Stable Atlas-based Mapped Prior (STAMP) machine-learning segmentation for multicenter large-scale MRI data

Deep Learning in Large and Multi-Site Structural Brain MR Imaging Datasets

Large, multi-site, heterogeneous brain imaging datasets are increasingly required for the training, validation, and testing of advanced deep learning (DL)-based automated tools, including structural magnetic resonance (MR) image-based diagnostic and treatment monitoring approaches. When assembling a number of smaller datasets to form a larger dataset, understanding the underlying variability between different acquisition and processing protocols across the aggregated dataset (termed “batch effects”) is critical. The presence of variation in the training dataset is important as it more closely reflects the true underlying data distribution and, thus, may enhance the overall generalizability of the tool. However, the impact of batch effects must be carefully evaluated in order to avoid undesirable effects that, for example, may reduce performance measures. Batch effects can result from many sources, including differences in acquisition equipment, imaging technique and parameters, as well as applied processing methodologies. Their impact, both beneficial and adversarial, must be considered when developing tools to ensure that their outputs are related to the proposed clinical or research question (i.e., actual disease-related or pathological changes) and are not simply due to the peculiarities of underlying batch effects in the aggregated dataset. We reviewed applications of DL in structural brain MR imaging that aggregated images from neuroimaging datasets, typically acquired at multiple sites. We examined datasets containing both healthy control participants and patients that were acquired using varying acquisition protocols. First, we discussed issues around Data Access and enumerated the key characteristics of some commonly used publicly available brain datasets. Then we reviewed methods for correcting batch effects by exploring the two main classes of approaches: Data Harmonization that uses data standardization, quality control protocols or other similar algorithms and procedures to explicitly understand and minimize unwanted batch effects; and Domain Adaptation that develops DL tools that implicitly handle the batch effects by using approaches to achieve reliable and robust results. In this narrative review, we highlighted the advantages and disadvantages of both classes of DL approaches, and described key challenges to be addressed in future studies.

Cerebellar White Matter Abnormalities in Charcot-Marie-Tooth Disease: A Combined Volumetry and Diffusion Tensor Imaging Analysis

Charcot–Marie–Tooth disease (CMT) is a genetically heterogeneous hereditary peripheral neuropathy. Brain volumetry and diffusion tensor imaging (DTI) were performed in 47 controls and 47 CMT patients with PMP22 duplication (n = 10), MFN2 (n = 15), GJB1 (n = 11), or NEFL mutations (n = 11) to investigate for structural changes in the cerebellum. Volume of cerebellar white matter (WM) was significantly reduced in CMT patients with NEFL mutations. Abnormal DTI findings were observed in the superior, middle, and inferior cerebellar peduncles, predominantly in NEFL mutations and partly in GJB1 mutations. Cerebellar ataxia was more prevalent in the NEFL mutation group (72.7%) than the GJB1 mutation group (9.1%) but was not observed in other genotypic subtypes, which indicates that structural cerebellar abnormalities were associated with the presence of cerebellar ataxia. However, NEFL and GJB1 mutations did not affect cerebellar gray matter (GM), and neither cerebellar GM nor WM abnormalities were observed in the PMP22 duplication or MFN2 mutation groups. We found structural evidence of cerebellar WM abnormalities in CMT patients with NEFL and GJB1 mutations and an association between cerebellar WM involvement and cerebellar ataxia in these genetic subtypes, especially in the NEFL subgroup. Therefore, we suggest that neuroimaging, such as MRI volumetry or DTI, for CMT patients could play an important role in detecting abnormalities of cerebellar WM.

Source-based morphometry: a decade of covarying structural brain patterns

In this paper, we review and discuss brain imaging studies which have used the source-based morphometry (SBM) approach over the past decade. SBM is a data-driven linear multivariate approach for decomposing structural brain imaging data into commonly covarying imaging components and subject-specific loading parameters. It is a well-established technique which has predominantly been used to study neuroanatomic differences between healthy controls and patients with neuropsychiatric diseases. We start by discussing the advantages of this technique over univariate analysis for imaging studies, followed by a discussion of results from recent studies which have successfully applied this methodology. We also present recent extensions of this framework including nonlinear SBM, biclustered independent component analysis (B-ICA) and conclude with the possible directions of work for future.

