Search

Segmentation of white matter lesions and deep grey matter structures is an important task in the quantification of magnetic resonance imaging in multiple sclerosis. In this paper we explore segmentation solutions based on convolutional neural networks (CNNs) for providing fast, reliable segmentations of lesions and grey-matter structures in multi-modal MR imaging, and the performance of these methods when applied to out-of-centre data. We trained two state-of-the-art fully convolutional CNN architectures on the 2016 MSSEG training dataset, which was annotated by seven independent human raters: a reference implementation of a 3D Unet, and a more recently proposed 3D-to-2D architecture (DeepSCAN). We then retrained those methods on a larger dataset from a single centre, with and without labels for other brain structures. We quantified changes in performance owing to dataset shift, and changes in performance by adding the additional brain-structure labels. We also compared performance with freely available reference methods. Both fully-convolutional CNN methods substantially outperform other approaches in the literature when trained and evaluated in cross-validation on the MSSEG dataset, showing agreement with human raters in the range of human inter-rater variability. Both architectures showed drops in performance when trained on single-centre data and tested on the MSSEG dataset. When trained with the addition of weak anatomical labels derived from Freesurfer, the performance of the 3D Unet degraded, while the performance of the DeepSCAN net improved. Overall, the DeepSCAN network predicting both lesion and anatomical labels was the best-performing network examined.

Background: Migraine with aura (MwA) in pediatric patients is clinically frequent. Clinically complex symptoms need to be differentiated to exclude mimicking conditions., Purpose: We hypothesize that MwA in children induces abnormalities readily visible in perfusion time to peak (TTP) maps as well as non-enhanced susceptibility weighted magnetic resonance imaging (SWI)., Materials and Methods: Between 2010 and 2018, we retrospectively evaluated symptoms and imaging of consecutive pediatric patients <18 years with MwA. We visually scored abnormalities on SWI and TTP maps in 12 regions of interest on both hemispheres on three axial slices, as normal, slightly, distinctly or severely abnormal., Results: 99 patients (69.7% female), mean age 14.07 y (±2.8) were included. Focally increased deoxygenation (FID) in SWI was present in 61.6%. FID on SWI was dominant for the left hemisphere (60.7% vs. 31.1%, (p < .001)), and in 8.2% symmetric. Side of aura symptoms and contralateral hemispheric imaging alterations in patients with FID correlated significantly (p = .002.). 61 of 99 patients had perfusion MR and 59% of these patients showed focal increase of TTP. Age correlated significantly with FID in SWI (r = -.248, p = .013) and increase of TTP in perfusion (r = -.252, p = .05). Focal abnormalities correlated significantly between SWI and TTP maps. Brain regions most often abnormal were the temporal superior, occipital and fronto-parietal regions., Conclusions: This study provides confidence in recognizing FID, and linking FID in SWI to acute MwA in pediatric patients. FID phenomenon had a left hemispheric significant dominance, and can be found bilaterally., Competing Interests: Declaration of competing interest Roland Wiest: UNRELATED: Grants/Grants Pending: Swiss National Foundation, Swiss Heart Foundation., (Copyright © 2020 European Paediatric Neurology Society. Published by Elsevier Ltd. All rights reserved.)

The detection of new or enlarged white-matter lesions is a vital task in the monitoring of patients undergoing disease-modifying treatment for multiple sclerosis. However, the definition of 'new or enlarged' is not fixed, and it is known that lesion-counting is highly subjective, with high degree of inter- and intra-rater variability. Automated methods for lesion quantification, if accurate enough, hold the potential to make the detection of new and enlarged lesions consistent and repeatable. However, the majority of lesion segmentation algorithms are not evaluated for their ability to separate radiologically progressive from radiologically stable patients, despite this being a pressing clinical use-case. In this paper, we explore the ability of a deep learning segmentation classifier to separate stable from progressive patients by lesion volume and lesion count, and find that neither measure provides a good separation. Instead, we propose a method for identifying lesion changes of high certainty, and establish on an internal dataset of longitudinal multiple sclerosis cases that this method is able to separate progressive from stable time-points with a very high level of discrimination (AUC = 0.999), while changes in lesion volume are much less able to perform this separation (AUC = 0.71). Validation of the method on two external datasets confirms that the method is able to generalize beyond the setting in which it was trained, achieving an accuracies of 75 % and 85 % in separating stable and progressive time-points., (Copyright © 2019. Published by Elsevier Inc.)

Importance: Various signs may be observed on brain magnetic resonance imaging (MRI) in patients with spontaneous intracranial hypotension (SIH). However, the lack of a classification system integrating these findings limits decision making in clinical practice., Objective: To develop a probability score based on the most relevant brain MRI findings to assess the likelihood of an underlying spinal cerebrospinal fluid (CSF) leak in patients with SIH., Design, Setting, and Participants: This case-control study in consecutive patients investigated for SIH was conducted at a single hospital department from February 2013 to October 2017. Patients with missing brain MRI data were excluded. Three blinded readers retrospectively reviewed the brain MRI scans of patients with SIH and a spinal CSF leak, patients with orthostatic headache without a CSF leak, and healthy control participants, evaluating 9 quantitative and 7 qualitative signs. A predictive diagnostic score based on multivariable backward logistic regression analysis was then derived. Its performance was validated internally in a prospective cohort of patients who had clinical suspicion for SIH., Main Outcomes and Measures: Likelihood of a spinal CSF leak based on the proposed diagnostic score., Results: A total of 152 participants (101 female [66.4%]; mean [SD] age, 46.1 [14.3] years) were studied. These included 56 with SIH and a spinal CSF leak, 16 with orthostatic headache without a CSF leak, 60 control participants, and 20 patients in the validation cohort. Six imaging findings were included in the final scoring system. Three were weighted as major (2 points each): pachymeningeal enhancement, engorgement of venous sinus, and effacement of the suprasellar cistern of 4.0 mm or less. Three were considered minor (1 point each): subdural fluid collection, effacement of the prepontine cistern of 5.0 mm or less, and mamillopontine distance of 6.5 mm or less. Patients were classified into groups at low, intermediate, or high probability of having a spinal CSF leak, with total scores of 2 points or fewer, 3 to 4 points, and 5 points or more, respectively, on a scale of 9 points. The discriminatory ability of the proposed score could be demonstrated in the validation cohort., Conclusions and Relevance: This 3-tier predictive scoring system is based on the 6 most relevant brain MRI findings and allows assessment of the likelihood (low, intermediate, or high) of a positive spinal imaging result in patients with SIH. It may be useful in identifying patients with SIH who are leak positive and in whom further invasive myelographic examinations are warranted before considering targeted therapy.