An Arbitrary Scale Super-Resolution Approach for 3D MR Images via Implicit Neural Representation.

High Resolution (HR) medical images provide rich anatomical structure details to facilitate early and accurate diagnosis. In magnetic resonance imaging (MRI), restricted by hardware capacity, scan time, and patient cooperation ability, isotropic 3-dimensional (3D) HR image acquisition typically requests long scan time and, results in small spatial coverage and low signal-to-noise ratio (SNR). Recent studies showed that, with deep convolutional neural networks, isotropic HR MR images could be recovered from low-resolution (LR) input via single image super-resolution (SISR) algorithms. However, most existing SISR methods tend to approach scale-specific projection between LR and HR images, thus these methods can only deal with fixed up-sampling rates. In this paper, we propose ArSSR, an Arbitrary Scale Super-Resolution approach for recovering 3D HR MR images. In the ArSSR model, the LR image and the HR image are represented using the same implicit neural voxel function with different sampling rates. Due to the continuity of the learned implicit function, a single ArSSR model is able to achieve arbitrary and infinite up-sampling rate reconstructions of HR images from any input LR image. Then the SR task is converted to approach the implicit voxel function via deep neural networks from a set of paired HR and LR training examples. The ArSSR model consists of an encoder network and a decoder network. Specifically, the convolutional encoder network is to extract feature maps from the LR input images and the fully-connected decoder network is to approximate the implicit voxel function. Experimental results on three datasets show that the ArSSR model can achieve state-of-the-art SR performance for 3D HR MR image reconstruction while using a single trained model to achieve arbitrary up-sampling scales.

Biochemically recurrent prostate cancer: rationalisation of the approach to imaging.

AIM: To assess the utility of magnetic resonance imaging (MRI) in addition to the additive benefit of the conventional imaging techniques, computed tomography (CT) and nuclear medicine (NM) bone scintigraphy, for investigation of biochemical recurrence (BCR) post-prostatectomy where access to prostate specific membrane antigen (PSMA) positron-emission tomography (PET)-CT is challenging. MATERIALS AND METHODS: Relevant imaging over a 5-year period was reviewed. Ethical approval was granted by the internal review board. All patients with suspected BCR, defined as a PSA ≥0.2 ng/ml on two separate occasions, underwent a retrospective imaging review. This was performed on PACS archive search database in a single centre using search terms "PSA" and "prostatectomy" in the three imaging methods; MRI, CT, and NM bone scintigraphy. All PSMA PET CT performed were recorded. RESULTS: One hundred and eighty-five patients were identified. Patients with an MRI pelvis that demonstrated distant metastases (i.e., pelvic bone metastases or lymph node involvement more cranial to the bifurcation of the common iliac arteries) were more likely to have a positive CT and/or NM bone scintigraphy. The Pearson correlation coefficient between the findings of M1 disease at MRI pelvis and the presence of distant metastases at CT thorax, abdomen, pelvis and NM bone scintigraphy was calculated at 0.81 (p<0.01) and 0.91 (p<0.01) respectively. CONCLUSION: An imaging strategy based on risk stratification and technique-specific selection criteria leads to more appropriate use of resources, and in turn, increases the yield of conventional imaging methods. MRI prostate findings can be used to predict the additive value of CT/NM bone scintigraphy allowing a more streamlined approach to their use.

A new classification and regression tree algorithm: Improved diagnostic sensitivity for HCC ≤ 3.0 cm using gadoxetate disodium-enhanced MRI.

