Search

The hippocampus is closely related to many brain diseases, such as Alzheimer's disease. Accurate measurement of the hippocampus is helpful for clinicians in identifying lesions and then diagnosing and treating the related brain diseases. Therefore, accurate segmentation of the hippocampus is of vital significance for the in-depth study of many brain diseases. However, the accurate measurement of the hippocampus depends on its accurate segmentation, and hippocampal segmentation has always been a challenging problem due to the small size, irregular shape, and fuzzy boundaries with surrounding tissues of the hippocampus. With the development of machine learning, many innovative methods have been proposed to segment the hippocampus. The purpose of this survey is to provide a comprehensive overview of hippocampal segmentation in brain MRI images using machine learning methods. First, a brief introduction to hippocampal segmentation in brain MRI images is given. Then, common evaluation metrics of hippocampal segmentation are introduced. Next, brain hippocampal segmentation methods based on traditional machine learning and deep learning are described. Subsequently, some common open datasets and toolkits applied to brain hippocampal segmentation are presented. Finally, objective conclusions regarding hippocampal segmentation in brain MRI images using machine learning methods are drawn, and future developments and trends are identified for brain hippocampal segmentation.

BACKGROUND AND PURPOSE: The hippocampus is a frequent focus of quantitative neuroimaging research, and structural hippocampal alterations are related to multiple neurocognitive disorders. An increasing number of neuroimaging studies are focusing on hippocampal subfield regional involvement in these disorders using various automated segmentation approaches. Direct comparisons among these approaches are limited. The purpose of this study was to compare the agreement between two automated hippocampal segmentation algorithms in an adult population. MATERIALS AND METHODS: We compared the results of 2 automated segmentation algorithms for hippocampal subfields (FreeSurfer v6.0 and volBrain) within a single imaging data set from adults (n = 176, 89 women) across a wide age range (20–79 years). Brain MR imaging was acquired on a single 3T scanner as part of the IXI Brain Development Dataset and included T1- and T2-weighted MR images. We also examined subfield volumetric differences related to age and sex and the impact of different intracranial volume and total hippocampal volume normalization methods. RESULTS: Estimated intracranial volume and total hippocampal volume of both protocols were strongly correlated (r = 0.93 and 0.9, respectively; both P < .001). Hippocampal subfield volumes were correlated (ranging from r = 0.42 for the subiculum to r = 0.78 for the cornu ammonis [CA]1, all P < .001). However, absolute volumes were significantly different between protocols. volBrain produced larger CA1 and CA4-dentate gyrus and smaller CA2-CA3 and subiculum volumes compared with FreeSurfer v6.0. Regional age- and sex-related differences in subfield volumes were qualitatively and quantitatively different depending on segmentation protocol and intracranial volume/total hippocampal volume normalization method. CONCLUSIONS: The hippocampal subfield volume relationship to demographic factors and disease states should undergo nuanced interpretation, especially when considering different segmentation protocols.

Background In vivo measurement of hippocampal volume with magnetic resonance imaging (MRI) has become an important element in neuroimaging research. However, hippocampal volumetric findings and their relationship with cardiovascular risk factors and memory performance are still controversial and inconsistent for non-demented adults. Purpose To compare total and regional hippocampal volumes from manual tracing and automated Freesurfer segmentation methods and their relationship with mid-life clinical data and late-life verbal episodic memory performance in older women. Material and Methods This study used structural MRI datasets from 161 women who were scanned in 2012 and underwent neuropsychological assessments. Of these participants, 135 women had completed baseline measures of cardiovascular risk factors in 1992. Results Our results showed a significant correlation between manual tracing and automated Freesurfer output segmentations of total (r = 0.71), anterior (r = 0.65), and posterior (r = 0.38) hippocampal volumes. Mid-life Framingham Cardiovascular Risk Profile score is not associated with late-life hippocampal volumes, adjusted for intracranial volume, age, education, and apolipoprotein E gene ε4 status. Anterior hippocampal volume segmented either with manual tracing or automated Freesurfer software is sensitive to changes in mid-life high-density lipoprotein (HDL) cholesterol level, while posterior hippocampal volume is linked with verbal episodic memory performance in elderly women. Conclusion These findings support the use of Freesurfer automated segmentation measures for large datasets as being highly correlated with the manual tracing method. In addition, our results suggest intervention strategies that target mid-life HDL cholesterol level in women.

