Automated model-based tissue classification of MR images of the brain

Baseline habitual dietary nitrate intake and Alzheimer's Disease related neuroimaging biomarkers in the Australian Imaging, Biomarkers and Lifestyle study of ageing

BACKGROUND

Dietary nitrate, as a nitric oxide (NO) precursor, may support brain health and protect against dementia.

OBJECTIVE

Our primary aim was to investigate whether dietary nitrate is associated with neuroimaging markers of brain health linked with Alzheimer's disease (AD).

PARTICIPANTS

Study participants were cognitively unimpaired individuals from the Australian Imaging, Biomarkers and Lifestyle Study of Ageing (AIBL) who had β-amyloid positron emission tomography (PET) scans (n = 554) and magnetic resonance imaging (MRI) scans (n = 335) and had completed a Food Frequency Questionnaire at baseline.

METHODS

Source-specific nitrate intakes were estimated using comprehensive nitrate food composition databases. Rates of cerebral β-amyloid (Aβ) deposition, measured using PET, and rates of brain atrophy, measured using MRI, were assessed between baseline and 126-months follow-up, at intervals of 18 months. Multivariable-adjusted linear mixed effect models were used to examine associations between baseline source-specific nitrate intake and rates of (i) cerebral Aβ deposition and (ii) brain atrophy, over the 126 months of follow-up. Analyses were carried out following stratification of the sample by established dementia Alzheimer's disease (AD) risk factors including sex and presence or absence of the apolipoprotein E (APOE) ε4 allele.

RESULTS

In women carriers of the APOE ε4 allele, higher plant sourced nitrate intake (median intake 121 mg/day), was associated with a slower rate of cerebral Aβ deposition [β: 4.47 versus 8.99 Centiloid (CL) /18 months, p < 0.05] and right hippocampal atrophy [-0.01 versus -0.03 mm3 /18 months, p < 0.01], after multivariable adjustments. Moderate intake showed protective associations in men carriers and in both men and women non-carriers of APOE ε4.

CONCLUSIONS

Associations were observed between plant-derived nitrate intake and cerebral Aβ deposition, particularly in high-risk populations (women and APOE ε4 carriers). Associations were also observed for brain volume atrophy, however these exhibited subgroup variability without clear patterns relative to sex and APOE ε4 allele carriage. These findings suggest a potential link between plant-sourced nitrate and AD related neuroimaging markers of brain health improved brain health, but further validation in larger studies is required.

Myelination across cortical hierarchies and depths in humans and macaques

Myelination is fundamental to brain function, enabling rapid neural communication and supporting neuroplasticity throughout the lifespan. While hierarchical patterns of myelin maturation across the cortical surface are well-documented in humans, it remains unclear which features reflect evolutionarily conserved developmental processes versus human-characteristic adaptations. Moreover, the laminar development of myelin across the primate cortical surface, which shapes hierarchies and supports functions ranging from sensory integration to network communication, has been largely unexplored. Using neuroimaging to measure the T1-weighted/T2-weighted ratio in tissue contrast as a proxy for myelin content, we systematically compared depth-dependent trajectories of myelination across the cortical surface in humans and macaques. We identified a conserved "inside-out" pattern, with deeper layers exhibiting steeper increases in myelination and earlier plateaus than superficial layers. This depth-dependent organization followed a hierarchical gradient across the cortical surface, progressing from early maturation in sensorimotor regions to prolonged development in association areas. Humans exhibited a markedly extended timeline of myelination across both cortical regions and depths compared to macaques, allowing for prolonged postnatal plasticity across the entire cortical hierarchy - from sensory and motor processing to higher-order association networks. This extended potential for plasticity may facilitate the shaping of cortical circuits through postnatal experience in ways that support human-characteristic perceptual and cognitive capabilities.

Histology-guided MRI segmentation of brainstem nuclei critical to consciousness

While substantial progress has been made in mapping the connectivity of cortical networks responsible for conscious awareness, neuroimaging analysis of subcortical arousal networks that modulate arousal (i.e., wakefulness) has been limited by a lack of a robust segmentation procedures for brainstem arousal nuclei. Automated segmentation of brainstem arousal nuclei is an essential step toward elucidating the physiology of arousal in human consciousness and the pathophysiology of disorders of consciousness. We created a probabilistic atlas of brainstem arousal nuclei built on diffusion MRI scans of five ex vivo human brain specimens scanned at 750 μm isotropic resolution. Labels of arousal nuclei used to generate the probabilistic atlas were manually annotated with reference to nucleus-specific immunostaining in two of the five brain specimens. We then developed a Bayesian segmentation algorithm that utilizes the probabilistic atlas as a generative model and automatically identifies brainstem arousal nuclei in a resolution- and contrast-agnostic manner. The segmentation method displayed high accuracy in both healthy and lesioned in vivo T1 MRI scans and high test-retest reliability across both T1 and T2 MRI contrasts. Finally, we show that the segmentation algorithm can detect volumetric changes and differences in magnetic susceptibility within brainstem arousal nuclei in Alzheimer's disease and traumatic coma, respectively. We release the probabilistic atlas and Bayesian segmentation tool in FreeSurfer to advance the study of human consciousness and its disorders.

A next-generation, histological atlas of the human brain and its application to automated brain MRI segmentation

Magnetic resonance imaging (MRI) is the standard tool to image the human brain in vivo. In this domain, digital brain atlases are essential for subject-specific segmentation of anatomical regions of interest (ROIs) and spatial comparison of neuroanatomy from different subjects in a common coordinate frame. High-resolution, digital atlases derived from histology (e.g., Allen atlas [7], BigBrain [13], Julich [15]), are currently the state of the art and provide exquisite 3D cytoarchitectural maps, but lack probabilistic labels throughout the whole brain. Here we present NextBrain, a next-generation probabilistic atlas of human brain anatomy built from serial 3D histology and corresponding highly granular delineations of five whole brain hemispheres. We developed AI techniques to align and reconstruct ~10,000 histological sections into coherent 3D volumes with joint geometric constraints (no overlap or gaps between sections), as well as to semi-automatically trace the boundaries of 333 distinct anatomical ROIs on all these sections. Comprehensive delineation on multiple cases enabled us to build the first probabilistic histological atlas of the whole human brain. Further, we created a companion Bayesian tool for automated segmentation of the 333 ROIs in any in vivo or ex vivo brain MRI scan using the NextBrain atlas. We showcase two applications of the atlas: automated segmentation of ultra-high-resolution ex vivo MRI and volumetric analysis of Alzheimer's disease and healthy brain ageing based on ~4,000 publicly available in vivo MRI scans. We publicly release: the raw and aligned data (including an online visualisation tool); the probabilistic atlas; the segmentation tool; and ground truth delineations for a 100 μm isotropic ex vivo hemisphere (that we use for quantitative evaluation of our segmentation method in this paper). By enabling researchers worldwide to analyse brain MRI scans at a superior level of granularity without manual effort or highly specific neuroanatomical knowledge, NextBrain holds promise to increase the specificity of MRI findings and ultimately accelerate our quest to understand the human brain in health and disease.

Medical image segmentation approach based on hybrid adaptive differential evolution and crayfish optimizer

Image segmentation plays a pivotal role in medical image analysis, particularly for accurately isolating tumors and lesions. Effective segmentation improves diagnostic precision and facilitates quantitative analysis, which is vital for medical professionals. However, traditional segmentation methods often struggle with multilevel thresholding due to the associated computational complexity. Therefore, determining the optimal threshold set is an NP-hard problem, highlighting the pressing need for efficient optimization strategies to overcome these challenges. This paper introduces a multi-threshold image segmentation (MTIS) method that integrates a hybrid approach combining Differential Evolution (DE) and the Crayfish Optimization Algorithm (COA), known as HADECO. Utilizing two-dimensional (2D) Kapur's entropy and a 2D histogram, this method aims to enhance the efficiency and accuracy of subsequent image analysis and diagnosis. HADECO is a hybrid algorithm that combines DE and COA by exchanging information based on predefined rules, leveraging the strengths of both for superior optimization results. It employs Latin Hypercube Sampling (LHS) to generate a high-quality initial population. HADECO introduces an improved DE algorithm (IDE) with adaptive and dynamic adjustments to key DE parameters and new mutation strategies to enhance its search capability. In addition, it incorporates an adaptive COA (ACOA) with dynamic adjustments to the switching probability parameter, effectively balancing exploration and exploitation. To evaluate the effectiveness of HADECO, its performance is initially assessed using CEC'22 benchmark functions. HADECO is evaluated against several contemporary algorithms using the Wilcoxon signed rank test (WSRT) and the Friedman test (FT) to integrate the results. The findings highlight HADECO's superior optimization abilities, demonstrated by its lowest average Friedman ranking of 1.08. Furthermore, the HADECO-based MTIS method is evaluated using MRI images for knee and CT scans for brain intracranial hemorrhage (ICH). Quantitative results in brain hemorrhage image segmentation show that the proposed method achieves a superior average peak signal-to-noise ratio (PSNR) and feature similarity index (FSIM) of 1.5 and 1.7 at the 6-level threshold. In knee image segmentation, it attains an average PSNR and FSIM of 1.3 and 1.2 at the 5-level threshold, demonstrating the method's effectiveness in solving image segmentation problems.

deepbet: Fast brain extraction of T1-weighted MRI using Convolutional Neural Networks

BACKGROUND

Brain extraction in magnetic resonance imaging (MRI) data is an important segmentation step in many neuroimaging preprocessing pipelines. Image segmentation is one of the research fields in which deep learning had the biggest impact in recent years. Consequently, traditional brain extraction methods are now being replaced by deep learning-based methods.

