Can FreeSurfer Compete with Manual Volumetric Measurements in Alzheimer's Disease?

Proof-of-concept recall-by-genotype study of extremely low and high Alzheimer's polygenic risk reveals autobiographical deficits and cingulate cortex correlates

BACKGROUND

Genome-wide association studies demonstrate that Alzheimer's disease (AD) has a highly polygenic architecture, where thousands of independent genetic variants explain risk with high classification accuracy. This AD polygenic risk score (AD-PRS) has been previously linked to preclinical cognitive and neuroimaging features observed in asymptomatic individuals. However, shared variance between AD-PRS and neurocognitive features are small, suggesting limited preclinical utility.

METHODS

Here, we recruited sixteen clinically asymptomatic individuals (mean age 67; range 58-76) with either extremely low / high AD-PRS (defined as at least 2 standard deviations from the wider sample mean (N = 4504; N EFFECTIVE = 90)) with comparable age sex and education level. We assessed group differences in autobiographical memory and T1-weighted structural neuroimaging features.

RESULTS

We observed marked reductions in autobiographical recollection (Cohen's d =  - 1.66; P FDR = 0.014) and midline structure (cingulate) thickness (Cohen's d =  - 1.55, P FDR = 0.05), with no difference in hippocampal volume (P > 0.3). We further confirm the negative association between AD-PRS and cingulate thickness in a larger study with a comparable age (N = 31,966, β =  - 0.002, P = 0.011), supporting the validity of our approach.

CONCLUSIONS

These observations conform with multiple streams of prior evidence suggesting alterations in cingulate structures may occur in individuals with higher AD genetic risk. We were able to use a genetically informed research design strategy that significantly improved the efficiency and power of the study. Thus, we further demonstrate that the recall-by-genotype of AD-PRS from wider samples is a promising approach for the detection, assessment, and intervention in specific individuals with increased AD genetic risk.

Considerations on brain age predictions from repeatedly sampled data across time

INTRODUCTION

Brain age, the estimation of a person's age from magnetic resonance imaging (MRI) parameters, has been used as a general indicator of health. The marker requires however further validation for application in clinical contexts. Here, we show how brain age predictions perform for the same individual at various time points and validate our findings with age-matched healthy controls.

METHODS

We used densely sampled T1-weighted MRI data from four individuals (from two densely sampled datasets) to observe how brain age corresponds to age and is influenced by acquisition and quality parameters. For validation, we used two cross-sectional datasets. Brain age was predicted by a pretrained deep learning model.

RESULTS

We found small within-subject correlations between age and brain age. We also found evidence for the influence of field strength on brain age which replicated in the cross-sectional validation data and inconclusive effects of scan quality.

CONCLUSION

The absence of maturation effects for the age range in the presented sample, brain age model bias (including training age distribution and field strength), and model error are potential reasons for small relationships between age and brain age in densely sampled longitudinal data. Clinical applications of brain age models should consider of the possibility of apparent biases caused by variation in the data acquisition process.

3D Capsule Networks for Brain Image Segmentation

BACKGROUND AND PURPOSE

Current autosegmentation models such as UNets and nnUNets have limitations, including the inability to segment images that are not represented during training and lack of computational efficiency. 3D capsule networks have the potential to address these limitations.

MATERIALS AND METHODS

We used 3430 brain MRIs, acquired in a multi-institutional study, to train and validate our models. We compared our capsule network with standard alternatives, UNets and nnUNets, on the basis of segmentation efficacy (Dice scores), segmentation performance when the image is not well-represented in the training data, performance when the training data are limited, and computational efficiency including required memory and computational speed.

RESULTS

The capsule network segmented the third ventricle, thalamus, and hippocampus with Dice scores of 95%, 94%, and 92%, respectively, which were within 1% of the Dice scores of UNets and nnUNets. The capsule network significantly outperformed UNets in segmenting images that were not well-represented in the training data, with Dice scores 30% higher. The computational memory required for the capsule network is less than one-tenth of the memory required for UNets or nnUNets. The capsule network is also >25% faster to train compared with UNet and nnUNet.

CONCLUSIONS

We developed and validated a capsule network that is effective in segmenting brain images, can segment images that are not well-represented in the training data, and is computationally efficient compared with alternatives.

