The purpose of this work was to assess the reproducibility of diffusion imaging, and in particular the apparent diffusion coefficient (ADC), intra-voxel incoherent motion (IVIM) parameters and diffusion tensor imaging (DTI) parameters, across multiple centres using clinically available protocols with limited harmonization between sequences.An ice–water phantom and nine healthy volunteers were scanned across fives centres on eight scanners (four Siemens 1.5T, four Philips 3T). The mean ADC, IVIM parameters (diffusion coefficient D and perfusion fraction f) and DTI parameters (mean diffusivity MD and fractional anisotropy FA), were measured in grey matter, white matter and specific brain sub-regions. A mixed effect model was used to measure the intra- and inter-scanner coefficient of variation (CV) for each of the five parameters.ADC, D, MD and FA had a good intra- and inter-scanner reproducibility in both grey and white matter, with a CV ranging between 1% and 7.4%; mean 2.6%. Other brain regions also showed high levels of reproducibility except for small structures such as the choroid plexus. The IVIM parameter f had a higher intra-scanner CV of 8.4% and inter-scanner CV of 24.8%. No major difference in the inter-scanner CV for ADC, D, MD and FA was observed when analysing the 1.5T and 3T scanners separately.ADC, D, MD and FA all showed good intra-scanner reproducibility, with the inter-scanner reproducibility being comparable or faring slightly worse, suggesting that using data from multiple scanners does not have an adverse effect compared with using data from the same scanner. The IVIM parameter f had a poorer inter-scanner CV when scanners of different field strengths were combined, and the parameter was also affected by the scan acquisition resolution. This study shows that the majority of diffusion MRI derived parameters are robust across 1.5T and 3T scanners and suitable for use in multi-centre clinical studies and trials. © 2015 The Authors NMR in Biomedicine Published by John Wiley & Sons Ltd.

Accurate and reliable intraoperative neuronavigation is crucial for achieving maximal safe resection of brain tumors. Intraoperative MRI (iMRI) has received significant attention as the next step in improving navigation. However, the immense cost and logistical challenge of iMRI precludes implementation in most centers worldwide. In comparison, intraoperative ultrasound (ioUS) is an affordable tool, easily incorporated into existing theatre infrastructure, and operative workflow. Historically, ultrasound has been perceived as difficult to learn and standardize, with poor, artifact-prone image quality. However, ioUS has dramatically evolved over the last decade, with vast improvements in image quality and well-integrated navigation tools. Advanced techniques, such as contrast-enhanced ultrasound (CEUS), have also matured and moved from the research field into actual clinical use. In this review, we provide a comprehensive and pragmatic guide to ioUS. A suggested protocol to facilitate learning ioUS and improve standardization is provided, and an outline of common artifacts and methods to minimize them given. The review also includes an update of advanced techniques and how they can be incorporated into clinical practice.

Background Variations in prognosis and treatment options for gliomas are dependent on tumour grading. When tissue is available for analysis, grade is established based on histological criteria. However, histopathological diagnosis is not always reliable or straight-forward due to tumour heterogeneity, sampling error and subjectivity, and hence there is great inter-observer variability in readings. Methods We trained convolutional neural network models to classify digital whole-slide histopathology images from The Cancer Genome Atlas. We tested a number of optimisation parameters. Results Data augmentation did not improve model training, while smaller batch size helped to prevent overfitting and led to improved model performance. There was no significant difference in performance between a modular 2-class model and a single 3-class model system. The best models trained achieved a mean accuracy of 73% in classifying glioblastoma from other grades, and 53% between WHO grade II and III gliomas. A visualisation method was developed to convey the model output in a clinically relevant manner by overlaying colour-coded predictions over the original whole slide image. Conclusions Our developed visualisation method reflects the clinical decision-making process by highlighting the intra-tumour heterogeneity and may be used in clinical setting to aid diagnosis. Explainable AI techniques may allow further evaluation of the model and underline areas for improvements such as biases. Due to intra-tumour heterogeneity, data annotation for training was imprecise, and hence performance was lower than expected. The models may be further improved by employing advanced data augmentation strategies and using more precise semi-automatic or manually labelled training data.