METHOD

Here, we used a unique dataset compilation comprising 7837 T1-weighted (T1w) MR images from 191 different OpenNeuro datasets in combination with advanced deep learning methods to build a fast, high-precision brain extraction tool called deepbet.

RESULTS

deepbet sets a novel state-of-the-art performance during cross-dataset validation with a median Dice score (DSC) of 99.0 on unseen datasets, outperforming the current best performing deep learning (DSC=97.9) and classic (DSC=96.5) methods. While current methods are more sensitive to outliers, deepbet achieves a Dice score of >97.4 across all 7837 images from 191 different datasets. This robustness was additionally tested in 5 external datasets, which included challenging clinical MR images. During visual exploration of each method's output which resulted in the lowest Dice score, major errors could be found for all of the tested tools except deepbet. Finally, deepbet uses a compute efficient variant of the UNet architecture, which accelerates brain extraction by a factor of ≈10 compared to current methods, enabling the processing of one image in ≈2 s on low level hardware.

CONCLUSIONS

In conclusion, deepbet demonstrates superior performance and reliability in brain extraction across a wide range of T1w MR images of adults, outperforming existing top tools. Its high minimal Dice score and minimal objective errors, even in challenging conditions, validate deepbet as a highly dependable tool for accurate brain extraction. deepbet can be conveniently installed via "pip install deepbet" and is publicly accessible at https://github.com/wwu-mmll/deepbet.

Cross-sectional associations between 24-hour time-use composition, grey matter volume and cognitive function in healthy older adults

BACKGROUND

Increasing physical activity (PA) is an effective strategy to slow reductions in cortical volume and maintain cognitive function in older adulthood. However, PA does not exist in isolation, but coexists with sleep and sedentary behaviour to make up the 24-hour day. We investigated how the balance of all three behaviours (24-hour time-use composition) is associated with grey matter volume in healthy older adults, and whether grey matter volume influences the relationship between 24-hour time-use composition and cognitive function.

METHODS

This cross-sectional study included 378 older adults (65.6 ± 3.0 years old, 123 male) from the ACTIVate study across two Australian sites (Adelaide and Newcastle). Time-use composition was captured using 7-day accelerometry, and T1-weighted magnetic resonance imaging was used to measure grey matter volume both globally and across regions of interest (ROI: frontal lobe, temporal lobe, hippocampi, and lateral ventricles). Pairwise correlations were used to explore univariate associations between time-use variables, grey matter volumes and cognitive outcomes. Compositional data analysis linear regression models were used to quantify associations between ROI volumes and time-use composition, and explore potential associations between the interaction between ROI volumes and time-use composition with cognitive outcomes.

RESULTS

After adjusting for covariates (age, sex, education), there were no significant associations between time-use composition and any volumetric outcomes. There were significant interactions between time-use composition and frontal lobe volume for long-term memory (p = 0.018) and executive function (p = 0.018), and between time-use composition and total grey matter volume for executive function (p = 0.028). Spending more time in moderate-vigorous PA was associated with better long-term memory scores, but only for those with smaller frontal lobe volume (below the sample mean). Conversely, spending more time in sleep and less time in sedentary behaviour was associated with better executive function in those with smaller total grey matter volume.

CONCLUSIONS

Although 24-hour time use was not associated with total or regional grey matter independently, total grey matter and frontal lobe grey matter volume moderated the relationship between time-use composition and several cognitive outcomes. Future studies should investigate these relationships longitudinally to assess whether changes in time-use composition correspond to changes in grey matter volume and cognition.

An efficient semi-supervised quality control system trained using physics-based MRI-artefact generators and adversarial training

Large medical imaging data sets are becoming increasingly available. A common challenge in these data sets is to ensure that each sample meets minimum quality requirements devoid of significant artefacts. Despite a wide range of existing automatic methods having been developed to identify imperfections and artefacts in medical imaging, they mostly rely on data-hungry methods. In particular, the scarcity of artefact-containing scans available for training has been a major obstacle in the development and implementation of machine learning in clinical research. To tackle this problem, we propose a novel framework having four main components: (1) a set of artefact generators inspired by magnetic resonance physics to corrupt brain MRI scans and augment a training dataset, (2) a set of abstract and engineered features to represent images compactly, (3) a feature selection process that depends on the class of artefact to improve classification performance, and (4) a set of Support Vector Machine (SVM) classifiers trained to identify artefacts. Our novel contributions are threefold: first, we use the novel physics-based artefact generators to generate synthetic brain MRI scans with controlled artefacts as a data augmentation technique. This will avoid the labour-intensive collection and labelling process of scans with rare artefacts. Second, we propose a large pool of abstract and engineered image features developed to identify 9 different artefacts for structural MRI. Finally, we use an artefact-based feature selection block that, for each class of artefacts, finds the set of features that provide the best classification performance. We performed validation experiments on a large data set of scans with artificially-generated artefacts, and in a multiple sclerosis clinical trial where real artefacts were identified by experts, showing that the proposed pipeline outperforms traditional methods. In particular, our data augmentation increases performance by up to 12.5 percentage points on the accuracy, F1, F2, precision and recall. At the same time, the computation cost of our pipeline remains low - less than a second to process a single scan - with the potential for real-time deployment. Our artefact simulators obtained using adversarial learning enable the training of a quality control system for brain MRI that otherwise would have required a much larger number of scans in both supervised and unsupervised settings. We believe that systems for quality control will enable a wide range of high-throughput clinical applications based on the use of automatic image-processing pipelines.

A fully convolutional neural network for comprehensive compartmentalization of abdominal adipose tissue compartments in MRI

BACKGROUND

Existing literature has highlighted structural, physiological, and pathological disparities among abdominal adipose tissue (AAT) sub-depots. Accurate separation and quantification of these sub-depots are crucial for advancing our understanding of obesity and its comorbidities. However, the absence of clear boundaries between the sub-depots in medical imaging data has challenged their separation, particularly for internal adipose tissue (IAT) sub-depots. To date, the quantification of AAT sub-depots remains challenging, marked by a time-consuming, costly, and complex process.

PURPOSE

To implement and evaluate a convolutional neural network to enable granular assessment of AAT by compartmentalization of subcutaneous adipose tissue (SAT) into superficial subcutaneous (SSAT) and deep subcutaneous (DSAT) adipose tissue, and IAT into intraperitoneal (IPAT), retroperitoneal (RPAT), and paraspinal (PSAT) adipose tissue.

MATERIAL AND METHODS

MRI datasets were retrospectively collected from Singapore Preconception Study for Long-Term Maternal and Child Outcomes (S-PRESTO: 389 women aged 31.4 ± 3.9 years) and Singapore Adult Metabolism Study (SAMS: 50 men aged 28.7 ± 5.7 years). For all datasets, ground truth segmentation masks were created through manual segmentation. A Res-Net based 3D-UNet was trained and evaluated via 5-fold cross-validation on S-PRESTO data (N = 300). The model's final performance was assessed on a hold-out (N = 89) and an external test set (N = 50, SAMS).

RESULTS

The proposed method enabled reliable segmentation of individual AAT sub-depots in 3D MRI volumes with high mean Dice similarity scores of 98.3%, 97.2%, 96.5%, 96.3%, and 95.9% for SSAT, DSAT, IPAT, RPAT, and PSAT respectively.

CONCLUSION

Convolutional neural networks can accurately sub-divide abdominal SAT into SSAT and DSAT, and abdominal IAT into IPAT, RPAT, and PSAT with high accuracy. The presented method has the potential to significantly contribute to advancements in the field of obesity imaging and precision medicine.

