Atypical imaging features of multiple sclerosis lesions include size >2 cm, mass effect, oedema and/or ring enhancement. This constellation is often referred to as ‘tumefactive multiple sclerosis’. Previous series emphasize their unifocal and clinically isolated nature, however, evolution of these lesions is not well defined. Biopsy may be required for diagnosis. We describe clinical and radiographic features in 168 patients with biopsy confirmed CNS inflammatory demyelinating disease (IDD). Lesions were analysed on pre- and post-biopsy magnetic resonance imaging (MRI) for location, size, mass effect/oedema, enhancement, multifocality and fulfilment of Barkhof criteria. Clinical data were correlated to MRI. Female to male ratio was 1.2 : 1, median age at onset, 37 years, duration between symptom onset and biopsy, 7.1 weeks and total disease duration, 3.9 years. Clinical course prior to biopsy was a first neurological event in 61%, relapsing–remitting in 29% and progressive in 4%. Presentations were typically polysymptomatic, with motor, cognitive and sensory symptoms predominating. Aphasia, agnosia, seizures and visual field defects were observed. At follow-up, 70% developed definite multiple sclerosis, and 14% had an isolated demyelinating syndrome. Median time to second attack was 4.8 years, and median EDSS at follow-up was 3.0. Multiple lesions were present in 70% on pre-biopsy MRI, and in 83% by last MRI, with Barkhof criteria fulfilled in 46% prior to biopsy and 55% by follow-up. Only 17% of cases remained unifocal. Median largest lesion size on T2-weighted images was 4 cm (range 0.5–12), with a discernible size of 2.1 cm (range 0.5–7.5). Biopsied lesions demonstrated mass effect in 45% and oedema in 77%. A strong association was found between lesion size, and presence of mass effect and/or oedema (P < 0.001). Ring enhancement was frequent. Most tumefactive features did not correlate with gender, course or diagnosis. Although lesion size >5 cm was associated with a slightly higher EDSS at last follow-up, long-term prognosis in patients with disease duration >10 years was better (EDSS 1.5) compared with a population-based multiple sclerosis cohort matched for disease duration (EDSS 3.5; P < 0.001). Given the retrospective nature of the study, the precise reason for biopsy could not always be determined. This study underscores the diagnostically challenging nature of CNS IDDs that present with atypical clinical or radiographic features. Most have multifocal disease at onset, and develop RRMS by follow-up. Although increased awareness of this broad spectrum may obviate need for biopsy in many circumstances, an important role for diagnostic brain biopsy may be required in some cases.

The rate of renal disease progression varies widely among patients with autosomal dominant polycystic kidney disease (ADPKD), necessitating optimal patient selection for enrollment into clinical trials. Patients from the Mayo Clinic Translational PKD Center with ADPKD (n=590) with computed tomography/magnetic resonance images and three or more eGFR measurements over $6 months were classified radiologically as typical (n=538) or atypical (n=52). Total kidney volume (TKV) was measured using stereology (TKVs) and ellipsoid equation (TKVe). Typical patients were randomly partitioned into development and internal validation sets and subclassified according to height-adjusted TKV (HtTKV) ranges for age (1A-1E, in increasing order). Consortium for Radiologic Imaging Study of PKD (CRISP) participants (n=173) were used for external validation. TKVe correlated strongly with TKVs, without systematic underestimation or overestimation. A longitudinal mixed regression model to predict eGFR decline showed that log 2 HtTKV and age significantly interacted with time in typical patients, but not in atypical patients. When 1A-1E classifications were used instead of log 2 HtTKV, eGFR slopes were significantly different among subclasses and, except for 1A, different from those in healthy kidney donors. The equation derived from the development set predicted eGFR in both validation sets. The frequency of ESRD at 10 years increased from subclass 1A (2.4%) to 1E (66.9%) in the Mayo cohort and from 1C (2.2%) to 1E (22.3%) in the younger CRISP cohort. Class and subclass designations were stable. An easily applied classification of ADPKD based on HtTKV and age should optimize patient selection for enrollment into clinical trials and for treatment when one becomes available.

Machine learning is a technique for recognizing patterns that can be applied to medical images. Although it is a powerful tool that can help in rendering medical diagnoses, it can be misapplied. Machine learning typically begins with the machine learning algorithm system computing the image features that are believed to be of importance in making the prediction or diagnosis of interest. The machine learning algorithm system then identifies the best combination of these image features for classifying the image or computing some metric for the given image region. There are several methods that can be used, each with different strengths and weaknesses. There are open-source versions of most of these machine learning methods that make them easy to try and apply to images. Several metrics for measuring the performance of an algorithm exist; however, one must be aware of the possible associated pitfalls that can result in misleading metrics. More recently, deep learning has started to be used; this method has the benefit that it does not require image feature identification and calculation as a first step; rather, features are identified as part of the learning process. Machine learning has been used in medical imaging and will have a greater influence in the future. Those working in medical imaging must be aware of how machine learning works.