PURPOSE: To develop and validate an effective algorithm, based on classification and regression tree (CART) analysis and LI-RADS features, for diagnosing HCC ≤ 3.0 cm with gadoxetate disodium­enhanced MRI (Gd-EOB-MRI). METHOD: We retrospectively included 299 and 90 high-risk patients with hepatic lesions ≤ 3.0 cm that underwent Gd-EOB-MRI from January 2018 to February 2021 in institution 1 (development cohort) and institution 2 (validation cohort), respectively. Through binary and multivariate regression analyses of LI-RADS features in the development cohort, we developed an algorithm using CART analysis, which comprised the targeted appearance and independently significant imaging features. On per-lesion basis, we compared the diagnostic performances of our algorithm, two previously reported CART algorithms, and LI-RADS LR-5 in development and validation cohorts. RESULTS: Our CART algorithm, presenting as a decision tree, included targetoid appearance, HBP hypointensity, nonrim arterial phase hyperenhancement (APHE), and transitional phase hypointensity plus mild-moderate T2 hyperintensity. For definite HCC diagnosis, the overall sensitivity of our algorithm (development cohort 93.2%, validation cohort 92.5%; P < 0.006) was significantly higher than those of Jiang's algorithm modified LR-5 (defined as targetoid appearance, nonperipheral washout, restricted diffusion, and nonrim APHE) and LI-RADS LR-5, with the comparable specificity (development cohort: 84.3%, validation cohort: 86.7%; P ≥ 0.006). Our algorithm, providing the highest balanced accuracy (development cohort: 91.2%, validation cohort: 91.6%), outperformed other criteria for identifying HCCs from non-HCC lesions. CONCLUSIONS: In high-risk patients, our CART algorithm developed with LI-RADS features showed promise for the early diagnosis of HCC ≤ 3.0 cm with Gd-EOB-MRI.

Cardiovascular magnetic resonance for evaluation of cardiac involvement in COVID-19: recommendations by the Society for Cardiovascular Magnetic Resonance.

Coronavirus disease 2019 (COVID-19) is an ongoing global pandemic that has affected nearly 600 million people to date across the world. While COVID-19 is primarily a respiratory illness, cardiac injury is also known to occur. Cardiovascular magnetic resonance (CMR) imaging is uniquely capable of characterizing myocardial tissue properties in-vivo, enabling insights into the pattern and degree of cardiac injury. The reported prevalence of myocardial involvement identified by CMR in the context of COVID-19 infection among previously hospitalized patients ranges from 26 to 60%. Variations in the reported prevalence of myocardial involvement may result from differing patient populations (e.g. differences in severity of illness) and the varying intervals between acute infection and CMR evaluation. Standardized methodologies in image acquisition, analysis, interpretation, and reporting of CMR abnormalities across would likely improve concordance between studies. This consensus document by the Society for Cardiovascular Magnetic Resonance (SCMR) provides recommendations on CMR imaging and reporting metrics towards the goal of improved standardization and uniform data acquisition and analytic approaches when performing CMR in patients with COVID-19 infection.

Liver fat quantification in photon counting CT in head to head comparison with clinical MRI - First experience.

PURPOSE: To compare liver fat quantification between MRI and photon-counting CT (PCCT). METHOD: A cylindrical phantom with inserts containing six concentrations of oil (0, 10, 20, 30, 50 and 100%) and oil-iodine mixtures (0, 10, 20, 30 and 50% fat +3 mg/mL iodine) was imaged with a PCCT (NAEOTOM Alpha) and a 1.5 T MRI system (MR 450w, IDEAL-IQ sequence), using clinical parameters. An IRB-approved prospective clinical evaluation included 12 obese adult patients with known fatty liver disease (seven women, mean age: 61.5 ± 13 years, mean BMI: 30.3 ± 4.7 kg/m2). Patients underwent a same-day clinical MRI and PCCT of the abdomen. Liver fat fractions were calculated for four segments (I, II, IVa and VII) using in- and opposed-phase on MRI ((Meanin - Meanopp)/2*Meanin) and iodine-fat, tissue decomposition analysis in PCCT (Syngo.Via VB60A). CT and MRI Fat fractions were compared using two-sample t-tests with equal variance. Statistical analysis was performed using RStudio (Version1.4.1717). RESULTS: Phantom results showed no significant differences between the known fat fractions (P = 0.32) or iodine (P = 0.6) in comparison to PCCT-measured concentrations, and no statistically significant difference between known and MRI-measured fat fractions (P = 0.363). In patients, the mean fat signal fraction measured on MRI and PCCT was 13.1 ± 9.9% and 12.0 ± 9.0%, respectively, with an average difference of 1.1 ± 1.9% between the modalities (P = 0.138). CONCLUSION: First experience shows promising accuracy of liver fat fraction quantification for PCCT in obese patients. This method may improve opportunistic screening for CT in the future.