Quantifying deep grey matter atrophy using automated segmentation approaches: A systematic review of structural MRI studies

The deep grey matter (DGM) nuclei of the brain play a crucial role in learning, behaviour, cognition, movement and memory. Although automated segmentation strategies can provide insight into the impact of multiple neurological conditions affecting these structures, such as Multiple Sclerosis (MS), Huntington's disease (HD), Alzheimer's disease (AD), Parkinson's disease (PD) and Cerebral Palsy (CP), there are a number of technical challenges limiting an accurate automated segmentation of the DGM. Namely, the insufficient contrast of T1 sequences to completely identify the boundaries of these structures, as well as the presence of iso-intense white matter lesions or extensive tissue loss caused by brain injury. Therefore in this systematic review, 269 eligible studies were analysed and compared to determine the optimal approaches for addressing these technical challenges. The automated approaches used among the reviewed studies fall into three broad categories, atlas-based approaches focusing on the accurate alignment of atlas priors, algorithmic approaches which utilise intensity information to a greater extent, and learning-based approaches that require an annotated training set. Studies that utilise freely available software packages such as FIRST, FreeSurfer and LesionTOADS were also eligible, and their performance compared. Overall, deep learning approaches achieved the best overall performance, however these strategies are currently hampered by the lack of large-scale annotated data. Improving model generalisability to new datasets could be achieved in future studies with data augmentation and transfer learning. Multi-atlas approaches provided the second-best performance overall, and may be utilised to construct a "silver standard" annotated training set for deep learning. To address the technical challenges, providing robustness to injury can be improved by using multiple channels, highly elastic diffeomorphic transformations such as LDDMM, and by following atlas-based approaches with an intensity driven refinement of the segmentation, which has been done with the Expectation Maximisation (EM) and level sets methods. Accounting for potential lesions should be achieved with a separate lesion segmentation approach, as in LesionTOADS. Finally, to address the issue of limited contrast, R2*, T2* and QSM sequences could be used to better highlight the DGM due to its higher iron content. Future studies could look to additionally acquire these sequences by retaining the phase information from standard structural scans, or alternatively acquiring these sequences for only a training set, allowing models to learn the "improved" segmentation from T1-sequences alone.

Apathy Is Related to Cognitive Control and Striatum Volumes in Prodromal Huntington's Disease

OBJECTIVES

Apathy is a debilitating symptom of Huntington's disease (HD) and manifests before motor diagnosis, making it an excellent therapeutic target in the preclinical phase of Huntington's disease (prHD). HD is a neurological genetic disorder characterized by cognitive and motor impairment, and psychiatric abnormalities. Apathy is not well characterized within the prHD. In previous literature, damage to the caudate and putamen has been correlated with increased apathy in other neurodegenerative and movement disorders. The objective of this study was to determine whether apathy severity in individuals with prHD is related to striatum volumes and cognitive control. We hypothesized that, within prHD individuals, striatum volumes and cognitive control scores would be related to apathy.

METHODS

We constructed linear mixed models to analyze striatum volumes and cognitive control, a composite measure that includes tasks assessing with apathy scores from 797 prHD participants. The outcome variable for each model was apathy, and the independent variables for the four separate models were caudate volume, putamen volume, cognitive control score, and motor symptom score. We also included depression as a covariate to ensure that our results were not solely related to mood.

RESULTS

Caudate and putamen volumes, as well as measures of cognitive control, were significantly related to apathy scores even after controlling for depression.

CONCLUSIONS

The behavioral apathy expressed by these individuals was related to regions of the brain commonly associated with isolated apathy, and not a direct result of mood symptoms. (JINS, 2019, 25, 462-469).