FastSurfer-HypVINN: Automated sub-segmentation of the hypothalamus and adjacent structures on high-resolutional brain MRI

The hypothalamus plays a crucial role in the regulation of a broad range of physiological, behavioral, and cognitive functions. However, despite its importance, only a few small-scale neuroimaging studies have investigated its substructures, likely due to the lack of fully automated segmentation tools to address scalability and reproducibility issues of manual segmentation. While the only previous attempt to automatically sub-segment the hypothalamus with a neural network showed promise for 1.0 mm isotropic T1-weighted (T1w) magnetic resonance imaging (MRI), there is a need for an automated tool to sub-segment also high-resolutional (HiRes) MR scans, as they are becoming widely available, and include structural detail also from multi-modal MRI. We, therefore, introduce a novel, fast, and fully automated deep-learning method named HypVINN for sub-segmentation of the hypothalamus and adjacent structures on 0.8 mm isotropic T1w and T2w brain MR images that is robust to missing modalities. We extensively validate our model with respect to segmentation accuracy, generalizability, in-session test-retest reliability, and sensitivity to replicate hypothalamic volume effects (e.g., sex differences). The proposed method exhibits high segmentation performance both for standalone T1w images as well as for T1w/T2w image pairs. Even with the additional capability to accept flexible inputs, our model matches or exceeds the performance of state-of-the-art methods with fixed inputs. We, further, demonstrate the generalizability of our method in experiments with 1.0 mm MR scans from both the Rhineland Study and the UK Biobank-an independent dataset never encountered during training with different acquisition parameters and demographics. Finally, HypVINN can perform the segmentation in less than a minute (graphical processing unit [GPU]) and will be available in the open source FastSurfer neuroimaging software suite, offering a validated, efficient, and scalable solution for evaluating imaging-derived phenotypes of the hypothalamus.

ROOD-MRI: Benchmarking the robustness of deep learning segmentation models to out-of-distribution and corrupted data in MRI

Deep artificial neural networks (DNNs) have moved to the forefront of medical image analysis due to their success in classification, segmentation, and detection challenges. A principal challenge in large-scale deployment of DNNs in neuroimage analysis is the potential for shifts in signal-to-noise ratio, contrast, resolution, and presence of artifacts from site to site due to variances in scanners and acquisition protocols. DNNs are famously susceptible to these distribution shifts in computer vision. Currently, there are no benchmarking platforms or frameworks to assess the robustness of new and existing models to specific distribution shifts in MRI, and accessible multi-site benchmarking datasets are still scarce or task-specific. To address these limitations, we propose ROOD-MRI: a novel platform for benchmarking the Robustness of DNNs to Out-Of-Distribution (OOD) data, corruptions, and artifacts in MRI. This flexible platform provides modules for generating benchmarking datasets using transforms that model distribution shifts in MRI, implementations of newly derived benchmarking metrics for image segmentation, and examples for using the methodology with new models and tasks. We apply our methodology to hippocampus, ventricle, and white matter hyperintensity segmentation in several large studies, providing the hippocampus dataset as a publicly available benchmark. By evaluating modern DNNs on these datasets, we demonstrate that they are highly susceptible to distribution shifts and corruptions in MRI. We show that while data augmentation strategies can substantially improve robustness to OOD data for anatomical segmentation tasks, modern DNNs using augmentation still lack robustness in more challenging lesion-based segmentation tasks. We finally benchmark U-Nets and vision transformers, finding robustness susceptibility to particular classes of transforms across architectures. The presented open-source platform enables generating new benchmarking datasets and comparing across models to study model design that results in improved robustness to OOD data and corruptions in MRI.

Improved interpretation of 18F-florzolotau PET in progressive supranuclear palsy using a normalization-free deep-learning classifier

While 18F-florzolotau tau PET is an emerging biomarker for progressive supranuclear palsy (PSP), its interpretation has been hindered by a lack of consensus on visual reading and potential biases in conventional semi-quantitative analysis. As clinical manifestations and regions of elevated 18F-florzolotau binding are highly overlapping in PSP and the Parkinsonian type of multiple system atrophy (MSA-P), developing a reliable discriminative classifier for 18F-florzolotau PET is urgently needed. Herein, we developed a normalization-free deep-learning (NFDL) model for 18F-florzolotau PET, which achieved significantly higher accuracy for both PSP and MSA-P compared to semi-quantitative classifiers. Regions driving the NFDL classifier's decision were consistent with disease-specific topographies. NFDL-guided radiomic features correlated with clinical severity of PSP. This suggests that the NFDL model has the potential for early and accurate differentiation of atypical parkinsonism and that it can be applied in various scenarios due to not requiring subjective interpretation, MR-dependent, and reference-based preprocessing.

Deep learning based automatic segmentation of organs-at-risk for 0.35 T MRgRT of lung tumors

BACKGROUND AND PURPOSE

Magnetic resonance imaging guided radiotherapy (MRgRT) offers treatment plan adaptation to the anatomy of the day. In the current MRgRT workflow, this requires the time consuming and repetitive task of manual delineation of organs-at-risk (OARs), which is also prone to inter- and intra-observer variability. Therefore, deep learning autosegmentation (DLAS) is becoming increasingly attractive. No investigation of its application to OARs in thoracic magnetic resonance images (MRIs) from MRgRT has been done so far. This study aimed to fill this gap.

MATERIALS AND METHODS

122 planning MRIs from patients treated at a 0.35 T MR-Linac were retrospectively collected. Using an 80/19/23 (training/validation/test) split, individual 3D U-Nets for segmentation of the left lung, right lung, heart, aorta, spinal canal and esophagus were trained. These were compared to the clinically used contours based on Dice similarity coefficient (DSC) and Hausdorff distance (HD). They were also graded on their clinical usability by a radiation oncologist.

RESULTS

Median DSC was 0.96, 0.96, 0.94, 0.90, 0.88 and 0.78 for left lung, right lung, heart, aorta, spinal canal and esophagus, respectively. Median 95th percentile values of the HD were 3.9, 5.3, 5.8, 3.0, 2.6 and 3.5 mm, respectively. The physician preferred the network generated contours over the clinical contours, deeming 85 out of 129 to not require any correction, 25 immediately usable for treatment planning, 15 requiring minor and 4 requiring major corrections.

CONCLUSIONS

We trained 3D U-Nets on clinical MRI planning data which produced accurate delineations in the thoracic region. DLAS contours were preferred over the clinical contours.

Learning from pseudo-labels: deep networks improve consistency in longitudinal brain volume estimation

Introduction

Brain atrophy is a critical biomarker of disease progression and treatment response in neurodegenerative diseases such as multiple sclerosis (MS). Confounding factors such as inconsistent imaging acquisitions hamper the accurate measurement of brain atrophy in the clinic. This study aims to develop and validate a robust deep learning model to overcome these challenges; and to evaluate its impact on the measurement of disease progression.

Methods

Voxel-wise pseudo-atrophy labels were generated using SIENA, a widely adopted tool for the measurement of brain atrophy in MS. Deformation maps were produced for 195 pairs of longitudinal 3D T1 scans from patients with MS. A 3D U-Net, namely DeepBVC, was specifically developed overcome common variances in resolution, signal-to-noise ratio and contrast ratio between baseline and follow up scans. The performance of DeepBVC was compared against SIENA using McLaren test-retest dataset and 233 in-house MS subjects with MRI from multiple time points. Clinical evaluation included disability assessment with the Expanded Disability Status Scale (EDSS) and traditional imaging metrics such as lesion burden.

Results

For 3 subjects in test-retest experiments, the median percent brain volume change (PBVC) for DeepBVC and SIENA was 0.105 vs. 0.198% (subject 1), 0.061 vs. 0.084% (subject 2), 0.104 vs. 0.408% (subject 3). For testing consistency across multiple time points in individual MS subjects, the mean (± standard deviation) PBVC difference of DeepBVC and SIENA were 0.028% (± 0.145%) and 0.031% (±0.154%), respectively. The linear correlation with baseline T2 lesion volume were r = -0.288 (p < 0.05) and r = -0.249 (p < 0.05) for DeepBVC and SIENA, respectively. There was no significant correlation of disability progression with PBVC as estimated by either method (p = 0.86, p = 0.84).

Discussion

DeepBVC is a deep learning powered brain volume change estimation method for assessing brain atrophy used T1-weighted images. Compared to SIENA, DeepBVC demonstrates superior performance in reproducibility and in the context of common clinical scan variances such as imaging contrast, voxel resolution, random bias field, and signal-to-noise ratio. Enhanced measurement robustness, automation, and processing speed of DeepBVC indicate its potential for utilisation in both research and clinical environments for monitoring disease progression and, potentially, evaluating treatment effectiveness.