Quality control and removal of technical variation of NMR metabolic biomarker data in ~120,000 UK Biobank participants.

Metabolic biomarker data quantified by nuclear magnetic resonance (NMR) spectroscopy in approximately 121,000 UK Biobank participants has recently been released as a community resource, comprising absolute concentrations and ratios of 249 circulating metabolites, lipids, and lipoprotein sub-fractions. Here we identify and characterise additional sources of unwanted technical variation influencing individual biomarkers in the data available to download from UK Biobank. These included sample preparation time, shipping plate well, spectrometer batch effects, drift over time within spectrometer, and outlier shipping plates. We developed a procedure for removing this unwanted technical variation, and demonstrate that it increases signal for genetic and epidemiological studies of the NMR metabolic biomarker data in UK Biobank. We subsequently developed an R package, ukbnmr, which we make available to the wider research community to enhance the utility of the UK Biobank NMR metabolic biomarker data and to facilitate rapid analysis.

Preoperative Assessment of Cervical Involvement in Endometrial Cancer by Transvaginal Ultrasound and Magnetic Resonance Imaging: A Systematic Review and Meta-Analysis.

OBJECTIVE: To compare the diagnostic accuracy of transvaginal ultrasound (TVS) and magnetic resonance imaging (MRI) for detecting cervical infiltration by endometrial carcinoma using meta-analysis assessment. METHODS: An extensive search of papers comparing TVS and MRI for assessing cervical infiltration in endometrial cancer in the same set of patients was performed in Medline (Pubmed), Web of Science, and the Cochrane Database. Quality was assessed using QUADAS-2 tool (Quality Assessment of Diagnostic Accuracy Studies-2). Quantitative meta-analysis was performed. RESULTS: Our extended search identified 12 articles that used both techniques in the same set of patients and were included in the meta-analysis. The risk of bias for most studies was high for patient selection and index tests in QUADAS-2. Overall, the pooled estimated sensitivity and specificity for diagnosing cervical infiltration in women with endometrial cancer were identical for both techniques [69 % (95 % CI, 51 %-82 %) and 93 % (95 % CI, 90 %-95 %) for TVS, and 69 % (95 % CI, 57 %-79 %) and 91 % (95 % CI, 90 %-95 %) for MRI, respectively]. No statistical differences were found when comparing both methods. Heterogeneity was high for sensitivity and moderate for specificity when analyzing TVS and moderate for both sensitivity and specificity in the case of MRI. CONCLUSION: TVS and MRI showed very similar diagnostic performance for diagnosing cervical involvement in women with endometrial cancer.

Rapid brain MRI protocols result in comparable differential diagnoses versus a full brain protocol in most canine and feline cases.