Comparing 3D, 2.5D, and 2D Approaches to Brain Image Auto-Segmentation

Deep-learning methods for auto-segmenting brain images either segment one slice of the image (2D), five consecutive slices of the image (2.5D), or an entire volume of the image (3D). Whether one approach is superior for auto-segmenting brain images is not known. We compared these three approaches (3D, 2.5D, and 2D) across three auto-segmentation models (capsule networks, UNets, and nnUNets) to segment brain structures. We used 3430 brain MRIs, acquired in a multi-institutional study, to train and test our models. We used the following performance metrics: segmentation accuracy, performance with limited training data, required computational memory, and computational speed during training and deployment. The 3D, 2.5D, and 2D approaches respectively gave the highest to lowest Dice scores across all models. 3D models maintained higher Dice scores when the training set size was decreased from 3199 MRIs down to 60 MRIs. 3D models converged 20% to 40% faster during training and were 30% to 50% faster during deployment. However, 3D models require 20 times more computational memory compared to 2.5D or 2D models. This study showed that 3D models are more accurate, maintain better performance with limited training data, and are faster to train and deploy. However, 3D models require more computational memory compared to 2.5D or 2D models.

Transcriptomic Signatures Associated With Regional Cortical Thickness Changes in Parkinson's Disease

Cortical atrophy is a common manifestation in Parkinson's disease (PD), particularly in advanced stages of the disease. To elucidate the molecular underpinnings of cortical thickness changes in PD, we performed an integrated analysis of brain-wide healthy transcriptomic data from the Allen Human Brain Atlas and patterns of cortical thickness based on T1-weighted anatomical MRI data of 149 PD patients and 369 controls. For this purpose, we used partial least squares regression to identify gene expression patterns correlated with cortical thickness changes. In addition, we identified gene expression patterns underlying the relationship between cortical thickness and clinical domains of PD. Our results show that genes whose expression in the healthy brain is associated with cortical thickness changes in PD are enriched in biological pathways related to sumoylation, regulation of mitotic cell cycle, mitochondrial translation, DNA damage responses, and ER-Golgi traffic. The associated pathways were highly related to each other and all belong to cellular maintenance mechanisms. The expression of genes within most pathways was negatively correlated with cortical thickness changes, showing higher expression in regions associated with decreased cortical thickness (atrophy). On the other hand, sumoylation pathways were positively correlated with cortical thickness changes, showing higher expression in regions with increased cortical thickness (hypertrophy). Our findings suggest that alterations in the balanced interplay of these mechanisms play a role in changes of cortical thickness in PD and possibly influence motor and cognitive functions.

Validity and reliability of the medial temporal lobe atrophy scale in a memory clinic population

Background

Visual rating of medial temporal lobe atrophy (MTA) is often performed in conjunction with dementia workup. Most prior studies involved patients with known or probable Alzheimer’s disease (AD). This study investigated the validity and reliability of MTA in a memory clinic population.

Methods

MTA was rated in 752 MRI examinations, of which 105 were performed in cognitively healthy participants (CH), 184 in participants with subjective cognitive impairment, 249 in subjects with mild cognitive impairment, and 214 in patients with dementia, including AD, subcortical vascular dementia and mixed dementia. Hippocampal volumes, measured manually or using FreeSurfer, were available in the majority of cases. Intra- and interrater reliability was tested using Cohen’s weighted kappa. Correlation between MTA and quantitative hippocampal measurements was ascertained with Spearman’s rank correlation coefficient. Moreover, diagnostic ability of MTA was assessed with receiver operating characteristic (ROC) analysis and suitable, age-dependent MTA thresholds were determined.

Results

Rater agreement was moderate to substantial. MTA correlation with quantitative volumetric methods ranged from -0.20 (p< 0.05) to -0.68 (p < 0.001) depending on the quantitative method used. Both MTA and FreeSurfer are able to distinguish dementia subgroups from CH. Suggested age-dependent MTA thresholds are 1 for the age group below 75 years and 1.5 for the age group 75 years and older.

Conclusions

MTA can be considered a valid marker of medial temporal lobe atrophy and may thus be valuable in the assessment of patients with cognitive impairment, even in a heterogeneous patient population.

