segmentations, and volumes of white matter hyperintensities (WMH) and stroke lesions were semi-automatically gathered using FSL BIANCA and ITK-SNAP, respectively. The predictive value of the classifier was measured, compared between classifier models and cross-validated. Results: Findings from 227 stroke survivors [age = 71.7 (11.3), males = (56.4%), stroke severity NIHSS = 3.8 (4.8)] were included. The best predictive accuracy (AUC = 0.876) was achieved by an SVM classifier with 19 features. The model with the fewest number of features that achieved statistically comparable accuracy (AUC = 0.850) was the 8-feature model. These features ranked by their weighting were; stroke lesion volume, WMH volume, left occipital and temporal cortical thickness, right cingulate cortical thickness, stroke severity (NIHSS), antiplatelet medication intake, and education. Conclusion: The rapid (<3 months) development of major NCD after stroke is possible to predict with an 87.6% accuracy and seems dependent on both neurodegenerative and vascular factors, as well as aspects of the stroke itself. In contrast to previous literature, we also found that vascular changes are more important than neurodegenerative ones. Although possible to predict with relatively high accuracy, our findings indicate that the development of rapid onset post-stroke NCD may be more complex than earlier suggested.

Pre-stroke cognitive impairment is associated with vascular imaging pathology: a prospective observational study.

BACKGROUND: Chronic brain pathology and pre-stroke cognitive impairment (PCI) is predictive of post-stroke dementia. The aim of the current study was to measure pre-stroke neurodegenerative and vascular disease burden found on brain MRI and to assess the association between pre-stroke imaging pathology and PCI, whilst also looking for potential sex differences. METHODS: This prospective brain MRI cohort is part of the multicentre Norwegian cognitive impairment after stroke (Nor-COAST) study. Patients hospitalized with acute ischemic or hemorrhagic stroke were included from five participating stroke units. Visual rating scales were used to categorize baseline MRIs (N = 410) as vascular, neurodegenerative, mixed, or normal, based on the presence of pathological imaging findings. Pre-stroke cognition was assessed by interviews of patients or caregivers using the Global Deterioration Scale (GDS). Stroke severity was assessed with the National Institute of Health Stroke Scale (NIHSS). Univariate and multiple logistic regression analyses were performed to investigate the association between imaging markers, PCI, and sex. RESULTS: Patients' (N = 410) mean (SD) age was 73.6 (±11) years; 182 (44%) participants were female, the mean (SD) NIHSS at admittance was 4.1 (±5). In 68% of the participants, at least one pathological imaging marker was found. Medial temporal lobe atrophy (MTA) was present in 30% of patients, white matter hyperintensities (WMH) in 38% of patients and lacunes in 35% of patients. PCI was found in 30% of the patients. PCI was associated with cerebrovascular pathology (OR 2.5; CI = 1.4 to 4.5, p = 0.001) and mixed pathology (OR 3.4; CI = 1.9 to 6.1, p = 0.001) but was not associated with neurodegeneration (OR 1.0; CI = 0.5 to 2.2; p = 0.973). Pathological MRI markers, including MTA and lacunes, were more prevalent among men, as was a history of clinical stroke prior to the index stroke. The OR of PCI for women was not significantly increased (OR 1.2; CI = 0.8 to 1.9; p = 0.3). CONCLUSIONS: Pre-stroke chronic brain pathology is common in stroke patients, with a higher prevalence in men. Vascular pathology and mixed pathology are associated with PCI. There were no significant sex differences for the risk of PCI. TRIAL REGISTRATION: NCT02650531 , date of registration: 08.01.2016.

Clinically accessible neuroimaging predictors of post-stroke neurocognitive disorder: a prospective observational study.