Evaluation of brain disease in veterinary patients uses a wide variety of MRI sequences. A shortened protocol that maintains consistency of interpretation would reduce radiologist reporting time, patient anesthetic time, and client cost. The aims of this retrospective, methods comparison, observer agreement study were to evaluate whether abbreviated MRI protocols alter differential diagnoses and recommendations compared to our institution's standard protocol; evaluate interobserver agreement on standard brain MRIs; and assess whether differential diagnoses change after postcontrast images. Normal and pathologic canine and feline brain MRIs were retrieved from hospital archives. Three protocols were created from each: a 5-sequence noncontrast enhanced Fast Brain Protocol 1 (FBP1); a 6-sequence contrast-enhanced Fast Brain Protocol 2 (FBP2); and an 11-sequence standard brain protocol (SBP). Three blinded veterinary radiologists interpreted FBP images for 98 cases (1 reader/case) and SBP images for 20 cases (3 readers/case). A fourth observer compared these interpretations to the original MRI reports (OMR). Overall agreement between FBPs and OMR was good (k = 0.75) and comparable to interobserver agreement for multiple reviews of SBP cases. Postcontrast images substantially altered conclusions in 17/97 cases (17.5%), as well as improved interobserver agreement compared to noncontrast studies. The conclusions reached with shortened brain protocols were comparable to those of a full brain study. The findings supported the use of a 6-sequence brain MRI protocol (sagittal T2-weighted [T2w] TSE; transverse T2w turbo spin echo fluid-attenuated inversion recovery, T2*-weighted gradient recalled echo, T1-weighted spin echo, and diffusion weighted imaging/apparent diffusion coefficient; and postcontrast transverse T1-weighted spin echo) for dogs and cats with suspected intracranial disease.

ASL-BIDS, the brain imaging data structure extension for arterial spin labeling.

Arterial spin labeling (ASL) is a non-invasive MRI technique that allows for quantitative measurement of cerebral perfusion. Incomplete or inaccurate reporting of acquisition parameters complicates quantification, analysis, and sharing of ASL data, particularly for studies across multiple sites, platforms, and ASL methods. There is a strong need for standardization of ASL data storage, including acquisition metadata. Recently, ASL-BIDS, the BIDS extension for ASL, was developed and released in BIDS 1.5.0. This manuscript provides an overview of the development and design choices of this first ASL-BIDS extension, which is mainly aimed at clinical ASL applications. Discussed are the structure of the ASL data, focussing on storage order of the ASL time series and implementation of calibration approaches, unit scaling, ASL-related BIDS fields, and storage of the labeling plane information. Additionally, an overview of ASL-BIDS compatible conversion and ASL analysis software and ASL example datasets in BIDS format is provided. We anticipate that large-scale adoption of ASL-BIDS will improve the reproducibility of ASL research.

Impact of MRI on decision-making in ICU patients with disorders of consciousness.

BACKGROUND: Recovery of consciousness is the most important survival factor in patients with acute brain injury and disorders of consciousness (DoC). Since most deaths in the intensive care unit (ICU) occur after withdrawal of life-support, medical decision-making is crucial for acute DoC patients. Neuroimaging informs decision-making, yet the precise effects of MRI on decision-making in the ICU are poorly understood. We investigated the impact of brain MRI on prognostication, therapeutic decisions and physician confidence in ICU patients with DoC. METHODS: In this simulated decision-making study utilizing a prospective ICU cohort, a panel of neurocritical experts first reviewed clinical information (without MRI) from 75 acute DoC patients and made decisions about diagnosis, prognosis and treatment. Following review of the MRI, the panel then decided if the initial decisions needed revision. In parallel, a blinded neuroradiologist reassessed all neuroimaging. RESULTS: MRI led to changes in clinical management of 57 (76%) of patients (Number-Needed-to-Test for any change: 1.32), including revised diagnoses (20%), levels of care (21%), diagnostic confidence (43%) and prognostications (33%). Decisions were revised more often with stroke than with other brain injuries (p = 0.02). However, although MRI revealed additional pathology in 81%, this did not predict revised clinical decision-making (p-values ≥0.08). CONCLUSION: MRI results changed decision-making in 3 of 4 ICU patients, but radiological findings were not predictive of clinical decision-making. This highlights the need to better understand the effects of neuroimaging on management decisions. How MRI influences decision-making in the ICU is an important avenue for research to improve acute DoC management.

Attention-guided deep learning for gestational age prediction using fetal brain MRI.