Contributions of Cerebro-Cerebellar Default Mode Connectivity Patterns to Memory Performance in Mild Cognitive Impairment

BACKGROUND

The cerebral default mode network (DMN) can be mapped onto specific regions in the cerebellum, which are specifically vulnerable to atrophy in Alzheimer's disease (AD) patients.

OBJECTIVE

We set out to determine whether there are specific differences in the interaction between the cerebral and cerebellar DMN in amnestic mild cognitive impairment (aMCI) patients compared to healthy controls using resting-state functional MRI and whether these differences are relevant for memory performance.

METHODS

Eighteen patients with aMCI were age and education-matched to eighteen older adults and underwent 3T MR-imaging. We performed seed-based functional connectivity analysis between the cerebellar DMN seeds and the cerebral DMN.

RESULTS

Our results showed that compared to healthy older adults, aMCI patients showed lower anti-correlation between the cerebellar DMN and several cerebral DMN regions. Additionally, we showed that degradation of the anti-correlation between the cerebellar DMN and the medial frontal cortex is correlated with worse memory performance in aMCI patients.

CONCLUSION

These findings provide evidence that the cerebellar DMN and cerebral DMN are negatively correlated during rest in older individuals, and suggest that the reduced anti-correlated impacts the modulatory role of the cerebellum on cognitive functioning, in particular on the executive component of memory functions in neurodegenerative diseases.

Regional brain volumetric changes despite 2 years of treatment initiated during acute HIV infection

OBJECTIVE

To assess changes in regional brain volumes after 24 months among individuals who initiated combination antiretroviral therapy (cART) within weeks of HIV exposure.

DESIGN

Prospective cohort study of Thai participants in the earliest stages of HIV-1infection.

METHODS

Thirty-four acutely HIV-infected individuals (AHI; Fiebig I-V) underwent brain magnetic resonance (MR) imaging and MR spectroscopy at 1.5 T and immediately initiated cART. Imaging was repeated at 24 months. Regional brain volumes were quantified using FreeSurfer's longitudinal pipeline. Voxel-wise analyses using tensor-based morphometry (TBM) were conducted to verify regional assessments. Baseline brain metabolite levels, blood and cerebrospinal fluid biomarkers assessed by ELISA, and peripheral blood monocyte phenotypes measured by flow cytometry were examined as predictors of significant volumetric change.

RESULTS

Participants were 31 ± 8 years old. The estimated mean duration of infection at cART initiation was 15 days. Longitudinal analyses revealed reductions in volumes of putamen (P < 0.001) and caudate (P = 0.006). TBM confirmed significant atrophy in the putamen and caudate, and also in thalamic and hippocampal regions. In exploratory post-hoc analyses, higher baseline frequency of P-selectin glycoprotein ligand-1 (PSGL-1)-expressing total monocytes correlated with greater caudate volumetric decrease (ρ = 0.67, P = 0.017), whereas the baseline density of PSGL-1-expressing inflammatory (CD14CD16) monocytes correlated with putamen atrophy (ρ = 0.65, P = 0.022).

CONCLUSION

Suppressive cART initiated during AHI may not prevent brain atrophy. Volumetric decrease appears greater than expected age-related decline, although examination of longitudinal change in demographically similar HIV-uninfected Thai individuals is needed. Mechanisms underlying progressive HIV-related atrophy may include early activation and enhanced adhesive and migratory capacity of circulating monocyte populations.

Brain volumes and their ratios in Alzheimer´s disease on magnetic resonance imaging segmented using Freesurfer 6.0

Ratios between opposing volumes from brain magnetic resonance imaging (MRI) can provide additional information to volumes in Alzheimer's disease (AD). Brain three-dimensional MPRAGE MRI at 3T were segmented into 44 regions using FreeSurfer v6 in 75 participants. The region's size in absolute volumes and relative proportions to the whole brain volume were compared between 39 AD patients and 36 age-, education- and sex-matched normal controls (NC). Volumes of the most atrophied parts were related to the opposing volumes of the most enlarged parts as ratios. The most atrophic structures in AD were both hippocampi. By contrast, the greatest enlargements in AD were inferior parts of both lateral ventricles. The best ratio for each side was the hippocampo-horn proportion calculated as ratio: the hippocampus / (the hippocampus + inferior lateral ventricle). Its optimal cut-off of 74% yielded sensitivity of 74% and specificity of 78% on the left and sensitivity of 74% and specificity of 78% on the right. The hippocampo-horn proportion is another measure to evaluate the degree of hippocampal atrophy on brain MRI in percentages. It has a potential to be simplified into a comparison of two-dimensional corresponding areas or a visual assessment.