Accurate Bayesian segmentation of thalamic nuclei using diffusion MRI and an improved histological atlas

The human thalamus is a highly connected brain structure, which is key for the control of numerous functions and is involved in several neurological disorders. Recently, neuroimaging studies have increasingly focused on the volume and connectivity of the specific nuclei comprising this structure, rather than looking at the thalamus as a whole. However, accurate identification of cytoarchitectonically designed histological nuclei on standard in vivo structural MRI is hampered by the lack of image contrast that can be used to distinguish nuclei from each other and from surrounding white matter tracts. While diffusion MRI may offer such contrast, it has lower resolution and lacks some boundaries visible in structural imaging. In this work, we present a Bayesian segmentation algorithm for the thalamus. This algorithm combines prior information from a probabilistic atlas with likelihood models for both structural and diffusion MRI, allowing segmentation of 25 thalamic labels per hemisphere informed by both modalities. We present an improved probabilistic atlas, incorporating thalamic nuclei identified from histology and 45 white matter tracts surrounding the thalamus identified in ultra-high gradient strength diffusion imaging. We present a family of likelihood models for diffusion tensor imaging, ensuring compatibility with the vast majority of neuroimaging datasets that include diffusion MRI data. The use of these diffusion likelihood models greatly improves identification of nuclear groups versus segmentation based solely on structural MRI. Dice comparison of 5 manually identifiable groups of nuclei to ground truth segmentations show improvements of up to 10 percentage points. Additionally, our chosen model shows a high degree of reliability, with median test-retest Dice scores above 0.85 for four out of five nuclei groups, whilst also offering improved detection of differential thalamic involvement in Alzheimer's disease (AUROC 81.98%). The probabilistic atlas and segmentation tool will be made publicly available as part of the neuroimaging package FreeSurfer (https://freesurfer.net/fswiki/ThalamicNucleiDTI).

An automatic pipeline for atlas-based fetal and neonatal brain segmentation and analysis

BACKGROUND AND OBJECTIVE

The automatic segmentation of perinatal brain structures in magnetic resonance imaging (MRI) is of utmost importance for the study of brain growth and related complications. While different methods exist for adult and pediatric MRI data, there is a lack for automatic tools for the analysis of perinatal imaging.

METHODS

In this work, a new pipeline for fetal and neonatal segmentation has been developed. We also report the creation of two new fetal atlases, and their use within the pipeline for atlas-based segmentation, based on novel registration methods. The pipeline is also able to extract cortical and pial surfaces and compute features, such as curvature, local gyrification index, sulcal depth, and thickness.

RESULTS

Results show that the introduction of the new templates together with our segmentation strategy leads to accurate results when compared to expert annotations, as well as better performances when compared to a reference pipeline (developing Human Connectome Project (dHCP)), for both early and late-onset fetal brains.

CONCLUSIONS

These findings show the potential of the presented atlases and the whole pipeline for application in both fetal, neonatal, and longitudinal studies, which could lead to dramatic improvements in the understanding of perinatal brain development.

CerebNet: A fast and reliable deep-learning pipeline for detailed cerebellum sub-segmentation

Quantifying the volume of the cerebellum and its lobes is of profound interest in various neurodegenerative and acquired diseases. Especially for the most common spinocerebellar ataxias (SCA), for which the first antisense oligonculeotide-base gene silencing trial has recently started, there is an urgent need for quantitative, sensitive imaging markers at pre-symptomatic stages for stratification and treatment assessment. This work introduces CerebNet, a fully automated, extensively validated, deep learning method for the lobular segmentation of the cerebellum, including the separation of gray and white matter. For training, validation, and testing, T1-weighted images from 30 participants were manually annotated into cerebellar lobules and vermal sub-segments, as well as cerebellar white matter. CerebNet combines FastSurferCNN, a UNet-based 2.5D segmentation network, with extensive data augmentation, e.g. realistic non-linear deformations to increase the anatomical variety, eliminating additional preprocessing steps, such as spatial normalization or bias field correction. CerebNet demonstrates a high accuracy (on average 0.87 Dice and 1.742mm Robust Hausdorff Distance across all structures) outperforming state-of-the-art approaches. Furthermore, it shows high test-retest reliability (average ICC >0.97 on OASIS and Kirby) as well as high sensitivity to disease effects, including the pre-ataxic stage of spinocerebellar ataxia type 3 (SCA3). CerebNet is compatible with FreeSurfer and FastSurfer and can analyze a 3D volume within seconds on a consumer GPU in an end-to-end fashion, thus providing an efficient and validated solution for assessing cerebellum sub-structure volumes. We make CerebNet available as source-code (https://github.com/Deep-MI/FastSurfer).

High-resolution atlasing and segmentation of the subcortex: Review and perspective on challenges and opportunities created by machine learning

This paper reviews almost three decades of work on atlasing and segmentation methods for subcortical structures in human brain MRI. In writing this survey, we have three distinct aims. First, to document the evolution of digital subcortical atlases of the human brain, from the early MRI templates published in the nineties, to the complex multi-modal atlases at the subregion level that are available today. Second, to provide a detailed record of related efforts in the automated segmentation front, from earlier atlas-based methods to modern machine learning approaches. And third, to present a perspective on the future of high-resolution atlasing and segmentation of subcortical structures in in vivo human brain MRI, including open challenges and opportunities created by recent developments in machine learning.

The Effect of Self-Paced Exercise Intensity and Cardiorespiratory Fitness on Frontal Grey Matter Volume in Cognitively Normal Older Adults: A Randomised Controlled Trial

OBJECTIVE

Exercise has been found to be important in maintaining neurocognitive health. However, the effect of exercise intensity level remains relatively underexplored. Thus, to test the hypothesis that self-paced high-intensity exercise and cardiorespiratory fitness (peak aerobic capacity; VO_2peak) increase grey matter (GM) volume, we examined the effect of a 6-month exercise intervention on frontal lobe GM regions that support the executive functions in older adults.

METHODS

Ninety-eight cognitively normal participants (age = 69.06 ± 5.2 years; n = 54 female) were randomised into either a self-paced high- or moderate-intensity cycle-based exercise intervention group, or a no-intervention control group. Participants underwent magnetic resonance imaging and fitness assessment pre-intervention, immediately post-intervention, and 12-months post-intervention.

RESULTS

The intervention was found to increase fitness in the exercise groups, as compared with the control group (F = 9.88, p = <0.001). Changes in pre-to-post-intervention fitness were associated with increased volume in the right frontal lobe (β = 0.29, p = 0.036, r = 0.27), right supplementary motor area (β = 0.30, p = 0.031, r = 0.29), and both right (β = 0.32, p = 0.034, r = 0.30) and left gyrus rectus (β = 0.30, p = 0.037, r = 0.29) for intervention, but not control participants. No differences in volume were observed across groups.

CONCLUSIONS

At an aggregate level, six months of self-paced high- or moderate-intensity exercise did not increase frontal GM volume. However, experimentally-induced changes in individual cardiorespiratory fitness was positively associated with frontal GM volume in our sample of older adults. These results provide evidence of individual variability in exercise-induced fitness on brain structure.

StRegA: Unsupervised anomaly detection in brain MRIs using a compact context-encoding variational autoencoder

Expert interpretation of anatomical images of the human brain is the central part of neuroradiology. Several machine learning-based techniques have been proposed to assist in the analysis process. However, the ML models typically need to be trained to perform a specific task, e.g., brain tumour segmentation or classification. Not only do the corresponding training data require laborious manual annotations, but a wide variety of abnormalities can be present in a human brain MRI - even more than one simultaneously, which renders a representation of all possible anomalies very challenging. Hence, a possible solution is an unsupervised anomaly detection (UAD) system that can learn a data distribution from an unlabelled dataset of healthy subjects and then be applied to detect out-of-distribution samples. Such a technique can then be used to detect anomalies - lesions or abnormalities, for example, brain tumours, without explicitly training the model for that specific pathology. Several Variational Autoencoder (VAE) based techniques have been proposed in the past for this task. Even though they perform very well on controlled artificially simulated anomalies, many of them perform poorly while detecting anomalies in clinical data. This research proposes a compact version of the "context-encoding" VAE (ceVAE) model, combined with pre and post-processing steps, creating a UAD pipeline (StRegA), which is more robust on clinical data and shows its applicability in detecting anomalies such as tumours in brain MRIs. The proposed pipeline achieved a Dice score of 0.642 ± 0.101 while detecting tumours in T2w images of the BraTS dataset and 0.859 ± 0.112 while detecting artificially induced anomalies, while the best performing baseline achieved 0.522 ± 0.135 and 0.783 ± 0.111, respectively.

Accurate segmentation of neonatal brain MRI with deep learning

An important step toward delivering an accurate connectome of the human brain is robust segmentation of 3D Magnetic Resonance Imaging (MRI) scans, which is particularly challenging when carried out on perinatal data. In this paper, we present an automated, deep learning-based pipeline for accurate segmentation of tissues from neonatal brain MRI and extend it by introducing an age prediction pathway. A major constraint to using deep learning techniques on developing brain data is the need to collect large numbers of ground truth labels. We therefore also investigate two practical approaches that can help alleviate the problem of label scarcity without loss of segmentation performance. First, we examine the efficiency of different strategies of distributing a limited budget of annotated 2D slices over 3D training images. In the second approach, we compare the segmentation performance of pre-trained models with different strategies of fine-tuning on a small subset of preterm infants. Our results indicate that distributing labels over a larger number of brain scans can improve segmentation performance. We also show that even partial fine-tuning can be superior in performance to a model trained from scratch, highlighting the relevance of transfer learning strategies under conditions of label scarcity. We illustrate our findings on large, publicly available T1- and T2-weighted MRI scans (n = 709, range of ages at scan: 26–45 weeks) obtained retrospectively from the Developing Human Connectome Project (dHCP) cohort.