Magnetic resonance imaging offers unrivaled visualization of the fetal brain, forming the basis for establishing age-specific morphologic milestones. However, gauging age-appropriate neural development remains a difficult task due to the constantly changing appearance of the fetal brain, variable image quality, and frequent motion artifacts. Here we present an end-to-end, attention-guided deep learning model that predicts gestational age with R2 score of 0.945, mean absolute error of 6.7 days, and concordance correlation coefficient of 0.970. The convolutional neural network was trained on a heterogeneous dataset of 741 developmentally normal fetal brain images ranging from 19 to 39 weeks in gestational age. We also demonstrate model performance and generalizability using independent datasets from four academic institutions across the U.S. and Turkey with R2 scores of 0.81-0.90 after minimal fine-tuning. The proposed regression algorithm provides an automated machine-enabled tool with the potential to better characterize in utero neurodevelopment and guide real-time gestational age estimation after the first trimester.

Accurate brain-age models for routine clinical MRI examinations.

Convolutional neural networks (CNN) can accurately predict chronological age in healthy individuals from structural MRI brain scans. Potentially, these models could be applied during routine clinical examinations to detect deviations from healthy ageing, including early-stage neurodegeneration. This could have important implications for patient care, drug development, and optimising MRI data collection. However, existing brain-age models are typically optimised for scans which are not part of routine examinations (e.g., volumetric T1-weighted scans), generalise poorly (e.g., to data from different scanner vendors and hospitals etc.), or rely on computationally expensive pre-processing steps which limit real-time clinical utility. Here, we sought to develop a brain-age framework suitable for use during routine clinical head MRI examinations. Using a deep learning-based neuroradiology report classifier, we generated a dataset of 23,302 'radiologically normal for age' head MRI examinations from two large UK hospitals for model training and testing (age range = 18-95 years), and demonstrate fast (< 5 s), accurate (mean absolute error [MAE] < 4 years) age prediction from clinical-grade, minimally processed axial T2-weighted and axial diffusion-weighted scans, with generalisability between hospitals and scanner vendors (Δ MAE < 1 year). The clinical relevance of these brain-age predictions was tested using 228 patients whose MRIs were reported independently by neuroradiologists as showing atrophy 'excessive for age'. These patients had systematically higher brain-predicted age than chronological age (mean predicted age difference = +5.89 years, 'radiologically normal for age' mean predicted age difference = +0.05 years, p < 0.0001). Our brain-age framework demonstrates feasibility for use as a screening tool during routine hospital examinations to automatically detect older-appearing brains in real-time, with relevance for clinical decision-making and optimising patient pathways.

Spatial Bayesian GLM on the cortical surface produces reliable task activations in individuals and groups.

The general linear model (GLM) is a widely popular and convenient tool for estimating the functional brain response and identifying areas of significant activation during a task or stimulus. However, the classical GLM is based on a massive univariate approach that does not explicitly leverage the similarity of activation patterns among neighboring brain locations. As a result, it tends to produce noisy estimates and be underpowered to detect significant activations, particularly in individual subjects and small groups. A recently proposed alternative, a cortical surface-based spatial Bayesian GLM, leverages spatial dependencies among neighboring cortical vertices to produce more accurate estimates and areas of functional activation. The spatial Bayesian GLM can be applied to individual and group-level analysis. In this study, we assess the reliability and power of individual and group-average measures of task activation produced via the surface-based spatial Bayesian GLM. We analyze motor task data from 45 subjects in the Human Connectome Project (HCP) and HCP Retest datasets. We also extend the model to multi-run analysis and employ subject-specific cortical surfaces rather than surfaces inflated to a sphere for more accurate distance-based modeling. Results show that the surface-based spatial Bayesian GLM produces highly reliable activations in individual subjects and is powerful enough to detect trait-like functional topologies. Additionally, spatial Bayesian modeling enhances reliability of group-level analysis even in moderately sized samples (n=45). Notably, the power of the spatial Bayesian GLM to detect activations above a scientifically meaningful effect size is nearly invariant to sample size, exhibiting high power even in small samples (n=10). The spatial Bayesian GLM is computationally efficient in individuals and groups and is convenient to implement with the open-source BayesfMRI R package.

