Search

Background and purposeBrain tumors are among the leading causes of cancer deaths in children. Initial diagnosis based on MR images can be a challenging task for radiologists, depending on the tumor type and location. Deep learning methods could support the diagnosis by predicting the tumor type.Materials and methodsA subset (181 subjects) of the data from “Children’s Brain Tumor Network” (CBTN) was used, including infratentorial and supratentorial tumors, with the main tumor types being low-grade astrocytomas, ependymomas, and medulloblastomas. T1w-Gd, T2-w, and ADC MR sequences were used separately. Classification was performed on 2D MR images using four different off-the-shelf deep learning models and a custom-designed shallow network all pretrained on adult MR images. Joint fusion was implemented to combine image and age data, and tumor type prediction was computed volume-wise. Matthew’s correlation coefficient (MCC), accuracy, and F1 scores were used to assess the models’ performance. Model explainability, using gradient-weighted class activation mapping (Grad-CAM), was implemented and the network’s attention on the tumor region was quantified.ResultsThe shallow custom network resulted in the highest classification performance when trained on T2-w or ADC MR images fused with age information, when considering infratentorial tumors only (MCC: 0.71 for ADC and 0.57 for T2-w), and both infra- and supratentorial tumors (MCC: 0.70 for ADC and 0.57 for T2-w).ConclusionClassification of pediatric brain tumors on MR images could be accomplished using deep learning, and the fusion of age information improved model performance.

The infiltrative nature of malignant gliomas results in active tumor spreading into the peritumoral edema, which is not visible in conventional magnetic resonance imaging (cMRI) even after contrast injection. MR relaxometry (qMRI) measures relaxation rates dependent on tissue properties, and can offer additional contrast mechanisms to highlight the non-enhancing infiltrative tumor. The aim of this study is to investigate if qMRI data provides additional information compared to cMRI sequences (T1w, T1wGd, T2w, FLAIR), when considering deep learning-based brain tumor (1) detection and (2) segmentation. A total of 23 patients with histologically confirmed malignant glioma were retrospectively included in the study. Quantitative MR imaging was used to obtain R1(1/T1), R2(1/T2) and proton density maps pre- and post-gadolinium contrast injection. Conventional MR imaging was also performed. A 2D CNN detection model and a 2D U-Net were trained on transversal slices (n=528) using either cMRI or a combination of qMRI pre- and post-contrast data for tumor detection and segmentation, respectively. Moreover, trends in quantitative R1and R2rates of regions identified as relevant for tumor detection by model explainability methods were qualitatively analyzed. Tumor detection and segmentation performance for models trained with a combination of qMRI pre- and post-contrast was the highest (detection MCC=0.72, segmentation Dice=0.90), however, improvements were not statistically significant compared to cMRI (detection MCC=0.67, segmentation Dice=0.90). The analysis of the relaxation rates of the relevant regions identified using model explainability methods showed no differences between models trained on cMRI or qMRI. Relevant regions which fell outside the annotation showed changes in relaxation rates after contrast injection similar to those within the annotation, when looking at majority of the individual cases. A similar trend could not be seen when looking at relaxation trends over all the dataset. In conclusion, models trained on qMRI data obtain similar performance to those trained on cMRI data, with the advantage of quantitatively measuring brain tissue properties within the scan time (11.8 minutes for qMRI with and without contrast, and 12.2 minutes for cMRI). Moreover, when considering individual patients, regions identified by model explainability methods as relevant for tumor detection outside the manual annotation of the tumor showed changes in quantitative relaxation rates after contrast injection similar to regions within the annotation, suggestive of infiltrative tumor in the peritumoral edema.

Stereotactic radiosurgery (SRS) can be an effective primary or adjuvant treatment option for intracranial tumors. However, it carries risks of various radiation toxicities, including radionecrosis and functional deficits. Current SRS inverse planning algorithms allow efficient inclusion of organs at risk (OARs) in the treatment planning process, which will be spared by setting a maximum radiation dose. In this work we propose using activation maps from functional MRI to map the eloquent regions of the brain and define functional OARs. We evaluate the effects of these functional OARs for Elekta Leksell Gamma Knife SRS inverse treatment planning on open data MRI from 5 subjects. Our results show that fMRI-derived functional OARs can effectively be used to reduce the radiation dose incident on the eloquent brain regions, while maintaining acceptable treatment planning metrics on the tumor targets.

Gadolinium depositions in the brain caused by Gadolinium Based Contrast Agents (GBCAs) represent a topical issue since depositions have been confirmed, but their potential effects remain unknown. The purpose of this study is to show that quantitative MRI provides an alternative to conventional MRI for detection of gadolinium depositions without the need for reference areas, while also avoiding methodological limitations associated with conventional MRI, facilitating future studies on the subject. This retrospective observational cohort study of longitudinally acquired data uses quantitative MRI to quantify gadolinium depositions in six areas of the brain; the dentate nuclei, globus pallidi, thalami, pons, white matter and frontal cortex grey matter of 43 patients who received two to five doses of Magnevist (a linear GBCA) from 2009 to 2015. R1 was measured at baseline and after four years in the six areas of the brain of each patient. The changes in R1 were examined using a Paired-Samples T test. The correlation between doses of Magnevist and R1 was analysed using an Independent-Samples T test. There was a significant increase in R1 (p1 was found in the thalami, white matter or pons. A correlation between doses of Magnevist and R1 was found in the dentate nuclei and grey matter, but not in the globus pallidus. The results suggest that quantitative MRI is an alternative to conventional MRI for early detection of gadolinium depositions, while additionally providing quantitative measurements for comparison and follow-up.