The relationship between chronic PTSD, cortical volumetry and white matter microstructure among Australian combat veterans

BACKGROUND

Posttraumatic stress disorder (PTSD) has been associated with volumetric and white matter microstructural changes among general and veteran populations. However, regions implicated have greatly varied and often conflict between studies, potentially due to confounding comorbidities within samples. This study compared grey matter volume and white matter microstructure among Australian combat veterans with and without a lifetime diagnosis of PTSD, in a homogenous sample assessed for known confounding comorbidities.

METHODS

Sixty-eight male trauma-exposed veterans (16 PTSD-diagnosed; mean age 69 years) completed a battery of psychometric assessments and underwent magnetic resonance and diffusion tensor imaging. Analyses included tract-based spatial statistics, voxel-wise analyses, diffusion connectome-based group-wise analysis, and volumetric analysis.

RESULTS

Significantly smaller grey matter volumes were observed in the left prefrontal cortex (P = 0.026), bilateral middle frontal gyrus (P = 0.021), and left anterior insula (P = 0.048) in the PTSD group compared to controls. Significant negative correlations were found between PTSD symptom severity and fractional anisotropy values in the left corticospinal tract (R² = 0.34, P = 0.024) and left inferior cerebellar peduncle (R² = 0.62, P = 0.016). No connectome-based differences in white matter properties were observed.

CONCLUSIONS

Findings from this study reinforce reports of white matter alterations, as indicated by reduced fractional anisotropy values, in relation to PTSD symptom severity, as well as patterns of reduced volume in the prefrontal cortex. These results contribute to the developing profile of neuroanatomical differences uniquely attributable to veterans who suffer from chronic PTSD.

Transcallosal and Corticospinal White Matter Disease and Its Association With Motor Impairment in Multiple Sclerosis

Purpose

In this cross-sectional, proof-of-concept study, we propose that using the more pathologically-specific neurite orientation dispersion and density imaging (NODDI) method, in conjunction with high-resolution probabilistic tractography, white matter tract templates can improve the assessment of regional axonal injury and its association with disability of people with multiple sclerosis (pwMS).

Methods

Parametric maps of the neurite density index, orientation dispersion index, and the apparent isotropic volume fraction (IVF) were estimated in 18 pwMS and nine matched healthy controls (HCs). Tract-specific values were measured in transcallosal (TC) fibers from the paracentral lobules and TC and corticospinal fibers from the ventral and dorsal premotor areas, presupplementary and supplementary motor areas, and primary motor cortex. The nonparametric Mann-Whitney U test assessed group differences in the NODDI-derived metrics; the Spearman's rank correlation analyses measured associations between the NODDI metrics and other clinical or radiological variables.

Results

IVF values of the TC fiber bundles from the paracentral, presupplementary, and supplementary motor areas were both higher in pwMS than in HCs (p ≤ 0.045) and in pwMS with motor disability compared to those without motor disability (p ≤ 0.049). IVF in several TC tracts was associated with the Expanded Disability Status Scale score (p ≤ 0.047), while regional and overall lesion burden correlated with the Timed 25-Foot Walking Test (p ≤ 0.049).

Conclusion

IVF alterations are present in pwMS even when the other NODDI metrics are still mostly preserved. Changes in IVF are biologically non-specific and may not necessarily drive irreversible functional loss. However, by possibly preceding downstream pathologies that are strongly associated with disability accretion, IVF changes are indicators of, otherwise, occult prelesional tissue injury.

Reduced cortical cholinergic innervation measured using [18F]-FEOBV PET imaging correlates with cognitive decline in mild cognitive impairment

•

Topographic FEOBV binding correlates with domain-specific cognitive performance.

•

Global and regional reductions in cholinergic innervation are observed in MCI.

•

Global FEOBV SUVR is associated with basal forebrain and hippocampal volumes.

•

Our results provide proof of concept for FEOBV PET to assess cholinergic terminal integrity.

Dysfunction of the cholinergic basal forebrain (BF) neurotransmitter system, including cholinergic axon denervation of the cortex, plays an important role in cognitive decline and dementia. A validated method to directly quantify cortical cholinergic terminal integrity enables exploration of the involvement of this system in diverse cognitive profiles associated with dementia, particularly at a prodromal stage. In this study, we used the radiotracer [¹⁸F]-fluoroethoxybenzovesamicol (FEOBV) as a direct measure of cholinergic terminal integrity and investigated its value for the assessment of cholinergic denervation in the cortex and associated cognitive deficits. Eighteen participants (8 with mild cognitive impairment (MCI) and 10 cognitively unimpaired controls) underwent neuropsychological assessment and brain imaging using FEOBV and [¹⁸F]-florbetaben for amyloid-β imaging. The MCI group showed a significant global reduction of FEOBV retention in the cortex and in the parietal and occipital cortices specifically compared to the control group. The global cortical FEOBV retention of all participants positively correlated with the BF, hippocampus and grey matter volumes, but no association was found between the global FEOBV retention and amyloid-β status. Topographic profiles from voxel-wise analysis of FEOBV images revealed significant positive correlations with the cognitive domains associated with the underlying cortical areas. Overlapping profiles of decreased FEOBV were identified in correlation with impairment in executive function, attention and language, which covered the anterior cingulate gyrus, olfactory cortex, calcarine cortex, middle temporal gyrus and caudate nucleus. However, the absence of cortical atrophy in these areas suggested that reduced cholinergic terminal integrity in the cortex is an important factor underlying the observed cognitive decline in early dementia. Our results provide support for the utility and validity of FEOBV PET for quantitative assessment of region-specific cholinergic terminal integrity that could potentially be used for early detection of cholinergic dysfunction in dementia following further validation in larger cohorts.

A cascaded nested network for 3T brain MR image segmentation guided by 7T labeling

Accurate segmentation of the brain into gray matter, white matter, and cerebrospinal fluid using magnetic resonance (MR) imaging is critical for visualization and quantification of brain anatomy. Compared to 3T MR images, 7T MR images exhibit higher tissue contrast that is contributive to accurate tissue delineation for training segmentation models. In this paper, we propose a cascaded nested network (CaNes-Net) for segmentation of 3T brain MR images, trained by tissue labels delineated from the corresponding 7T images. We first train a nested network (Nes-Net) for a rough segmentation. The second Nes-Net uses tissue-specific geodesic distance maps as contextual information to refine the segmentation. This process is iterated to build CaNes-Net with a cascade of Nes-Net modules to gradually refine the segmentation. To alleviate the misalignment between 3T and corresponding 7T MR images, we incorporate a correlation coefficient map to allow well-aligned voxels to play a more important role in supervising the training process. We compared CaNes-Net with SPM and FSL tools, as well as four deep learning models on 18 adult subjects and the ADNI dataset. Our results indicate that CaNes-Net reduces segmentation errors caused by the misalignment and improves segmentation accuracy substantially over the competing methods.

SynthMorph: Learning Contrast-Invariant Registration Without Acquired Images

We introduce a strategy for learning image registration without acquired imaging data, producing powerful networks agnostic to contrast introduced by magnetic resonance imaging (MRI). While classical registration methods accurately estimate the spatial correspondence between images, they solve an optimization problem for every new image pair. Learning-based techniques are fast at test time but limited to registering images with contrasts and geometric content similar to those seen during training. We propose to remove this dependency on training data by leveraging a generative strategy for diverse synthetic label maps and images that exposes networks to a wide range of variability, forcing them to learn more invariant features. This approach results in powerful networks that accurately generalize to a broad array of MRI contrasts. We present extensive experiments with a focus on 3D neuroimaging, showing that this strategy enables robust and accurate registration of arbitrary MRI contrasts even if the target contrast is not seen by the networks during training. We demonstrate registration accuracy surpassing the state of the art both within and across contrasts, using a single model. Critically, training on arbitrary shapes synthesized from noise distributions results in competitive performance, removing the dependency on acquired data of any kind. Additionally, since anatomical label maps are often available for the anatomy of interest, we show that synthesizing images from these dramatically boosts performance, while still avoiding the need for real intensity images. Our code is available at doic https://w3id.org/synthmorph.

Factorisation-Based Image Labelling

Segmentation of brain magnetic resonance images (MRI) into anatomical regions is a useful task in neuroimaging. Manual annotation is time consuming and expensive, so having a fully automated and general purpose brain segmentation algorithm is highly desirable. To this end, we propose a patched-based labell propagation approach based on a generative model with latent variables. Once trained, our Factorisation-based Image Labelling (FIL) model is able to label target images with a variety of image contrasts. We compare the effectiveness of our proposed model against the state-of-the-art using data from the MICCAI 2012 Grand Challenge and Workshop on Multi-Atlas Labelling. As our approach is intended to be general purpose, we also assess how well it can handle domain shift by labelling images of the same subjects acquired with different MR contrasts.