Characterizing Fatigue-Related White Matter Changes in MS: A Proton Magnetic Resonance Spectroscopy Study

Few cross-sectional studies have investigated the correlation between neurochemical changes and multiple sclerosis (MS) fatigue, but little is known on the fatigue-related white matter differences between time points. We aim to investigate the longitudinal neurometabolite profile of white matter in MS fatigue. Forty-eight relapsing remitting multiple sclerosis (RRMS) patients with an expanded disability status scale (EDSS) ≤ 4 underwent high field ¹H-multivoxel magnetic resonance spectroscopy (MRS) at baseline and year 1. Fatigue severity was evaluated by the fatigue severity scale (FSS). Patients were divided into low (LF, FSS ≤ 3), moderate (MF, FSS = 3.1–5), and high fatigue (HF, FSS ≥ 5.1) groups. In a two-way analysis of variance (ANOVA), we observed a decline in the ratio of the sum of N-acetylaspartate (NAA) and N-acetylaspartylglutamate (NAAG) to the sum of creatine (Cr) and phosphocreatine (PCr) in the right anterior quadrant (RAQ) and left anterior quadrant (LAQ) of the MRS grid in the HF group at baseline and year 1. This decline was significant when compared with the LF group (p = 0.018 and 0.020). In a one-way ANOVA, the fatigue group effect was significant and the ratio difference in the right posterior quadrant (RPQ) and left posterior quadrant (LPQ) of the HF group was also significant (p = 0.012 and 0.04). Neurochemical changes in the bilateral frontal white matter and possibly parietooccipital areas were noted in the HF group at two different time points. Our findings may shed some light on the pathology of MS fatigue.

Sex-specific hippocampus volume changes in obstructive sleep apnea

Introduction

Obstructive sleep apnea (OSA) patients show hippocampal-related autonomic and neurological symptoms, including impaired memory and depression, which differ by sex, and are mediated in distinct hippocampal subfields. Determining sites and extent of hippocampal sub-regional injury in OSA could reveal localized structural damage linked with OSA symptoms.

Methods

High-resolution T1-weighted images were collected from 66 newly-diagnosed, untreated OSA (mean age ± SD: 46.3 ± 8.8 years; mean AHI ± SD: 34.1 ± 21.5 events/h;50 male) and 59 healthy age-matched control (46.8 ± 9.0 years;38 male) participants. We added age-matched controls with T1-weighted scans from two datasets (IXI, OASIS-MRI), for 979 controls total (426 male/46.5 ± 9.9 years). We segmented the hippocampus and analyzed surface structure with “FSL FIRST” software, scaling volumes for brain size, and evaluated group differences with ANCOVA (covariates: total-intracranial-volume, sex; P < .05, corrected).

Results

In OSA relative to controls, the hippocampus showed small areas larger volume bilaterally in CA1 (surface displacement ≤0.56 mm), subiculum, and uncus, and smaller volume in right posterior CA3/dentate (≥ − 0.23 mm). OSA vs. control males showed higher bilateral volume (≤0.61 mm) throughout CA1 and subiculum, extending to head and tail, with greater right-sided increases; lower bilateral volumes (≥ − 0.45 mm) appeared in mid- and posterior-CA3/dentate. OSA vs control females showed only right-sided effects, with increased CA1 and subiculum/uncus volumes (≤0.67 mm), and decreased posterior CA3/dentate volumes (≥ − 0.52 mm). Unlike males, OSA females showed volume decreases in the right hippocampus head and tail.

Conclusions

The hippocampus shows lateralized and sex-specific, OSA-related regional volume differences, which may contribute to sex-related expression of symptoms in the sleep disorder. Volume increases suggest inflammation and glial activation, whereas volume decreases suggest long-lasting neuronal injury; both processes may contribute to dysfunction in OSA.

