Search

Adolescents exhibit remarkable heterogeneity in the structural architecture of brain development. However, due to limited large-scale longitudinal neuroimaging studies, existing research has largely focused on population averages, and the neurobiological basis underlying individual heterogeneity remains poorly understood. Here we identify, using the IMAGEN adolescent cohort followed up over 9 years (14–23 y), three groups of adolescents characterized by distinct developmental patterns of whole-brain gray matter volume (GMV). Group 1 show continuously decreasing GMV associated with higher neurocognitive performances than the other two groups during adolescence. Group 2 exhibit a slower rate of GMV decrease and lower neurocognitive performances compared with Group 1, which was associated with epigenetic differences and greater environmental burden. Group 3 show increasing GMV and lower baseline neurocognitive performances due to a genetic variation. Using the UK Biobank, we show these differences may be attenuated in mid-to-late adulthood. Our study reveals clusters of adolescent neurodevelopment based on GMV and the potential long-term impact. Longitudinal analysis of neuroimaging data are useful for analysing heterogeneity in adolescent brain development. Here the authors cluster adolescent participants of the IMAGEN study into groups based on gray matter volume developmental patterns and investigate genome-wide and epigenome-wide associations with these groups. [ABSTRACT FROM AUTHOR]

Objective: To report imaging protocol and scheduling variance in routine care of glioblastoma patients in order to demonstrate challenges of integrating deep‐learning models in glioblastoma care pathways. Additionally, to understand the most common imaging studies and image contrasts to inform the development of potentially robust deep‐learning models. Methods: MR imaging data were analysed from a random sample of five patients from the prospective cohort across five participating sites of the ZGBM consortium. Reported clinical and treatment data alongside DICOM header information were analysed to understand treatment pathway imaging schedules. Results: All sites perform all structural imaging at every stage in the pathway except for the presurgical study, where in some sites only contrast‐enhanced T1‐weighted imaging is performed. Diffusion MRI is the most common non‐structural imaging type, performed at every site. Conclusion: The imaging protocol and scheduling varies across the UK, making it challenging to develop machine‐learning models that could perform robustly at other centres. Structural imaging is performed most consistently across all centres. Advances in knowledge: Successful translation of deep‐learning models will likely be based on structural post‐treatment imaging unless there is significant effort made to standardise non‐structural or peri‐operative imaging protocols and schedules. [ABSTRACT FROM AUTHOR]

• Delays in radiological reporting can result in poorer patient outcomes and inflated healthcare costs. To reduce reporting times for abnormal scans, we present a deep learning framework for detecting clinically-relevant abnormalities in minimally processed, hospital-grade axial T2-weighted and axial diffusion-weighted head MRI scans. • The models were trained at scale using a Transformer-based neuroradiology report classifier to generate a labelled dataset of > 70,000 examinations from two large UK hospital networks, and demonstrate fast, accurate, and interpretable classification, with good generalisability between hospitals. • Through a simulation study we show that our best model would considerably reduce the mean reporting time for abnormal examinations at the two hospital networks, demonstrating feasibility for use in a clinical triage environment. The growing demand for head magnetic resonance imaging (MRI) examinations, along with a global shortage of radiologists, has led to an increase in the time taken to report head MRI scans in recent years. For many neurological conditions, this delay can result in poorer patient outcomes and inflated healthcare costs. Potentially, computer vision models could help reduce reporting times for abnormal examinations by flagging abnormalities at the time of imaging, allowing radiology departments to prioritise limited resources into reporting these scans first. To date, however, the difficulty of obtaining large, clinically-representative labelled datasets has been a bottleneck to model development. In this work, we present a deep learning framework, based on convolutional neural networks, for detecting clinically-relevant abnormalities in minimally processed, hospital-grade axial T2-weighted and axial diffusion-weighted head MRI scans. The models were trained at scale using a Transformer-based neuroradiology report classifier to generate a labelled dataset of 70,206 examinations from two large UK hospital networks, and demonstrate fast (< 5 s), accurate (area under the receiver operating characteristic curve (AUC) > 0.9), and interpretable classification, with good generalisability between hospitals (ΔAUC ≤ 0.02). Through a simulation study we show that our best model would reduce the mean reporting time for abnormal examinations from 28 days to 14 days and from 9 days to 5 days at the two hospital networks, demonstrating feasibility for use in a clinical triage environment. [Display omitted] [ABSTRACT FROM AUTHOR]

Background: Multi-center magnetic resonance imaging (MRI) studies present an opportunity to advance research by pooling data. However, brain measurements derived from MR-images are susceptible to differences in MR-sequence parameters. It is therefore necessary to determine whether there is an interaction between the sequence parameters and the effect of interest, and to minimise any such interaction by careful choice of acquisition parameters. As an exemplar of the issues involved in multi-center studies, we present data from a study in which we aimed to optimize a set of volumetric MRI-protocols to define a protocol giving data that are consistent and reproducible across two centers and over time., Methods: Optimization was achieved based on data quality and quantitative measures, in our case using FreeSurfer and Voxel Based Morphometry approaches. Our approach consisted of a series of five comparisons. Firstly, a single-center dataset was collected, using a range of candidate pulse-sequences and parameters chosen on the basis of previous literature. Based on initial results, a number of minor changes were implemented to optimize the pulse-sequences, and a second single-center dataset was collected. FreeSurfer data quality measures were compared between datasets in order to determine the best performing sequence(s), which were taken forward to the next stage of testing. We subsequently acquired short-term and long-term two-center reproducibility data, and quantitative measures were again assessed to determine the protocol with the highest reproducibility across centers. Effects of a scanner software and hardware upgrade on the reproducibility of the protocols at one of the centers were also evaluated., Results: Assessing the quality measures from the first two datasets allowed us to define artefact-free protocols, all with high image quality as assessed by FreeSurfer. Comparing the quantitative test and retest measures, we found high within-center reproducibility for all protocols, but lower between-center reproducibility for some protocols than others. The upgrade showed no important effects., Conclusions: We were able to determine (for the scanners used in this study) an optimised protocol, which gave the highest within- and between-center reproducibility of those assessed, and give details of this protocol here. More generally, we discuss some of the issues raised by multi-center studies and describe a methodical approach to take towards optimization and standardization, and recommend performing this kind of procedure to other investigators.