Higher Coffee Consumption Is Associated With Slower Cognitive Decline and Less Cerebral Aβ-Amyloid Accumulation Over 126 Months: Data From the Australian Imaging, Biomarkers, and Lifestyle Study

Background: Worldwide, coffee is one of the most popular beverages consumed. Several studies have suggested a protective role of coffee, including reduced risk of Alzheimer's disease (AD). However, there is limited longitudinal data from cohorts of older adults reporting associations of coffee intake with cognitive decline, in distinct domains, and investigating the neuropathological mechanisms underpinning any such associations. Methods: The aim of the current study was to investigate the relationship between self-reported habitual coffee intake, and cognitive decline assessed using a comprehensive neuropsychological battery in 227 cognitively normal older adults from the Australian Imaging, Biomarkers, and Lifestyle (AIBL) study, over 126 months. In a subset of individuals, we also investigated the relationship between habitual coffee intake and cerebral Aβ-amyloid accumulation (n = 60) and brain volumes (n = 51) over 126 months. Results: Higher baseline coffee consumption was associated with slower cognitive decline in executive function, attention, and the AIBL Preclinical AD Cognitive Composite (PACC; shown reliably to measure the first signs of cognitive decline in at-risk cognitively normal populations), and lower likelihood of transitioning to mild cognitive impairment or AD status, over 126 months. Higher baseline coffee consumption was also associated with slower Aβ-amyloid accumulation over 126 months, and lower risk of progressing to "moderate," "high," or "very high" Aβ-amyloid burden status over the same time-period. There were no associations between coffee intake and atrophy in total gray matter, white matter, or hippocampal volume. Discussion: Our results further support the hypothesis that coffee intake may be a protective factor against AD, with increased coffee consumption potentially reducing cognitive decline by slowing cerebral Aβ-amyloid accumulation, and thus attenuating the associated neurotoxicity from Aβ-amyloid-mediated oxidative stress and inflammatory processes. Further investigation is required to evaluate whether coffee intake could be incorporated as a modifiable lifestyle factor aimed at delaying AD onset.

Differences Between MR Brain Region Segmentation Methods: Impact on Single-Subject Analysis

For the segmentation of magnetic resonance brain images into anatomical regions, numerous fully automated methods have been proposed and compared to reference segmentations obtained manually. However, systematic differences might exist between the resulting segmentations, depending on the segmentation method and underlying brain atlas. This potentially results in sensitivity differences to disease and can further complicate the comparison of individual patients to normative data. In this study, we aim to answer two research questions: 1) to what extent are methods interchangeable, as long as the same method is being used for computing normative volume distributions and patient-specific volumes? and 2) can different methods be used for computing normative volume distributions and assessing patient-specific volumes? To answer these questions, we compared volumes of six brain regions calculated by five state-of-the-art segmentation methods: Erasmus MC (EMC), FreeSurfer (FS), geodesic information flows (GIF), multi-atlas label propagation with expectation–maximization (MALP-EM), and model-based brain segmentation (MBS). We applied the methods on 988 non-demented (ND) subjects and computed the correlation (PCC-v) and absolute agreement (ICC-v) on the volumes. For most regions, the PCC-v was good (>0.75), indicating that volume differences between methods in ND subjects are mainly due to systematic differences. The ICC-v was generally lower, especially for the smaller regions, indicating that it is essential that the same method is used to generate normative and patient data. To evaluate the impact on single-subject analysis, we also applied the methods to 42 patients with Alzheimer’s disease (AD). In the case where the normative distributions and the patient-specific volumes were calculated by the same method, the patient’s distance to the normative distribution was assessed with the z-score. We determined the diagnostic value of this z-score, which showed to be consistent across methods. The absolute agreement on the AD patients’ z-scores was high for regions of thalamus and putamen. This is encouraging as it indicates that the studied methods are interchangeable for these regions. For regions such as the hippocampus, amygdala, caudate nucleus and accumbens, and globus pallidus, not all method combinations showed a high ICC-z. Whether two methods are indeed interchangeable should be confirmed for the specific application and dataset of interest.

Whole brain segmentation with full volume neural network

Whole brain segmentation is an important neuroimaging task that segments the whole brain volume into anatomically labeled regions-of-interest. Convolutional neural networks have demonstrated good performance in this task. Existing solutions, usually segment the brain image by classifying the voxels, or labeling the slices or the sub-volumes separately. Their representation learning is based on parts of the whole volume whereas their labeling result is produced by aggregation of partial segmentation. Learning and inference with incomplete information could lead to sub-optimal final segmentation result. To address these issues, we propose to adopt a full volume framework, which feeds the full volume brain image into the segmentation network and directly outputs the segmentation result for the whole brain volume. The framework makes use of complete information in each volume and can be implemented easily. An effective instance in this framework is given subsequently. We adopt the 3D high-resolution network (HRNet) for learning spatially fine-grained representations and the mixed precision training scheme for memory-efficient training. Extensive experiment results on a publicly available 3D MRI brain dataset show that our proposed model advances the state-of-the-art methods in terms of segmentation performance.

A Contrast Augmentation Approach to Improve Multi-Scanner Generalization in MRI

Most data-driven methods are very susceptible to data variability. This problem is particularly apparent when applying Deep Learning (DL) to brain Magnetic Resonance Imaging (MRI), where intensities and contrasts vary due to acquisition protocol, scanner- and center-specific factors. Most publicly available brain MRI datasets originate from the same center and are homogeneous in terms of scanner and used protocol. As such, devising robust methods that generalize to multi-scanner and multi-center data is crucial for transferring these techniques into clinical practice. We propose a novel data augmentation approach based on Gaussian Mixture Models (GMM-DA) with the goal of increasing the variability of a given dataset in terms of intensities and contrasts. The approach allows to augment the training dataset such that the variability in the training set compares to what is seen in real world clinical data, while preserving anatomical information. We compare the performance of a state-of-the-art U-Net model trained for segmenting brain structures with and without the addition of GMM-DA. The models are trained and evaluated on single- and multi-scanner datasets. Additionally, we verify the consistency of test-retest results on same-patient images (same and different scanners). Finally, we investigate how the presence of bias field influences the performance of a model trained with GMM-DA. We found that the addition of the GMM-DA improves the generalization capability of the DL model to other scanners not present in the training data, even when the train set is already multi-scanner. Besides, the consistency between same-patient segmentation predictions is improved, both for same-scanner and different-scanner repetitions. We conclude that GMM-DA could increase the transferability of DL models into clinical scenarios.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

TorchIO: A Python library for efficient loading, preprocessing, augmentation and patch-based sampling of medical images in deep learning

BACKGROUND AND OBJECTIVE

Processing of medical images such as MRI or CT presents different challenges compared to RGB images typically used in computer vision. These include a lack of labels for large datasets, high computational costs, and the need of metadata to describe the physical properties of voxels. Data augmentation is used to artificially increase the size of the training datasets. Training with image subvolumes or patches decreases the need for computational power. Spatial metadata needs to be carefully taken into account in order to ensure a correct alignment and orientation of volumes.

METHODS

We present TorchIO, an open-source Python library to enable efficient loading, preprocessing, augmentation and patch-based sampling of medical images for deep learning. TorchIO follows the style of PyTorch and integrates standard medical image processing libraries to efficiently process images during training of neural networks. TorchIO transforms can be easily composed, reproduced, traced and extended. Most transforms can be inverted, making the library suitable for test-time augmentation and estimation of aleatoric uncertainty in the context of segmentation. We provide multiple generic preprocessing and augmentation operations as well as simulation of MRI-specific artifacts.

RESULTS

Source code, comprehensive tutorials and extensive documentation for TorchIO can be found at http://torchio.rtfd.io/. The package can be installed from the Python Package Index (PyPI) running pip install torchio. It includes a command-line interface which allows users to apply transforms to image files without using Python. Additionally, we provide a graphical user interface within a TorchIO extension in 3D Slicer to visualize the effects of transforms.

CONCLUSION

TorchIO was developed to help researchers standardize medical image processing pipelines and allow them to focus on the deep learning experiments. It encourages good open-science practices, as it supports experiment reproducibility and is version-controlled so that the software can be cited precisely. Due to its modularity, the library is compatible with other frameworks for deep learning with medical images.