•

The hippocampus in OSA shows areas of increased and decreased volume.

•

The injury is sex-specific, in subregions related to symptoms in females and males.

•

Injury may be inflammation (volume increases) or cell death (volume decreases).

Relationship of Hippocampal Volumes and Posttraumatic Stress Disorder Symptoms Over Early Posttrauma Periods

BACKGROUND

Smaller hippocampal volume is associated with more severe posttraumatic stress disorder (PTSD) symptoms years after traumatic experiences. Posttraumatic stress symptoms appear early following trauma, but the relationship between hippocampal volume and PTSD symptom severity during early posttrauma periods is not well understood. It is possible that the inverse relationship between hippocampal volume and PTSD symptom severity is already present soon after trauma. To test this possibility, we prospectively examined the association between hippocampal volumes and severity of PTSD symptoms within weeks to months after trauma due to a motor vehicle collision.

METHODS

Structural magnetic resonance imaging scans of 44 survivors were collected about 2 weeks and again at 3 months after a motor vehicle collision to measure hippocampal volumes. The PTSD Checklist was used to evaluate PTSD symptoms at each scan time. Full (n = 5) or partial (n = 6) PTSD was evaluated using the Clinician-Administered PTSD Scale at 3 months.

RESULTS

Left hippocampal volumes at both time points negatively correlated with PTSD Checklist scores, and with subscores for re-experiencing symptoms at 3 months. Left hippocampal volumes at 3 months also negatively correlated with hyperarousal symptoms at 3 months. Finally, neither left nor right hippocampal volumes significantly changed between 2 weeks and 3 months posttrauma.

CONCLUSIONS

The results suggest that small hippocampal volume at early posttrauma weeks is associated with increased risk for PTSD development. Furthermore, the inverse relationship between hippocampal volume and PTSD symptoms at 3 months did not arise from posttrauma shifts in hippocampal volume between 2 weeks and 3 months after trauma.

Thalamic Volume Is Reduced in Cervical and Laryngeal Dystonias

BACKGROUND

Dystonia, a debilitating movement disorder characterized by abnormal fixed positions and/or twisting postures, is associated with dysfunction of motor control networks. While gross brain lesions can produce secondary dystonias, advanced neuroimaging techniques have been required to identify network abnormalities in primary dystonias. Prior neuroimaging studies have provided valuable insights into the pathophysiology of dystonia, but few directly assessed the gross volume of motor control regions, and to our knowledge, none identified abnormalities common to multiple types of idiopathic focal dystonia.

METHODS

We used two gross volumetric segmentation techniques and one voxelwise volumetric technique (voxel based morphometry, VBM) to compare regional volume between matched healthy controls and patients with idiopathic primary focal dystonia (cervical, n = 17, laryngeal, n = 7). We used (1) automated gross volume measures of eight motor control regions using the FreeSurfer analysis package; (2) blinded, anatomist-supervised manual segmentation of the whole thalamus (also gross volume); and (3) voxel based morphometry, which measures local T1-weighted signal intensity and estimates gray matter density or volume at the level of single voxels, for both whole-brain and thalamus.

RESULTS

Using both automated and manual gross volumetry, we found a significant volume decrease only in the thalamus in two focal dystonias. Decreases in whole-thalamic volume were independent of head and brain size, laterality of symptoms, and duration. VBM measures did not differ between dystonia and control groups in any motor control region.

CONCLUSIONS

Reduced thalamic gross volume, detected in two independent analyses, suggests a common anatomical abnormality in cervical dystonia and spasmodic dysphonia. Defining the structural underpinnings of dystonia may require such complementary approaches.