BACKGROUND AND PURPOSE: Administration of a gadolinium-based contrast agent is an important diagnostic biomarker for blood-brain barrier damage. In clinical use, detection is based on subjective comparison of native and postgadolinium-based contrast agent T1-weighted images. Quantitative MR imaging studies have suggested a relation between the longitudinal relaxation rate and proton-density in the brain parenchyma, which is disturbed by gadolinium-based contrast agents. This discrepancy can be used to synthesize a contrast-enhancement map based solely on the postgadolinium-based contrast agent acquisition. The aim of this study was to compare synthetic enhancement maps with subtraction maps of native and postgadolinium-based contrast agent images. MATERIALS AND METHODS: For 14 patients with high-grade gliomas, quantitative MR imaging was performed before and after gadolinium-based contrast agent administration. The quantification sequence was multidynamic and multiecho, with a scan time of 6 minutes. The 2 image stacks were coregistered using in-plane transformation. The longitudinal relaxation maps were subtracted and correlated with the synthetic longitudinal relaxation enhancement maps on the basis of the postgadolinium-based contrast agent images only. ROIs were drawn for tumor delineation. RESULTS: Linear regression of the subtraction and synthetic longitudinal relaxation enhancement maps showed a slope of 1.02 ± 0.19 and an intercept of 0.05 ± 0.12. The Pearson correlation coefficient was 0.861 ± 0.059, and the coefficient of variation was 0.18 ± 0.04. On average, a volume of 1.71 ± 1.28 mL of low-intensity enhancement was detected in the synthetic enhancement maps outside the borders of the drawn ROI. CONCLUSIONS: The study shows that there was a good correlation between subtraction longitudinal relaxation enhancement maps and synthetic longitudinal relaxation enhancement maps in patients with high-grade gliomas. The method may improve the sensitivity and objectivity for the detection of gadolinium-based contrast agent enhancement.

BACKGROUND AND PURPOSE: Contrast-enhancing MS lesions are important markers of active inflammation in the diagnostic work-up of MS and in disease monitoring with MR imaging. Because intravenous contrast agents involve an expense and a potential risk of adverse events, it would be desirable to identify active lesions without using a contrast agent. The purpose of this study was to evaluate whether pre-contrast injection tissue-relaxation rates and proton density of MS lesions, by using a new quantitative MR imaging sequence, can identify active lesions. MATERIALS AND METHODS: Forty-four patients with a clinical suspicion of MS were studied. MR imaging with a standard clinical MS protocol and a quantitative MR imaging sequence was performed at inclusion (baseline) and after 1 year. ROIs were placed in MS lesions, classified as nonenhancing or enhancing. Longitudinal and transverse relaxation rates, as well as proton density were obtained from the quantitative MR imaging sequence. Statistical analyses of ROI values were performed by using a mixed linear model, logistic regression, and receiver operating characteristic analysis. RESULTS: Enhancing lesions had a significantly (P < .001) higher mean longitudinal relaxation rate (1.22 ± 0.36 versus 0.89 ± 0.24), a higher mean transverse relaxation rate (9.8 ± 2.6 versus 7.4 ± 1.9), and a lower mean proton density (77 ± 11.2 versus 90 ± 8.4) than nonenhancing lesions. An area under the receiver operating characteristic curve value of 0.832 was obtained. CONCLUSIONS: Contrast-enhancing MS lesions often have proton density and relaxation times that differ from those in nonenhancing lesions, with lower proton density and shorter relaxation times in enhancing lesions compared with nonenhancing lesions.

Background and purpose Brain atrophy is related to clinical deterioration in multiple sclerosis (MS) but its association with intrathecal markers of inflammation or neurodegeneration is unclear. Our aim was to investigate whether cerebrospinal fluid (CSF) markers of inflammation or neurodegeneration are associated with brain volume change in natalizumab-treated MS and whether this change is reflected in non-lesional white matter metabolites. Methods About 25 patients with natalizumab-treated MS were followed for 3 years with assessment of percentage brain volume change (PBVC) and absolute quantification of metabolites with proton magnetic resonance spectroscopy (1H MRS). Analyses of inflammatory [interleukin 1β (IL-1β), IL-6, C-X-C motif chemokine 8 (CXCL8), CXCL10, CXCL11, C-C motif chemokine 22] and neurodegenerative [neurofilament light protein (NFL), glial fibrillary acidic protein, myelin basic protein, tau proteins] markers were done at baseline and 1-year follow-up. Results The mean decline in PBVC was 3% at the 3-year follow-up, although mean 1H MRS metabolite levels in non-lesional white matter were unchanged. CSF levels of NFL and tau at baseline correlated negatively with PBVC over 3 years (r = −0.564, P = 0.012, and r = −0.592, P = 0.010, respectively). Conclusions A significant 3-year whole-brain atrophy was not reflected in mean metabolite change of non-lesional white matter. In addition, our results suggest that CSF levels of NFL and tau correlate with brain atrophy development and may be used for evaluating treatment response in inflammatory active MS. Funding agencies: Medical Research Council [K2013-61X-22310-01-4]; Swedish Society of Neurologically Disabled; Swedish Society of Medicine; National Research Council (VR/NT); Linkoping University; Linkoping University Hospital; County Council of Ostergotland