Joint super-resolution and synthesis of 1 mm isotropic MP-RAGE volumes from clinical MRI exams with scans of different orientation, resolution and contrast

Most existing algorithms for automatic 3D morphometry of human brain MRI scans are designed for data with near-isotropic voxels at approximately 1 mm resolution, and frequently have contrast constraints as well-typically requiring T1-weighted images (e.g., MP-RAGE scans). This limitation prevents the analysis of millions of MRI scans acquired with large inter-slice spacing in clinical settings every year. In turn, the inability to quantitatively analyze these scans hinders the adoption of quantitative neuro imaging in healthcare, and also precludes research studies that could attain huge sample sizes and hence greatly improve our understanding of the human brain. Recent advances in convolutional neural networks (CNNs) are producing outstanding results in super-resolution and contrast synthesis of MRI. However, these approaches are very sensitive to the specific combination of contrast, resolution and orientation of the input images, and thus do not generalize to diverse clinical acquisition protocols - even within sites. In this article, we present SynthSR, a method to train a CNN that receives one or more scans with spaced slices, acquired with different contrast, resolution and orientation, and produces an isotropic scan of canonical contrast (typically a 1 mm MP-RAGE). The presented method does not require any preprocessing, beyond rigid coregistration of the input scans. Crucially, SynthSR trains on synthetic input images generated from 3D segmentations, and can thus be used to train CNNs for any combination of contrasts, resolutions and orientations without high-resolution real images of the input contrasts. We test the images generated with SynthSR in an array of common downstream analyses, and show that they can be reliably used for subcortical segmentation and volumetry, image registration (e.g., for tensor-based morphometry), and, if some image quality requirements are met, even cortical thickness morphometry. The source code is publicly available at https://github.com/BBillot/SynthSR.

Synthetic microbleeds generation for classifier training without ground truth

BACKGROUND AND OBJECTIVE

Cerebral microbleeds (CMB) are important biomarkers of cerebrovascular diseases and cognitive dysfunctions. Susceptibility weighted imaging (SWI) is a common MRI sequence where CMB appear as small hypointense blobs. The prevalence of CMB in the population and in each scan is low, resulting in tedious and time-consuming visual assessment. Automated detection methods would be of value but are challenged by the CMB low prevalence, the presence of mimics such as blood vessels, and the difficulty to obtain sufficient ground truth for training and testing. In this paper, synthetic CMB (sCMB) generation using an analytical model is proposed for training and testing machine learning methods. The main aim is creating perfect synthetic ground truth as similar as reals, in high number, with a high diversity of shape, volume, intensity, and location to improve training of supervised methods.

METHOD

sCMB were modelled with a random Gaussian shape and added to healthy brain locations. We compared training on our synthetic data to standard augmentation techniques. We performed a validation experiment using sCMB and report result for whole brain detection using a 10-fold cross validation design with an ensemble of 10 neural networks.

RESULTS

Performance was close to state of the art (~9 false positives per scan), when random forest was trained on synthetic only and tested on real lesion. Other experiments showed that top detection performance could be achieved when training on synthetic CMB only. Our dataset is made available, including a version with 37,000 synthetic lesions, that could be used for benchmarking and training.

CONCLUSION

Our proposed synthetic microbleeds model is a powerful data augmentation approach for CMB classification with and should be considered for training automated lesion detection system from MRI SWI.

An atlas of classifiers-a machine learning paradigm for brain MRI segmentation

We present the Atlas of Classifiers (AoC)-a conceptually novel framework for brain MRI segmentation. The AoC is a spatial map of voxel-wise multinomial logistic regression (LR) functions learned from the labeled data. Upon convergence, the resulting fixed LR weights, a few for each voxel, represent the training dataset. It can, therefore, be considered as a light-weight learning machine, which despite its low capacity does not underfit the problem. The AoC construction is independent of the actual intensities of the test images, providing the flexibility to train it on the available labeled data and use it for the segmentation of images from different datasets and modalities. In this sense, it does not overfit the training data, as well. The proposed method has been applied to numerous publicly available datasets for the segmentation of brain MRI tissues and is shown to be robust to noise and outreach commonly used methods. Promising results were also obtained for multi-modal, cross-modality MRI segmentation. Finally, we show how AoC trained on brain MRIs of healthy subjects can be exploited for lesion segmentation of multiple sclerosis patients.

Assessing brain injury topographically using MR neurite orientation dispersion and density imaging in multiple sclerosis

BACKGROUND AND PURPOSE

Axonal injury is a key player of disability in persons with multiple sclerosis (pwMS). Yet, detecting and measuring it in vivo is challenging. The neurite orientation dispersion and density imaging (NODDI) proposes a novel framework for probing axonal integrity in vivo. NODDI at 3.0 Tesla was used to quantify tissue damage in pwMS and its relationship with disease progression.

METHODS

Eighteen pwMS (4 clinically isolated syndrome, 11 relapsing remitting, and 3 secondary progressive MS) and nine age- and sex-matched healthy controls underwent a brain MRI, inclusive of clinical sequences and a multi-shell diffusion acquisition. Parametric maps of axial diffusivity (AD), neurite density index (ndi), apparent isotropic volume fraction (ivf), and orientation dispersion index (odi) were fitted. Anatomically matched regions of interest were used to quantify AD and NODDI-derived metrics and to assess the relations between these measures and those of disease progression.

RESULTS

AD, ndi, ivf, and odi significantly differed between chronic black holes (cBHs) and T2-lesions, and between the latter and normal appearing white matter (NAWM). All metrics except ivf significantly differed between NAWM located next to a cBH and that situated contra-laterally. Only NAWM odi was significantly associated with T2-lesion volume, the timed 25-foot walk test and disease duration.

CONCLUSIONS

NODDI is sensitive to tissue injury but its relationship with clinical progression remains limited.

GABA concentrations in the anterior cingulate and dorsolateral prefrontal cortices: Associations with chronic cigarette smoking, neurocognition, and decision making

Chronic cigarette smoking is associated with regional metabolite abnormalities in choline-containing compounds, creatine-containing compounds, glutamate, and N-acetylaspartate. The effects of cigarette smoking on anterior frontal cortical gamma-aminobutyric acid (GABA) concentration are unknown. This study compared chronic smokers (n = 33) and nonsmokers (n = 31) on anterior cingulate cortex (ACC) and right dorsolateral prefrontal cortex (DLPFC) GABA+ (the sum of GABA and coedited macromolecules) concentrations and associations of GABA+ levels in these regions with seven neurocognitive domains of functioning, decision making, and impulsivity measures. Smokers had significantly lower right DLPFC GABA+ concentration than nonsmokers, but groups were equivalent on ACC GABA+ level. Across groups, greater number of days since end of menstrual cycle was related to higher GABA+ level in the ACC but not right DLPFC GABA+ concentration. In exploratory correlation analyses, higher ACC and right DLPFC GABA+ levels were associated with faster processing speed and better auditory-verbal memory, respectively, in the combined group of smokers and nonsmokers; in smokers only, higher ACC GABA+ was related to better decision making and auditory-verbal learning. This study contributes additional novel data on the adverse effects of chronic cigarette smoking on the adult human brain and demonstrated ACC and DLPFC GABA+ concentrations were associated with neurocognition and decision making/impulsivity in active cigarette smokers. Longitudinal studies on the effects of smoking cessation on regional brain GABA levels, with a greater number of female participants, are required to determine if the observed metabolite abnormalities are persistent or normalize with smoking cessation.

LEARNING MRI CONTRAST-AGNOSTIC REGISTRATION

We introduce a strategy for learning image registration without acquired imaging data, producing powerful networks agnostic to magnetic resonance imaging (MRI) contrast. While classical methods accurately estimate the spatial correspondence between images, they solve an optimization problem for every new image pair. Learning methods are fast at test time but limited to images with contrasts and geometric content similar to those seen during training. We propose to remove this dependency using a generative strategy that exposes networks to a wide range of images synthesized from segmentations during training, forcing them to generalize across contrasts. We show that networks trained within this framework generalize to a broad array of unseen MRI contrasts and surpass classical state-of-the-art brain registration accuracy by up to 12.4 Dice points for a variety of tested contrast combinations. Critically, training on arbitrary shapes synthesized from noise distributions results in competitive performance, removing the dependency on acquired data of any kind. Additionally, since anatomical label maps are often available for the anatomy of interest, we show that synthesizing images from these dramatically boosts performance, while still avoiding the need for real intensity images during training.

Automatic segmentation of white matter hyperintensities from brain magnetic resonance images in the era of deep learning and big data - A systematic review

BACKGROUND

White matter hyperintensities (WMH), of presumed vascular origin, are visible and quantifiable neuroradiological markers of brain parenchymal change. These changes may range from damage secondary to inflammation and other neurological conditions, through to healthy ageing. Fully automatic WMH quantification methods are promising, but still, traditional semi-automatic methods seem to be preferred in clinical research. We systematically reviewed the literature for fully automatic methods developed in the last five years, to assess what are considered state-of-the-art techniques, as well as trends in the analysis of WMH of presumed vascular origin.

METHOD

We registered the systematic review protocol with the International Prospective Register of Systematic Reviews (PROSPERO), registration number - CRD42019132200. We conducted the search for fully automatic methods developed from 2015 to July 2020 on Medline, Science direct, IEE Explore, and Web of Science. We assessed risk of bias and applicability of the studies using QUADAS 2.