Advancing motion denoising of multiband resting-state functional connectivity fMRI data.

Simultaneous multi-slice (multiband) accelerated functional magnetic resonance imaging (fMRI) provides dramatically improved temporal and spatial resolution for resting-state functional connectivity (RSFC) studies of the human brain in health and disease. However, multiband acceleration also poses unique challenges for denoising of subject motion induced data artifacts, the presence of which is a major confound in RSFC research that substantively diminishes reliability and reproducibility. We comprehensively evaluated existing and novel approaches to volume censoring-based motion denoising in the Human Connectome Project (HCP) dataset. We show that assumptions underlying common metrics for evaluating motion denoising pipelines, especially those based on quality control-functional connectivity (QC-FC) correlations and differences between high- and low-motion participants, are problematic, and appear to be inappropriate in their current widespread use as indicators of comparative pipeline performance and as targets for investigators to use when tuning pipelines for their own datasets. We further develop two new quantitative metrics that are instead agnostic to QC-FC correlations and other measures that rely upon the null assumption that no true relationships exist between trait measures of subject motion and functional connectivity, and demonstrate their use as benchmarks for comparing volume censoring methods. Finally, we develop and validate quantitative methods for determining dataset-specific optimal volume censoring parameters prior to the final analysis of a dataset, and provide straightforward recommendations and code for all investigators to apply this optimized approach to their own RSFC datasets.

Isolating the sources of pipeline-variability in group-level task-fMRI results.

Task-fMRI researchers have great flexibility as to how they analyze their data, with multiple methodological options to choose from at each stage of the analysis workflow. While the development of tools and techniques has broadened our horizons for comprehending the complexities of the human brain, a growing body of research has highlighted the pitfalls of such methodological plurality. In a recent study, we found that the choice of software package used to run the analysis pipeline can have a considerable impact on the final group-level results of a task-fMRI investigation (Bowring et al., 2019, BMN). Here we revisit our work, seeking to identify the stages of the pipeline where the greatest variation between analysis software is induced. We carry out further analyses on the three datasets evaluated in BMN, employing a common processing strategy across parts of the analysis workflow and then utilizing procedures from three software packages (AFNI, FSL, and SPM) across the remaining steps of the pipeline. We use quantitative methods to compare the statistical maps and isolate the main stages of the workflow where the three packages diverge. Across all datasets, we find that variation between the packages' results is largely attributable to a handful of individual analysis stages, and that these sources of variability were heterogeneous across the datasets (e.g., choice of first-level signal model had the most impact for the balloon analog risk task dataset, while first-level noise model and group-level model were more influential for the false belief and antisaccade task datasets, respectively). We also observe areas of the analysis workflow where changing the software package causes minimal differences in the final results, finding that the group-level results were largely unaffected by which software package was used to model the low-frequency fMRI drifts.

White matter hyperintensities segmentation using an ensemble of neural networks.

White matter hyperintensities (WMHs) represent the most common neuroimaging marker of cerebral small vessel disease (CSVD). The volume and location of WMHs are important clinical measures. We present a pipeline using deep fully convolutional network and ensemble models, combining U-Net, SE-Net, and multi-scale features, to automatically segment WMHs and estimate their volumes and locations. We evaluated our method in two datasets: a clinical routine dataset comprising 60 patients (selected from Chinese National Stroke Registry, CNSR) and a research dataset composed of 60 patients (selected from MICCAI WMH Challenge, MWC). The performance of our pipeline was compared with four freely available methods: LGA, LPA, UBO detector, and U-Net, in terms of a variety of metrics. Additionally, to access the model generalization ability, another research dataset comprising 40 patients (from Older Australian Twins Study and Sydney Memory and Aging Study, OSM), was selected and tested. The pipeline achieved the best performance in both research dataset and the clinical routine dataset with DSC being significantly higher than other methods (p < .001), reaching .833 and .783, respectively. The results of model generalization ability showed that the model trained on the research dataset (DSC = 0.736) performed higher than that trained on the clinical dataset (DSC = 0.622). Our method outperformed widely used pipelines in WMHs segmentation. This system could generate both image and text outputs for whole brain, lobar and anatomical automatic labeling WMHs. Additionally, software and models of our method are made publicly available at https://www.nitrc.org/projects/what_v1.