BACKGROUND: Neurocognitive disorder (NCD) is common in stroke survivors. We aimed to identify clinically accessible imaging markers of stroke and chronic pathology that are associated with early post-stroke NCD. METHODS: We included 231 stroke survivors from the "Norwegian Cognitive Impairment after Stroke (Nor-COAST)" study who underwent a standardized cognitive assessment 3 months after the stroke. Any NCD (mild cognitive impairment and dementia) and major NCD (dementia) were diagnosed according to "Diagnostic and Statistical Manual of Mental Disorders (DSM-5)" criteria. Clinically accessible imaging findings were analyzed on study-specific brain MRIs in the early phase after stroke. Stroke lesion volumes were semi automatically quantified and strategic stroke locations were determined by an atlas based coregistration. White matter hyperintensities (WMH) and medial temporal lobe atrophy (MTA) were visually scored. Logistic regression was used to identify neuroimaging findings associated with major NCD and any NCD. RESULTS: Mean age was 71.8 years (SD 11.1), 101 (43.7%) were females, mean time from stroke to imaging was 8 (SD 16) days. At 3 months 63 (27.3%) had mild NCD and 65 (28.1%) had major NCD. Any NCD was significantly associated with WMH pathology (odds ratio (OR) = 2.73 [1.56 to 4.77], p = 0.001), MTA pathology (OR = 1.95 [1.12 to 3.41], p = 0.019), and left hemispheric stroke (OR = 1.8 [1.05 to 3.09], p = 0.032). Major NCD was significantly associated with WMH pathology (OR = 2.54 [1.33 to 4.84], p = 0.005) and stroke lesion volume (OR (per ml) =1.04 [1.01 to 1.06], p = 0.001). CONCLUSION: WMH pathology, MTA pathology and left hemispheric stroke were associated with the development of any NCD. Stroke lesion volume and WMH pathology were associated with the development of major NCD 3 months after stroke. These imaging findings may be used in the routine clinical setting to identify patients at risk for early post-stroke NCD. TRIAL REGISTRATION: ClinicalTrials.gov, NCT02650531 , Registered 8 January 2016 - Retrospectively registered.

Analysis of diffusion tensor imaging metrics for gliomas grading at 3 T.

OBJECTIVES: To assess the diagnostic accuracy of axial diffusivity (AD), radial diffusivity (RD), apparent diffusion coefficient (ADC) and fractional anisotropy (FA) values derived from DTI for grading of glial tumors, and to estimate the correlation between DTI parameters and tumor grades. METHODS: Seventy-eight patients with glial tumors underwent DTI. AD, RD, ADC and FA values of tumor, peritumoral edema and contralateral normal-appearing white matter (NAWM) and AD, RD, ADC and FA ratios: lowest average AD, RD, ADC and FA values in tumor or peritumoral edema to AD, RD, ADC and FA of NAWM were calculated. DTI parameters and tumor grades were analyzed statistically and with Pearson correlation. Receiver operating characteristic (ROC) curve analysis was also performed. RESULTS: The differences in ADC, AD and RD tumor values, and ADC and RD tumor ratios were statistically significant between grades II and III, grades II and IV, and between grades II and III-IV. The AD tumor ratio differed significantly among all tumor grades. Tumor ADC, AD, RD and glial tumor grades were strongly correlated. In the ROC curve analysis, the area under the curve (AUC) of the parameter tumor ADC was the largest for distinguishing grade II from grades III to IV (98.5%), grade II from grade IV (98.9%) and grade II from grade III (97.0%). CONCLUSION: ADC, RD and AD are useful DTI parameters for differentiation between low- and high-grade gliomas with a diagnostic accuracy of more than 90%. Our study revealed a good inverse correlation between ADC, RD, AD and WHO grades II-IV astrocytic tumors.

[High-grade gliomas in adults]. / Høygradige gliomer hos voksne.

BACKGROUND: High-grade glioma is a primary malignant brain tumour which affects about 200 Norwegian patients each year. Diagnosis and treatment of high-grade gliomas in adults has been reviewed. MATERIAL AND METHODS: The article is based on recent literature retrieved through a non-systematic search in PubMed and the authors' experience with the patient group. RESULTS: The most common symptoms are focal neurological deficits, epileptic seizures and pressure symptoms. The patients should be examined by magnetic resonance (MR) imaging and the diagnosis confirmed with biopsy. No curative treatment is currently available for high-grade gliomas. The standard treatment is surgical resection followed by radiation therapy alone or in combination with chemotherapy (temozolomid). Five-year survival is only 6.1 %. INTERPRETATION: The diagnosis is composite with both neurological symptoms and cognitive problems. This requires good communication with the patient and close cooperation between various departments and the primary health services. Symptomatic treatment and multidisciplinary follow-up is necessary.