RESULTS

The search yielded 2327 papers after removing 104 duplicates. After screening titles, abstracts and full text, 37 were selected for detailed analysis. Of these, 16 proposed a supervised segmentation method, 10 proposed an unsupervised segmentation method, and 11 proposed a deep learning segmentation method. Average DSC values ranged from 0.538 to 0.91, being the highest value obtained from an unsupervised segmentation method. Only four studies validated their method in longitudinal samples, and eight performed an additional validation using clinical parameters. Only 8/37 studies made available their methods in public repositories.

CONCLUSIONS

We found no evidence that favours deep learning methods over the more established k-NN, linear regression and unsupervised methods in this task. Data and code availability, bias in study design and ground truth generation influence the wider validation and applicability of these methods in clinical research.

Development and evaluation of a high performance T1-weighted brain template for use in studies on older adults

Τhe accuracy of template-based neuroimaging investigations depends on the template's image quality and representativeness of the individuals under study. Yet a thorough, quantitative investigation of how available standardized and study-specific T1-weighted templates perform in studies on older adults has not been conducted. The purpose of this work was to construct a high-quality standardized T1-weighted template specifically designed for the older adult brain, and systematically compare the new template to several other standardized and study-specific templates in terms of image quality, performance in spatial normalization of older adult data and detection of small inter-group morphometric differences, and representativeness of the older adult brain. The new template was constructed with state-of-the-art spatial normalization of high-quality data from 222 older adults. It was shown that the new template (a) exhibited high image sharpness, (b) provided higher inter-subject spatial normalization accuracy and (c) allowed detection of smaller inter-group morphometric differences compared to other standardized templates, (d) had similar performance to that of study-specific templates constructed with the same methodology, and (e) was highly representative of the older adult brain.

Perilesional neurodegenerative injury in multiple sclerosis: Relation to focal lesions and impact on disability

BACKGROUND

Axonal injury is the primary source of irreversible neurological decline in persons with multiple sclerosis (pwMS). Identifying and quantifying myelin and axonal loss in lesional and perilesional tissue in vivo is fundamental for a better understanding of multiple sclerosis (MS) outcomes and patient impairment. Using advanced magnetic resonance imaging (MRI) methods, consisting of selective inversion recovery quantitative magnetization transfer imaging (SIR-qMT) and multi-compartment diffusion MRI with the spherical mean technique (SMT), we conducted a cross-sectional pilot study to assess myelin and axonal damage in the normal appearing white matter (NAWM) surrounding chronic black holes (cBHs) and how this pathology correlates with disability in vivo. We hypothesized that lesional axonal transection propagates tissue injury in the surrounding NAWM and that the degree of this injury is related to patient disability.

METHODS

Eighteen pwMS underwent a 3.0 Tesla conventional clinical MRI, inclusive of T1 and T2 weighted protocols, as well as SIR-qMT and SMT. Regions of interests (ROIs) were manually delineated in cBHs, NAWM neighboring cBHs (perilesional NAWM), distant ipsilateral NAWM and contra-lateral distant NAWM. SIR-qMT-derived macromolecular-to-free pool size ratio (PSR) and SMT-derived apparent axonal volume fraction (V_ax) were extracted to infer on myelin and axonal content, respectively. Group differences were assessed using mixed-effects regression models and correlation analyses were obtained by bootstrapping 95% confidence interval.

RESULTS

In comparison to perilesional NAWM, both PSR and V_ax values were reduced in cBHs (p < 0.0001) and increased in distant contra-lateral NAWM ROIs (p < 0.001 for PSR and p < 0.0001 for V_ax) but not ipsilateral NAWM (p = 0.176 for PSR and p = 0.549 for V_ax). V_ax values measured in cBHs correlated with those in perilesional NAWM (Pearson rho = 0.63, p < 0.001). No statistically relevant associations were seen between PSR/V_ax values and clinical and/or MRI metrics of the disease with the exception of cBH PSR values, which correlated with the Expanded Disability Status Scale (Pearson rho = -0.63, p = 0.03).

CONCLUSIONS

Our results show that myelin and axonal content, detected by PSR and V_ax, are reduced in perilesional NAWM, as a function of the degree of focal cBH axonal injury. This finding is indicative of an ongoing anterograde/retrograde degeneration and suggests that treatment prevention of cBH development is a key factor for preserving NAWM integrity in surrounding tissue. It also suggests that measuring changes in perilesional areas over time may be a useful measure of outcome for proof-of-concept clinical trials on neuroprotection and repair. PSR and V_ax largely failed to capture associations with clinical and MRI characteristics, likely as a result of the small sample size and cross-sectional design, however, longitudinal assessment of a larger cohort may unravel the impact of this pathology on disease progression.

Breast Cancer Image Classification: A Review

BACKGROUND

Breast cancer represents uncontrolled breast cell growth. Breast cancer is the most diagnosed cancer in women worldwide. Early detection of breast cancer improves the chances of survival and increases treatment options. There are various methods for screening breast cancer, such as mammogram, ultrasound, computed tomography and Magnetic Resonance Imaging (MRI). MRI is gaining prominence as an alternative screening tool for early detection and breast cancer diagnosis. Nevertheless, MRI can hardly be examined without the use of a Computer-Aided Diagnosis (CAD) framework, due to the vast amount of data.

OBJECTIVE

This paper aims to cover the approaches used in the CAD system for the detection of breast cancer.

METHODS

In this paper, the methods used in CAD systems are categories into two classes: the conventional approach and artificial intelligence (AI) approach.

RESULTS

The conventional approach covers the basic steps of image processing, such as preprocessing, segmentation, feature extraction and classification. The AI approach covers the various convolutional and deep learning networks used for diagnosis.

CONCLUSION

This review discusses some of the core concepts used in breast cancer and presents a comprehensive review of efforts in the past to address this problem.

Quantitative definition of neurobehavior, vision, hearing and brain volumes in macaques congenitally exposed to Zika virus

Congenital Zika virus (ZIKV) exposure results in a spectrum of disease ranging from severe birth defects to delayed onset neurodevelopmental deficits. ZIKV-related neuropathogenesis, predictors of birth defects, and neurodevelopmental deficits are not well defined in people. Here we assess the methodological and statistical feasibility of a congenital ZIKV exposure macaque model for identifying infant neurobehavior and brain abnormalities that may underlie neurodevelopmental deficits. We inoculated five pregnant macaques with ZIKV and mock-inoculated one macaque in the first trimester. Following birth, growth, ocular structure/function, brain structure, hearing, histopathology, and neurobehavior were quantitatively assessed during the first week of life. We identified the typical pregnancy outcomes of congenital ZIKV infection, with fetal demise and placental abnormalities. We estimated sample sizes needed to define differences between groups and demonstrated that future studies quantifying brain region volumes, retinal structure, hearing, and visual pathway function require a sample size of 14 animals per group (14 ZIKV, 14 control) to detect statistically significant differences in at least half of the infant exam parameters. Establishing the parameters for future studies of neurodevelopmental outcomes following congenital ZIKV exposure in macaques is essential for robust and rigorous experimental design.

BISON: Brain tissue segmentation pipeline using T1 -weighted magnetic resonance images and a random forest classifier

PURPOSE

Tissue segmentation from T₁ -weighted (T1W) MRI is a critical requirement in many neuroscience and clinical applications. However, accurate tissue segmentation is challenging because of the variabilities in tissue intensity profiles caused by differences in scanner models, acquisition protocols, and age. In addition, many methods assume healthy anatomy and fail in the presence of pathology such as white matter hyperintensities (WMHs). We present BISON (Brain tISsue segmentatiON), a new pipeline for tissue segmentation using a random forest classifier and a set of intensity and location priors based on T1W MRI.

METHODS

BISON was developed and cross-validated using multiscanner manual labels of 72 subjects aged 5 to 96 years. We also assessed the test-retest reliability of BISON on two data sets: 20 subjects with scan/rescan MR images and manual segmentations and 90 scans from a single individual. The results were compared against Atropos, a state-of-the-art commonly used tissue classification method from advanced normalization tools (ANTs).

RESULTS

BISON cross-validation dice kappa values against manual segmentations of 72 MRI volumes yielded κ_GM = 0.88, κ_WM = 0.85, κ_CSF = 0.77, outperforming Atropos (κ_GM = 0.79, κ_WM = 0.84, κ_CSF = 0.64), test-retest values on 20 subjects of κ_GM = 0.94, κ_WM = 0.92, κ_CSF = 0.77 outperforming both manual (κ_GM = 0.92, κ_WM = 0.91, κ_CSF =0.74) and Atropos (κ_GM = 0.87, κ_WM = 0.92, κ_CSF = 0.79). Finally, BISON outperformed Atropos, FAST (fast automated segmentation tool) from the FMRIB (Functional Magnetic Resonance Imaging of the Brain) Software Library, and SPM12 (statistical parametric mapping 12) in the presence of WMHs.

CONCLUSION

BISON can provide accurate and robust segmentations in data from various age ranges and scanner models, making it ideal for performing tissue classification in large multicenter and multiscanner databases.