High and Low Levels of an NTRK2-Driven Genetic Profile Affect Motor- and Cognition-Associated Frontal Gray Matter in Prodromal Huntington's Disease

This study assessed how BDNF (brain-derived neurotrophic factor) and other genes involved in its signaling influence brain structure and clinical functioning in pre-diagnosis Huntington's disease (HD). Parallel independent component analysis (pICA), a multivariate method for identifying correlated patterns in multimodal datasets, was applied to gray matter concentration (GMC) and genomic data from a sizeable PREDICT-HD prodromal cohort (N = 715). pICA identified a genetic component highlighting NTRK2, which encodes BDNF's TrkB receptor, that correlated with a GMC component including supplementary motor, precentral/premotor cortex, and other frontal areas (p < 0.001); this association appeared to be driven by participants with high or low levels of the genetic profile. The frontal GMC profile correlated with cognitive and motor variables (Trail Making Test A (p = 0.03); Stroop Color (p = 0.017); Stroop Interference (p = 0.04); Symbol Digit Modalities Test (p = 0.031); Total Motor Score (p = 0.01)). A top-weighted NTRK2 variant (rs2277193) was protectively associated with Trail Making Test B (p = 0.007); greater minor allele numbers were linked to a better performance. These results support the idea of a protective role of NTRK2 in prodromal HD, particularly in individuals with certain genotypes, and suggest that this gene may influence the preservation of frontal gray matter that is important for clinical functioning.

Comparison of medical image classification accuracy among three machine learning methods

BACKGROUND

Low-quality medical images may influence the accuracy of the machine learning process.

OBJECTIVE

This study was undertaken to compare accuracy of medical image classification among machine learning methods, as classification is a basic aspect of clinical image inspection.

METHODS

Three types of machine learning methods were used, which include Support Vector Machine (SVM), Artificial Neural Network (ANN), and Convolution Neural Network (CNN). To investigate changes in accuracy related to image quality, we constructed a single dataset using two different file formats of DICOM (Digital Imaging and Communications in Medicine) and JPEG (Joint Photographic Experts Group).

RESULTS

The JPEG format contains less color information and data capacity than the DICOM format. CNN classification was accurate for both datasets, whereas SVM and ANN accuracy decreased with the loss of data from DICOM to JPEG formats.

CONCLUSIONS

CNN is more accurate than conventional machine learning methods that utilize the manual feature extraction.

Robust automated constellation-based landmark detection in human brain imaging

A robust fully automated algorithm for identifying an arbitrary number of landmark points in the human brain is described and validated. The proposed method combines statistical shape models with trained brain morphometric measures to estimate midbrain landmark positions reliably and accurately. Gross morphometric constraints provided by automatically identified eye centers and the center of the head mass are shown to provide robust initialization in the presence of large rotations in the initial head orientation. Detection of primary midbrain landmarks are used as the foundation from which extended detection of an arbitrary set of secondary landmarks in different brain regions by applying a linear model estimation and principle component analysis. This estimation model sequentially uses the knowledge of each additional detected landmark as an improved foundation for improved prediction of the next landmark location. The accuracy and robustness of the presented method was evaluated by comparing the automatically generated results to two manual raters on 30 identified landmark points extracted from each of 30 T1-weighted magnetic resonance images. For the landmarks with unambiguous anatomical definitions, the average discrepancy between the algorithm results and each human observer differed by less than 1 mm from the average inter-observer variability when the algorithm was evaluated on imaging data collected from the same site as the model building data. Similar results were obtained when the same model was applied to a set of heterogeneous image volumes from seven different collection sites representing 3 scanner manufacturers. This method is reliable for general application in large-scale multi-site studies that consist of a variety of imaging data with different orientations, spacings, origins, and field strengths.

