By combining karyotyping and RNA sequencing, we identified the 2 first ever reported ALK rearrangements in CNS tumors. Such rearrangements may represent the hallmark of a new entity of pediatric glioma characterized by both ependymal and astrocytic features. Our findings are of particular importance because crizotinib, a selective ALK inhibitor, has demonstrated effect in patients with lung cancer harboring ALK rearrangements. Thus, ALK emerges as an interesting therapeutic target in patients with ependymal tumors carrying ALK fusions.

The purpose of this study was to assess the clinical value of a deep learning (DL) model for automatic detection and segmentation of brain metastases, in which a neural network is trained on four distinct MRI sequences using an input-level dropout layer, thus simulating the scenario of missing MRI sequences by training on the full set and all possible subsets of the input data. This retrospective, multicenter study, evaluated 165 patients with brain metastases. The proposed input-level dropout (ILD) model was trained on multisequence MRI from 100 patients and validated/tested on 10/55 patients, in which the test set was missing one of the four MRI sequences used for training. The segmentation results were compared with the performance of a state-of-the-art DeepLab V3 model. The MR sequences in the training set included pre-gadolinium and post-gadolinium (Gd) T1-weighted 3D fast spin echo, post-Gd T1-weighted inversion recovery (IR) prepped fast spoiled gradient echo, and 3D fluid attenuated inversion recovery (FLAIR), whereas the test set did not include the IR prepped image-series. The ground truth segmentations were established by experienced neuroradiologists. The results were evaluated using precision, recall, Intersection over union (IoU)-score and Dice score, and receiver operating characteristics (ROC) curve statistics, while the Wilcoxon rank sum test was used to compare the performance of the two neural networks. The area under the ROC curve (AUC), averaged across all test cases, was 0.989 ± 0.029 for the ILD-model and 0.989 ± 0.023 for the DeepLab V3 model (p = 0.62). The ILD-model showed a significantly higher Dice score (0.795 ± 0.104 vs. 0.774 ± 0.104, p = 0.017), and IoU-score (0.561 ± 0.225 vs. 0.492 ± 0.186, p < 0.001) compared to the DeepLab V3 model, and a significantly lower average false positive rate of 3.6/patient vs. 7.0/patient (p < 0.001) using a 10 mm3 lesion-size limit. The ILD-model, trained on all possible combinations of four MRI sequences, may facilitate accurate detection and segmentation of brain metastases on a multicenter basis, even when the test cohort is missing input MRI sequences.

Manual segmentation of HGG is very time-consuming and using the SAM could increase the efficiency of this process. However, the accuracy of the SAM ultimately depends on the expert doing the editing. Our study confirmed a considerable inter-observer variability among experts defining tumor volume from volumetric MRIs.

No abstract

PurposeThis study aimed to investigate the hemodynamic status of cerebral metastases prior to and after stereotactic radiation surgery (SRS) and to identify the vascular characteristics that are associated with the development of pseudoprogression from radiation-induced damage with and without a radionecrotic component.Methods and materialsTwenty-four patients with 29 metastases from non-small cell lung cancer or malignant melanoma received SRS with dose of 15 Gy to 25 Gy. Magnetic resonance imaging (MRI) scans were acquired prior to SRS, every 3 months during the first year after SRS, and every 6 months thereafter. On the basis of the follow-up MRI scans or histology after SRS, metastases were classified as having response, tumor progression, or pseudoprogression. Advanced perfusion MRI enabled the estimation of vascular status in tumor regions including fractions of abnormal vessel architecture, underperfused tissue, and vessel pruning.ResultsPrior to SRS, metastases that later developed pseudoprogression had a distinct poor vascular function in the peritumoral zone compared with responding metastases (P < .05; number of metastases = 15). In addition, differences were found between the peritumoral zone of pseudoprogressing metastases and normal-appearing brain tissue (P < .05). In contrast, for responding metastases, no differences in vascular status between peritumoral and normal-appearing brain tissue were observed. The dysfunctional peritumoral vasculature persisted in pseudoprogressing metastases after SRS.ConclusionsOur results suggest that the vascular status of peritumoral tissue prior to SRS plays a defining role in the development of pseudoprogression and that advanced perfusion MRI may provide new insights into patients' susceptibility to radiation-induced effects.