Quantitative analysis of brain MRI is routine for many neurological diseases and conditions and relies on accurate segmentation of structures of interest. Deep learning-based segmentation approaches for brain MRI are gaining interest due to their self-learning and generalization ability over large amounts of data. As the deep learning architectures are becoming more mature, they gradually outperform previous state-of-the-art classical machine learning algorithms. This review aims to provide an overview of current deep learning-based segmentation approaches for quantitative brain MRI. First we review the current deep learning architectures used for segmentation of anatomical brain structures and brain lesions. Next, the performance, speed, and properties of deep learning approaches are summarized and discussed. Finally, we provide a critical assessment of the current state and identify likely future developments and trends.

The identification of pathological processes that could be targeted by therapeutic interventions is a major goal of research into multiple sclerosis (MS). Pathological assessment is the gold standard for such identification, but has intrinsic limitations owing to the limited availability of autopsy and biopsy tissue. MRI has gained a leading role in the assessment of MS because it allows doctors to obtain an ante mortem picture of the degree of CNS involvement. A number of correlative pathological and MRI studies have helped to define in vivo the pathological substrates of MS in focal lesions and normal-appearing white matter, not only in the brain, but also in the spinal cord. These studies have resulted in the identification of aspects of pathophysiology that were previously neglected, including grey matter involvement and vascular pathology. Despite these important achievements, numerous open questions still need to be addressed to resolve controversies about how the pathology of MS results in fixed neurological disability.

IMPORTANCEPer the World Health Organization 2016 integrative classification, newly diagnosed glioblastomas are separated into isocitrate dehydrogenase gene 1 or 2 (IDH)-wild-type and IDH-mutant subtypes, with median patient survival of 1.2 and 3.6 years, respectively. Although maximal resection of contrast-enhanced (CE) tumor is associated with longer survival, the prognostic importance of maximal resection within molecular subgroups and the potential importance of resection of non-contrast-enhanced (NCE) disease is poorly understood.OBJECTIVE To assess the association of resection of CE and NCE tumors in conjunction with molecular and clinical information to develop a new road map for cytoreductive surgery.

A recent joint meeting was held on January 30, 2014, with the US Food and Drug Administration (FDA), National Cancer Institute (NCI), clinical scientists, imaging experts, pharmaceutical and biotech companies, clinical trials cooperative groups, and patient advocate groups to discuss imaging endpoints for clinical trials in glioblastoma. This workshop developed a set of priorities and action items including the creation of a standardized MRI protocol for multicenter studies. The current document outlines consensus recommendations for a standardized Brain Tumor Imaging Protocol (BTIP), along with the scientific and practical justifications for these recommendations, resulting from a series of discussions between various experts involved in aspects of neuro-oncology neuroimaging for clinical trials. The minimum recommended sequences include: (i) parameter-matched precontrast and postcontrast inversion recovery-prepared, isotropic 3D T1-weighted gradient-recalled echo; (ii) axial 2D T2-weighted turbo spin-echo acquired after contrast injection and before postcontrast 3D T1-weighted images to control timing of images after contrast administration; (iii) precontrast, axial 2D T2-weighted fluid-attenuated inversion recovery; and (iv) precontrast, axial 2D, 3-directional diffusion-weighted images. Recommended ranges of sequence parameters are provided for both 1.5 T and 3 T MR systems.

rtificial intelligence (AI) algorithms have existed for decades and have recently been propelled to the forefront of medical imaging research. To a large extent, this is related to improvements in computing power, availability of a large amount of training data, and innovative and improved neural network architectures, with the recognition that certain types of algorithms are well suited to image analysis. The latter discovery was accelerated by the ImageNet competition and represents a fundamental transformation in research mechanics and methods in computer vision.Currently, in most studies, researchers collect data, perform analysis, and publish results. The same researchers may continue to augment and expand the data set and perform subsequent analysis with resulting publications. The data for each study are held quite closely and are rarely shared among institutions outside of multicenter trials. Competitions represent a different model of research: Research data are made available to the public, usually with a baseline performance metric. Groups around the world are invited to analyze the data and create algorithms to beat the performance of the prior generation. For example, the baseline performance metric for this challenge was set by the previous skeletal age model developed by Larson et al (1).The Radiological Society of North America (RSNA) Pediatric Bone Age Machine Learning Challenge was created to evaluate the performance of computer algorithms in executing a common image analysis activity that is familiar to many pediatric radiologists: estimating the bone age of pediatric patients based on radiographs of their hand (1-5). This challenge used a data set of pediatric