Cognitive Control, Learning, and Clinical Motor Ratings Are Most Highly Associated with Basal Ganglia Brain Volumes in the Premanifest Huntington's Disease Phenotype

OBJECTIVES

Huntington's disease (HD) is a debilitating genetic disorder characterized by motor, cognitive and psychiatric abnormalities associated with neuropathological decline. HD pathology is the result of an extended chain of CAG (cytosine, adenine, guanine) trinucleotide repetitions in the HTT gene. Clinical diagnosis of HD requires the presence of an otherwise unexplained extrapyramidal movement disorder in a participant at risk for HD. Over the past 15 years, evidence has shown that cognitive, psychiatric, and subtle motor dysfunction is evident decades before traditional motor diagnosis. This study examines the relationships among subcortical brain volumes and measures of emerging disease phenotype in prodromal HD, before clinical diagnosis.

METHODS

The dataset includes 34 cognitive, motor, psychiatric, and functional variables and five subcortical brain volumes from 984 prodromal HD individuals enrolled in the PREDICT HD study. Using cluster analyses, seven distinct clusters encompassing cognitive, motor, psychiatric, and functional domains were identified. Individual cluster scores were then regressed against the subcortical brain volumetric measurements.

RESULTS

Accounting for site and genetic burden (the interaction of age and CAG repeat length) smaller caudate and putamen volumes were related to clusters reflecting motor symptom severity, cognitive control, and verbal learning.

CONCLUSIONS

Variable reduction of the HD phenotype using cluster analysis revealed biologically related domains of HD and are suitable for future research with this population. Our cognitive control cluster scores show sensitivity to changes in basal ganglia both within and outside the striatum that may not be captured by examining only motor scores. (JINS, 2017, 23, 159-170).

Patterns of Co-Occurring Gray Matter Concentration Loss across the Huntington Disease Prodrome

Huntington disease (HD) is caused by an abnormally expanded cytosine-adenine-guanine (CAG) trinucleotide repeat in the HTT gene. Age and CAG-expansion number are related to age at diagnosis and can be used to index disease progression. However, observed onset-age variability suggests that other factors also modulate progression. Indexing prodromal (pre-diagnosis) progression may highlight therapeutic targets by isolating the earliest-affected factors. We present the largest prodromal HD application of the univariate method voxel-based morphometry (VBM) and the first application of the multivariate method source-based morphometry (SBM) to, respectively, compare gray matter concentration (GMC) and capture co-occurring GMC patterns in control and prodromal participants. Using structural MRI data from 1050 (831 prodromal, 219 control) participants, we characterize control-prodromal, whole-brain GMC differences at various prodromal stages. Our results provide evidence for (1) regional co-occurrence and differential patterns of decline across the prodrome, with parietal and occipital differences commonly co-occurring, and frontal and temporal differences being relatively independent from one another, (2) fronto-striatal circuits being among the earliest and most consistently affected in the prodrome, (3) delayed degradation in some movement-related regions, with increasing subcortical and occipital differences with later progression, (4) an overall superior-to-inferior gradient of GMC reduction in frontal, parietal, and temporal lobes, and (5) the appropriateness of SBM for studying the prodromal HD population and its enhanced sensitivity to early prodromal and regionally concurrent differences.

Linking white matter and deep gray matter alterations in premanifest Huntington disease

Huntington disease (HD) is a fatal progressive neurodegenerative disorder for which only symptomatic treatment is available. A better understanding of the pathology, and identification of biomarkers will facilitate the development of disease-modifying treatments. HD is potentially a good model of a neurodegenerative disease for development of biomarkers because it is an autosomal-dominant disease with complete penetrance, caused by a single gene mutation, in which the neurodegenerative process can be assessed many years before onset of signs and symptoms of manifest disease. Previous MRI studies have detected abnormalities in gray and white matter starting in premanifest stages. However, the understanding of how these abnormalities are related, both in time and space, is still incomplete. In this study, we combined deep gray matter shape diffeomorphometry and white matter DTI analysis in order to provide a better mapping of pathology in the deep gray matter and subcortical white matter in premanifest HD. We used 296 MRI scans from the PREDICT-HD database. Atrophy in the deep gray matter, thalamus, hippocampus, and nucleus accumbens was analyzed by surface based morphometry, and while white matter abnormalities were analyzed in (i) regions of interest surrounding these structures, using (ii) tractography-based analysis, and using (iii) whole brain atlas-based analysis. We detected atrophy in the deep gray matter, particularly in putamen, from early premanifest stages. The atrophy was greater both in extent and effect size in cases with longer exposure to the effects of the CAG expansion mutation (as assessed by greater CAP-scores), and preceded detectible abnormalities in the white matter. Near the predicted onset of manifest HD, the MD increase was widespread, with highest indices in the deep and posterior white matter. This type of in-vivo macroscopic mapping of HD brain abnormalities can potentially indicate when and where therapeutics could be targeted to delay the onset or slow the disease progression.