Hippocampus is one of the first affected brain regions in Alzheimer's disease (AD). To help AD diagnosis, hippocampal atrophy have been studied extensively, but accurate segmentation of hippocampus has always been a difficult problem. In this study, we propose a 3D multi-scale multi-attention UNet with the use of multi-scale inception module, residual connections and attention modules for automatic hippocampal segmentation. Experimental results on the HarP dataset show that our proposed method can achieve average Dice coefficient of 0.827 for hippocampal segmentation, which outperform classic 3D UNet. The efficiency and accuracy of the proposed methods suggests that it may facilitate the study of the hippocampus for large-scale studies.

The assessment of the hippocampus is indispensable for the supplementary diagnosis of Alzheimer's disease. In response to the small size of the hippocampus and the difficulty of accurate segmentation, this paper proposes a 3DUnet-CBAM based hippocampal segmentation model. A 3DUnet convolutional network is used to avoid information loss between two-dimensional slices, and a CBAM attention mechanism is incorporated in the last layer of downsampling. The deep and superficial features of the hippocampus are effectively combined. The segmentation accuracy of the hippocampus is finally improved. Experimental results show that the model achieves 89.01%, 89.04% and 88.97% for dice, precision and recall under the validation set, respectively, which can better assist doctors in diagnosing Alzheimer's disease.

Hippocampal segmentation from infant brain MR images is indispensable for studying early brain development. However, most of the hippocampal segmentation methods were developed for population-based adult brain images, which are not suitable for longitudinal infant brain images acquired in the first year of life due to the low image contrast and variable development patterns of the hippocampal structure. To address these challenges, we propose a classification-guided boundary regression method to first detect hippocampal boundaries in the longitudinal infant brain images and then use those detected boundaries to guide the deformable model for final segmentation. Specifically, we first employ a classification-guided regression forest to predict the 3D displacements from individual image voxels to the potential hippocampal boundaries. These predicted displacements then determine the boundary maps by a voting strategy. Second, we iteratively enhance the voted hippocampal boundary map by incorporating the spatial context information given the tentative boundary estimation of the current time point. Besides, the longitudinal context information from all time points of the temporal sequence of the same subject (i.e., given their tentative segmentation results) is also utilized to facilitate accurate segmentation. Finally, a deformable model is applied to the enhanced voted boundary maps for achieving the longitudinal hippocampal segmentation. The experiments on infant brain MR images acquired from 2-week-old to 1-year-old show promising hippocampal segmentation results, indicating the applicability of our method in early brain development study.

The hippocampus (Hc) is of great importance in various psychiatric diseases in adults, children and adolescents. Automated Hc segmentation has been widely used in adults, implying sufficient overlap with manual segmentation. However, estimation biases related to the Hc volume have been pointed out. This may particularly apply to children who show age-related Hc volume changes, thus, questioning the accuracy of automated Hc segmentation in this age group. The aim of this study was to compare manual segmentation with automated segmentation using the widely adopted FreeSurfer (FS) and MAGeT-Brain software. In 70 children and adolescents (5-16 years, mean age 10.6 years), T1-weighted images were acquired on one of two identical 3T scanners. Automated segmentation was performed using the FS subcortical segmentation, the FS hippocampal subfields segmentation and the MAGeT-Brain software. In comparison with manual segmentation, volume differences, Dice similarity coefficient (DSC), Bland-Altman plot, intraclass correlation coefficient (ICC) and left-right consistency of automated segmentation were calculated. The average percentage of volume differences (PVD) with manual segmentation was 56.8% for FS standard segmentation, 32.2% for FS subfield segmentation and - 15.6% for MAGeT-Brain. The FS Hc subfields segmentation (left/right DSC = 0.86/0.87) and MAGeT-Brain (both hemispheres DSC = 0.91) resulted in a higher volume overlap with manual segmentation compared with the FS subcortical segmentation (DSC = 0.79/0.78). In children aged 5-10.5 years, MAGeT-Brain yielded the highest overlap (DSC = 0.92/0.93). Contrary volume estimation biases were detected in FS and MAGeT-Brain: FS showed larger volume overestimation in smaller Hc volumes, while MAGeT-Brain showed more pronounced volume underestimation in larger Hc volumes. While automated Hc segmentation using FS hippocampal subfields or MAGeT-Brain resulted in adequate volume overlap with manual segmentation, estimation biases compromised the reliability of automated procedures in children and adolescents.