Measurements of diagnostic examination performance and correlation analysis using microvascular leakage, cerebral blood volume, and blood flow derived from 3T dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging in glial tumor grading.

INTRODUCTION: To assess the diagnostic accuracy of microvascular leakage (MVL), cerebral blood volume (CBV) and blood flow (CBF) values derived from dynamic susceptibility-weighted contrast-enhanced perfusion MR imaging (DSC-MR imaging) for grading of cerebral glial tumors, and to estimate the correlation between vascular permeability/perfusion parameters and tumor grades. METHODS: A prospective study of 79 patients with cerebral glial tumors underwent DSC-MR imaging. Normalized relative CBV (rCBV) and relative CBF (rCBF) from tumoral (rCBVt and rCBFt), peri-enhancing region (rCBVe and rCBFe), and the value in the tumor divided by the value in the peri-enhancing region (rCBVt/e and rCBFt/e), as well as MVL, expressed as the leakage coefficient K(2) were calculated. Hemodynamic variables and tumor grades were analyzed statistically and with Pearson correlations. Receiver operating characteristic (ROC) curve analyses were also performed for each of the variables. RESULTS: The differences in rCBVt and the maximum MVL (MVL(max)) values were statistically significant among all tumor grades. Correlation analysis using Pearson was as follows: rCBVt and tumor grade, r = 0.774; rCBFt and tumor grade, r = 0.417; MVL(max) and tumor grade, r = 0.559; MVL(max) and rCBVt, r = 0.440; MVL(max) and rCBFt, r = 0.192; and rCBVt and rCBFt, r = 0.605. According to ROC analyses for distinguishing tumor grade, rCBVt showed the largest areas under ROC curve (AUC), except for grade III from IV. CONCLUSION: Both rCBVt and MVL(max) showed good discriminative power in distinguishing all tumor grades. rCBVt correlated strongly with tumor grade; the correlation between MVL(max) and tumor grade was moderate.

Proton magnetic resonance spectroscopy in the distinction of high-grade cerebral gliomas from single metastatic brain tumors.

BACKGROUND: Brain metastases and primary high-grade gliomas, including glioblastomas multiforme (GBM) and anaplastic astrocytomas (AA), may be indistinguishable by conventional magnetic resonance (MR) imaging. Identification of these tumors may have therapeutic consequences. PURPOSE: To assess the value of MR spectroscopy (MRS) using short and intermediate echo time (TE) in differentiating solitary brain metastases and high-grade gliomas on the basis of differences in metabolite ratios in the intratumoral and peritumoral region. MATERIAL AND METHODS: We performed MR imaging and MRS in 73 patients with histologically verified intraaxial brain tumors: 53 patients with high-grade gliomas (34 GBM and 19 AA) and 20 patients with metastatic brain tumors. The metabolite ratios of Cho/Cr, Cho/NAA, and NAA/Cr at intermediate TE and the presence of lipids at short TE were assessed from spectral maps in the tumoral core, peritumoral edema, and contralateral normal-appearing white matter. The differences in the metabolite ratios between high-grade gliomas/GBM/AA and metastases were analyzed statistically. Cutoff values of Cho/Cr, Cho/NAA, and NAA/Cr ratios in the peritumoral edema, as well as Cho/Cr and NAA/Cr ratios in the tumoral core for distinguishing high-grade gliomas/GBM/AA from metastases were determined by receiver operating characteristic (ROC) curve analysis. RESULTS: Significant differences were noted in the peritumoral Cho/Cr, Cho/NAA, and NAA/ Cr ratios between high-grade gliomas/GBM/AA and metastases. ROC analysis demonstrated a cutoff value of 1.24 for peritumoral Cho/Cr ratio to provide sensitivity, specificity, positive (PPV), and negative predictive values (NPV) of 100%, 88.9%, 80.0%, and 100%, respectively, for discrimination between high-grade gliomas and metastases. By using a cutoff value of 1.11 for peritumoral Cho/NAA ratio, the sensitivity was 100%, the specificity was 91.1%, the PPV was 83.3%, and the NPV was 100%. CONCLUSION: The results of this study demonstrate that MRS can differentiate high-grade gliomas from metastases, especially with peritumoral measurements, supporting the hypothesis that MRS can detect infiltration of tumor cells in the peritumoral edema.