Neuroimaging of a minipig model of Huntington's disease: Feasibility of volumetric, diffusion-weighted and spectroscopic assessments

BACKGROUND

As novel treatment approaches for Huntington's disease (HD) evolve, the use of transgenic (tg) large animal models has been considered for preclinical safety and efficacy assessments. It is hoped that large animal models may provide higher reliability in translating preclinical findings to humans, e.g., by using similar endpoints and biomarkers.

NEW METHOD

We here investigated the feasibility to conduct MRI assessments in a recently developed tgHD model in the Libechov minipig. The model is characterized by high genetic homology to humans and a similar body mass and compartments. The minipig brain provides anatomical features that are attractive for imaging studies and could be used as endpoints for disease modifying preclinical studies similar to human HD.

RESULTS

We demonstrate that complex MRI protocols can be successfully acquired with tgHD and wild type (wt) Libechov minipigs. We show that acquisition of anatomical images applicable for volumetric assessments is feasible and outline the development of a segmented MRI brain atlas. Similarly diffusion-weighted imaging (DWI) including fiber tractography is presented. We also demonstrate the feasibility to conduct in vivo metabolic assessments using MR spectroscopy.

COMPARISON WITH EXISTING METHODS

In human HD, these MRI methods are already validated and used as reliable biomarker of disease progression even before the onset of a clinical motor phenotype.

CONCLUSIONS

The results show that the minipig brain is well suited for MRI assessments in preclinical studies. We conclude that further characterization of phenotypical differences between tg and wt animals in sufficiently powered cross-sectional and longitudinal studies is warranted.

Preliminary analysis using multi-atlas labeling algorithms for tracing longitudinal change

Multicenter longitudinal neuroimaging has great potential to provide efficient and consistent biomarkers for research of neurodegenerative diseases and aging. In rare disease studies it is of primary importance to have a reliable tool that performs consistently for data from many different collection sites to increase study power. A multi-atlas labeling algorithm is a powerful brain image segmentation approach that is becoming increasingly popular in image processing. The present study examined the performance of multi-atlas labeling tools for subcortical identification using two types of in-vivo image database: Traveling Human Phantom (THP) and PREDICT-HD. We compared the accuracy (Dice Similarity Coefficient; DSC and intraclass correlation; ICC), multicenter reliability (Coefficient of Variance; CV), and longitudinal reliability (volume trajectory smoothness and Akaike Information Criterion; AIC) of three automated segmentation approaches: two multi-atlas labeling tools, MABMIS and MALF, and a machine-learning-based tool, BRAINSCut. In general, MALF showed the best performance (higher DSC, ICC, lower CV, AIC, and smoother trajectory) with a couple of exceptions. First, the results of accumben, where BRAINSCut showed higher reliability, were still premature to discuss their reliability levels since their validity is still in doubt (DSC < 0.7, ICC < 0.7). For caudate, BRAINSCut presented slightly better accuracy while MALF showed significantly smoother longitudinal trajectory. We discuss advantages and limitations of these performance variations and conclude that improved segmentation quality can be achieved using multi-atlas labeling methods. While multi-atlas labeling methods are likely to help improve overall segmentation quality, caution has to be taken when one chooses an approach, as our results suggest that segmentation outcome can vary depending on research interest.