Combining multiple anatomical MRI measures improves Alzheimer's disease classification

Several anatomical MRI markers for Alzheimer's disease (AD) have been identified. Hippocampal volume, cortical thickness, and grey matter density have been used successfully to discriminate AD patients from controls. These anatomical MRI measures have so far mainly been used separately. The full potential of anatomical MRI scans for AD diagnosis might thus not yet have been used optimally. In this study, we therefore combined multiple anatomical MRI measures to improve diagnostic classification of AD. For 21 clinically diagnosed AD patients and 21 cognitively normal controls, we calculated (i) cortical thickness, (ii) cortical area, (iii) cortical curvature, (iv) grey matter density, (v) subcortical volumes, and (vi) hippocampal shape. These six measures were used separately and combined as predictors in an elastic net logistic regression. We made receiver operating curve plots and calculated the area under the curve (AUC) to determine classification performance. AUC values for the single measures ranged from 0.67 (cortical thickness) to 0.94 (grey matter density). The combination of all six measures resulted in an AUC of 0.98. Our results demonstrate that the different anatomical MRI measures contain complementary information. A combination of these measures may therefore improve accuracy of AD diagnosis in clinical practice. Hum Brain Mapp 37:1920-1929, 2016. © 2016 Wiley Periodicals, Inc.

Measuring Global Brain Atrophy with the Brain Volume/Cerebrospinal Fluid Index: Normative Values, Cut-Offs and Clinical Associations

BACKGROUND

Global brain atrophy is present in normal aging and different neurodegenerative disorders such as Alzheimer's disease (AD) and is becoming widely used to monitor disease progression.

SUMMARY

The brain volume/cerebrospinal fluid index (BV/CSF index) is validated in this study as a measurement of global brain atrophy. We tested the ability of the BV/CSF index to detect global brain atrophy, investigated the influence of confounders, provided normative values and cut-offs for mild, moderate and severe brain atrophy, and studied associations with different outcome variables. A total of 1,009 individuals were included [324 healthy controls, 408 patients with mild cognitive impairment (MCI) and 277 patients with AD]. Magnetic resonance images were segmented using FreeSurfer, and the BV/CSF index was calculated and studied both cross-sectionally and longitudinally (1-year follow-up). Both AD patients and MCI patients who progressed to AD showed greater global brain atrophy compared to stable MCI patients and controls. Atrophy was associated with older age, larger intracranial volume, less education and presence of the ApoE ε4 allele. Significant correlations were found with clinical variables, CSF biomarkers and several cognitive tests.

KEY MESSAGES

The BV/CSF index may be useful for staging individuals according to the degree of global brain atrophy, and for monitoring disease progression. It also shows potential for predicting clinical changes and for being used in the clinical routine.

Structural Image Analysis of the Brain in Neuropsychology Using Magnetic Resonance Imaging (MRI) Techniques

Magnetic resonance imaging (MRI) of the brain provides exceptional image quality for visualization and neuroanatomical classification of brain structure. A variety of image analysis techniques provide both qualitative as well as quantitative methods to relate brain structure with neuropsychological outcome and are reviewed herein. Of particular importance are more automated methods that permit analysis of a broad spectrum of anatomical measures including volume, thickness and shape. The challenge for neuropsychology is which metric to use, for which disorder and the timing of when image analysis methods are applied to assess brain structure and pathology. A basic overview is provided as to the anatomical and pathoanatomical relations of different MRI sequences in assessing normal and abnormal findings. Some interpretive guidelines are offered including factors related to similarity and symmetry of typical brain development along with size-normalcy features of brain anatomy related to function. The review concludes with a detailed example of various quantitative techniques applied to analyzing brain structure for neuropsychological outcome studies in traumatic brain injury.

MIR137HG risk variant rs1625579 genotype is related to corpus callosum volume in schizophrenia

Genome-wide association studies implicate the MIR137HG risk variant rs1625579 (MIR137HGrv) within the host gene for microRNA-137 as a potential regulator of schizophrenia susceptibility. We examined the influence of MIR137HGrv genotype on 17 subcortical and callosal volumes in a large sample of individuals with schizophrenia and healthy controls (n=841). Although the volumes were overall reduced relative to healthy controls, for individuals with schizophrenia the homozygous MIR137HGrv risk genotype was associated with attenuated reduction of mid-posterior corpus callosum volume (p=0.001), along with trend-level effects in the adjacent central and posterior corpus callosum. These findings are unique in the literature and remain robust after analysis in ethnically homogenous and single-scanner subsets of the larger sample. Thus, our study suggests that the mechanisms whereby MIR137HGrv works to increase schizophrenia risk are not those that generate the corpus callosum volume reductions commonly found in the disorder.