Background MRI may provide insights into longitudinal responses in the diffusivity and vascular function of the irradiated normal-appearing brain following stereotactic radiosurgery (SRS) of brain metastases. Methods Forty patients with brain metastases from non-small cell lung cancer (N = 26) and malignant melanoma (N = 14) received SRS (15–25 Gy). Longitudinal MRI was performed pre-SRS and at 3, 6, 9, 12, and 18 months post-SRS. Measures of tissue diffusivity and vascularity were assessed by diffusion-weighted and perfusion MRI, respectively. All maps were normalized to white matter receiving less than 1 Gy. Longitudinal responses were assessed in normal-appearing brain, excluding tumor and edema, in the LowDose (1–10 Gy) and HighDose (&gt;10 Gy) regions. The Eastern Cooperative Oncology Group (ECOG) performance status was recorded pre-SRS. Results Following SRS, the diffusivity in the LowDose region increased continuously for 1 year (105.1% ± 6.2%; P &lt; .001), before reversing toward pre-SRS levels at 18 months. Transient reductions in microvascular cerebral blood volume (P &lt; .05), blood flow (P &lt; .05), and vessel densities (P &lt; .05) were observed in LowDose at 6–9 months post-SRS. Correspondingly, vessel calibers in LowDose transiently increased at 3–9 months (P &lt; .01). The responses in HighDose displayed similar trends as in LowDose, but with larger interpatient variations. Vascular responses followed pre-SRS ECOG status. Conclusions Our results imply that even low doses of radiation to normal-appearing brain following cerebral SRS induce increased diffusivity and reduced vascular function for up until 18 months. In particular, the vascular responses indicate the reduced ability of the normal-appearing brain tissue to form new capillaries. Assessing the potential long-term neurologic effects of SRS on the normal-appearing brain is warranted.

Dynamic susceptibility contrast (DSC) imaging is a widely used technique for assessment of cerebral blood volume (CBV). With combined gradient-echo and spin-echo DSC techniques, measures of the underlying vessel size and vessel architecture can be obtained from the vessel size index (VSI) and vortex area, respectively. However, how noise, and specifically the contrast-to-noise ratio (CNR), affect the estimations of these parameters has largely been overlooked. In order to address this issue, we have performed simulations to generate DSC signals with varying levels of CNR, defined by the peak of relaxation rate curve divided by the standard deviation of the baseline. Moreover, DSC data from 59 brain cancer patients were acquired at two different 3 T-scanners (N = 29 and N = 30, respectively), where CNR and relative parameter maps were obtained. Our simulations showed that the measured parameters were affected by CNR in different ways, where low CNR led to overestimations of CBV and underestimations of VSI and vortex area. In addition, a higher noise-sensitivity was found in vortex area than in CBV and VSI. Results from clinical data were consistent with simulations, and indicated that CNR < 4 gives highly unreliable measurements. Moreover, we have shown that the distribution of values in the tumour regions could change considerably when voxels with CNR below a given cut off are excluded when generating the relative parameter maps. The widespread use of CBV and attractive potential of VSI and vortex area, makes the noise-sensitivity of these parameters found in our study relevant for further use and development of the DSC imaging technique. Our results suggest that the CNR has considerable impact on the measured parameters, with the potential to affect the clinical interpretation of DSC-MRI, and should therefore be taken into account in the clinical decision-making process.

Purpose Magnetic resonance (MR) imaging is an essential diagnostic tool in clinical medicine. Recently, a variety of deep‐learning methods have been applied to segmentation tasks in medical images, with promising results for computer‐aided diagnosis. For MR images, effectively integrating different pulse sequences is important to optimize performance. However, the best way to integrate different pulse sequences remains unclear. In addition, networks trained with a certain subset of pulse sequences as input are unable to perform when given a subset of those pulse sequences. In this study, we evaluate multiple architectural features and characterize their effects in the task of metastasis segmentation while creating a method to robustly train a network to be able to work given any strict subset of the pulse sequences available during training. Methods We use a 2.5D DeepLabv3 segmentation network to segment metastases lesions on brain MR's with four pulse sequence inputs. To study how we can best integrate MR pulse sequences for this task, we consider (1) different pulse sequence integration schemas, combining our features at early, middle, and late points within a deep network, (2) different modes of weight sharing for parallel network branches, and (3) a novel integration level dropout layer, which will allow the networks to be robust to performing inference on input with only a subset of pulse sequences available at the training. Results We find that levels of integration and modes of weight sharing that favor low variance work best in our regime of small amounts of training data (n = 100). By adding an input‐level dropout layer, we could preserve the overall performance of these networks while allowing for inference on inputs with missing pulse sequences. We illustrate not only the generalizability of the network but also the utility of this robustness when applying the trained model to data from a different center, which does not use the same pulse sequences. Finally, we apply network visualization methods to better understand which input features are most important for network performance. Conclusions Together, these results provide a framework for building networks with enhanced robustness to missing data while maintaining comparable performance in medical imaging applications.