The bright side and dark side of daydreaming predict creativity together through brain functional connectivity.

Daydreaming and creativity have similar cognitive processes and neural basis. However, few empirical studies have examined the relationship between daydreaming and creativity using cognitive neuroscience methods. The present study explored the relationship between different types of daydreaming and creativity and their common neural basis. The behavioral results revealed that positive constructive daydreaming is positively related to creativity, while poor attentional control is negatively related to it. Machine learning framework was adopted to examine the predictive effect of daydreaming-related brain functional connectivity (FC) on creativity. The results demonstrated that task FCs related to positive constructive daydreaming and task FCs related to poor attentional control both predicted an individual's creativity score successfully. In addition, task FCs combining the positive constructive daydreaming and poor attentional control also had significant predictive effect on creativity score. Furthermore, predictive analysis based on resting-state FCs showed similar patterns. Both of the subscale-related FCs and combined FCs had significant predictive effect on creativity score. Further analysis showed the task and the resting-state FCs both mainly located in the default mode network, central executive network, salience network, and attention network. These results showed that daydreaming was closely related to creativity, as they shared common FC basis.

Reliability of structural MRI measurements: The effects of scan session, head tilt, inter-scan interval, acquisition sequence, FreeSurfer version and processing stream.

BACKGROUND: Large-scale longitudinal and multi-centre studies are used to explore neuroimaging markers of normal ageing, and neurodegenerative and mental health disorders. Longitudinal changes in brain structure are typically small, therefore the reliability of automated techniques is crucial. Determining the effects of different factors on reliability allows investigators to control those adversely affecting reliability, calculate statistical power, or even avoid particular brain measures with low reliability. This study examined the impact of several image acquisition and processing factors and documented the test-retest reliability of structural MRI measurements. METHODS: In Phase I, 20 healthy adults (11 females; aged 20-30 years) were scanned on two occasions three weeks apart on the same scanner using the ADNI-3 protocol. On each occasion, individuals were scanned twice (repetition), after re-entering the scanner (reposition) and after tilting their head forward. At one year follow-up, nine returning individuals and 11 new volunteers were recruited for Phase II (11 females; aged 22-31 years). Scans were acquired on two different scanners using the ADNI-2 and ADNI-3 protocols. Structural images were processed using FreeSurfer (v5.3.0, 6.0.0 and 7.1.0) to provide subcortical and cortical volume, cortical surface area and thickness measurements. Intra-class correlation coefficients (ICC) were calculated to estimate test-retest reliability. We examined the effect of repetition, reposition, head tilt, time between scans, MRI sequence and scanner on reliability of structural brain measurements. Mean percentage differences were also calculated in supplementary analyses. RESULTS: Using the FreeSurfer v7.1.0 longitudinal pipeline, we observed high reliability for subcortical and cortical volumes, and cortical surface areas at repetition, reposition, three weeks and one year (mean ICCs>0.97). Cortical thickness reliability was lower (mean ICCs>0.82). Head tilt had the greatest adverse impact on ICC estimates, for example reducing mean right cortical thickness to ICC=0.74. In contrast, changes in ADNI sequence or MRI scanner had a minimal effect. We observed an increase in reliability for updated FreeSurfer versions, with the longitudinal pipeline consistently having a higher reliability than the cross-sectional pipeline. DISCUSSION: Longitudinal studies should monitor or control head tilt to maximise reliability. We provided the ICC estimates and mean percentage differences for all FreeSurfer brain regions, which may inform power analyses for clinical studies and have implications for the design of future longitudinal studies.