LAUSR

The greatest novelties in the research presentation sessions (RPS) on neuroimaging at the European Congress of Radiology (ECR) 2020 are related to progress in advanced magnetic resonance imaging (MRI) techniques and artificial intelligence (AI). But, as is the case every year at ECR, advances in the fields of stroke, brain tumours, neurodegenerative disease, and multiple sclerosis (MS) are included in the programme too. All abstracts are published in the ECR 2020 Book of Abstracts [1], with session and presentation numbers mentioned in brackets. In brain tumour imaging, the recently revised WHO classification has resulted in an increasing use of deep learning in the study of new biomarkers. The central focus is the methods in isocitrate dehydrogenase (IDH) genotype prediction and the presence or absence of O-methylguanine-methyltransferase (MGMT) in glioblastoma. In a preliminary study of 25 patients, classifiers provided natural boundaries for efficient discrimination of IDH-wild type and IDH-mutation (RPS 1011a4). The usefulness of rCBV measurements to differentiate the IDH genotype is also reported (RPS 1011a-7). The proteomic footprint of IDH-mutation is the presence of 2-hydroxy-glutarate (2HG). The authors studied 39 patients and found 80% sensitivity and 75% specificity of single voxel MR spectroscopy in the detection of 2HG (RPS 1011a-5). The study of the topographical distribution of MGMT promoter methylation in 436 IDH-wild type glioblastoma revealed no statistically significant differences between methylated and unmethylated status (RPS 1011a-6). There is an increasing use of radiomics for volumetry and for assessing response to treatment in brain tumour imaging. In a series of 124 patients, it was shown that it was possible to predict patient survival by analysing radiomics features quantifying tumour intensity and heterogeneity (RPS 1011b-3). In stroke imaging, several abstracts focus on improvement in stroke patient selection for thrombolysis. One study reports that the mothership and drip-and-ship strategies do not differ significantly in impact on outcome, provided that ‘bridging’ intravenous thrombolysis is given to the acute stroke patient in the drip-and-ship strategy (RPS 1411b-1). Automated attenuation measurements on CT angiography source images present excellent performance in detecting ischaemia in acute stroke patients and bear the potential to reduce the use of CT perfusion (RPS 1411b-7). AI can provide an e-CT angiography collateral score for patients eligible for thrombectomy (RPS 1411b-8). Automated assessment of regional hypoattenuations was able to identify expert-classified visual ASPECTS cut-off values and may be helpful for patient selection for thrombectomy (RPS 1411b-9). Dynamic CT angiography weighting (arterial/arteriovenous/venous) did not impact collateral grade analysis and was not useful in predicting infarct size and clinical outcome (RPS 1111-5). High-resolution intracranial vessel wall imaging has been used to study the atherosclerotic plaque features and differences were observed in patients with a good and poor leptomeningeal collateral status which may help determine therapeutic decision-making (RPS 1111-6). Another focus in stroke research deals with postthrombectomy imaging. The post-thrombectomy CT perfusion parameter time-topeak was an independent predictor of treatment failure and CT perfusion parameters improve the treatment outcome prediction in acute stroke patients beyond clinical/angiographic parameters, ASPECTS, and pre-treatment perfusion parameters (RPS 1411b-2). The AURORAmeta-analysis provides data on 458 patients who underwent thrombectomy more than 6 hours after last * Philippe Demaerel [email protected]