Shown in every neuroanatomy textbook, a key morphological feature is the bumpy ridges, which we refer to as hippocampal dentation, on the inferior aspect of the hippocampus. Like the folding of the cerebral cortex, hippocampal dentation allows for greater surface area in a confined space. However, examining numerous approaches to hippocampal segmentation and morphology analysis, virtually all published 3D renderings of the hippocampus show the inferior surface to be quite smooth or mildly irregular; we have rarely seen the characteristic bumpy structure on reconstructed 3D surfaces. The only exception is a 9.4T postmortem study (Yushkevich et al. [2009]: NeuroImage 44:385-398). An apparent question is, does this indicate that this specific morphological signature can only be captured using ultra high-resolution techniques? Or, is such information buried in the data we commonly acquire, awaiting a computation technique that can extract and render it clearly? In this study, we propose an automatic and robust super-resolution technique that captures the fine scale morphometric features of the hippocampus based on common 3T MR images. The method is validated on 9.4T ultra-high field images and then applied on 3T data sets. This method opens possibilities of future research on the hippocampus and other sub-cortical structural morphometry correlating the degree of dentation with a range of diseases including epilepsy, Alzheimer's disease, and schizophrenia. Hum Brain Mapp 39:472-490, 2018. © 2017 Wiley Periodicals, Inc.

Introduction A harmonized protocol (HarP) for manual hippocampal segmentation on magnetic resonance imaging (MRI) has recently been developed by an international European Alzheimer's Disease Consortium–Alzheimer's Disease Neuroimaging Initiative project. We aimed at providing consensual certified HarP hippocampal labels in Montreal Neurological Institute (MNI) standard space to serve as reference in automated image analyses. Methods Manual HarP tracings on the high-resolution MNI152 standard space template of four expert certified HarP tracers were combined to obtain consensual bilateral hippocampus labels. Utility and validity of these reference labels is demonstrated in a simple atlas-based morphometry approach for automated calculation of HarP-compliant hippocampal volumes within SPM software. Results Individual tracings showed very high agreement among the four expert tracers (pairwise Jaccard indices 0.82–0.87). Automatically calculated hippocampal volumes were highly correlated (rL/R = 0.89/0.91) with gold standard volumes in the HarP benchmark data set (N = 135 MRIs), with a mean volume difference of 9% (standard deviation 7%). Conclusion The consensual HarP hippocampus labels in the MNI152 template can serve as a reference standard for automated image analyses involving MNI standard space normalization.

We provide and evaluate an open-source software solution for automatically hippocampal segmentation from T1-weighted (T1w) magnetic resonance imaging (MRI). The method is applied for measuring the hippocampal volume, which allows discriminate between patients with Alzheimer's disease (AD) or mild cognitive impairment (MCI) and elderly controls (NC). The method is based on a fast patch-based label fusion method, whose selected patches and their weights are calculated from a combination of similarity measures between patches using intensity-based distances and labeling-based distances. These combined similarity measures produces better selection of the patches, and their weights are more robust. The algorithm is trained with the Harmonized Hippocampal Protocol (HarP). The proposal is compared with FreeSurfer and other label fusion methods. To evaluate the performance and the robustness of the proposed label fusion method, we employ two databases of T1w MRI of human brains. For AD vs NC, we obtain a high degree of accuracy, approximately 90 %. For MCI vs NC, we obtain accuracies around 75 %. The average time for the hippocampal segmentation from a T1w MRI is less than 17 minutes.

Recently, the level set has become a popular method in many research fields. The main reason is that it can be modified into many variants. One such case is our proposed method. We describe a contrast-enhancement method to segment the hippocampal region from the background. However, the hippocampus region has quite similar intensities to the neighboring pixel intensities. In addition, to handle the inhomogeneous intensities of the hippocampus, we used a bias correction before hippocampal segmentation. Thus, we developed a contrast-enhanced bias-corrected distance-regularized level set (CBDLS) to segment the hippocampus in magnetic resonance imaging (MRI). It shows better performance than the distance-regularized level set evolution (DLS) and bias-corrected distance-regularized level set (BDLS) methods in 33 MRI images of one normal patient. Segmentation after contrast enhancement and bias correction can be done more accurately than segmentation while not using a bias-correction method and without contrast enhancement.

The morphological analysis of hippocampus based on magnetic resonance imaging is significant for computer-aided neurological disease diagnosis. The segmentation of the hippocampus is a prerequisite for calculating its volume and measuring its shape. In this study, we propose a dual dense context-aware network (DDCNet) for automatic hippocampal segmentation, mainly including a multi-scale input approach and multi-resolution feature fusion. First, considering that a single-scale network cannot accurately segment the hippocampus and background with large difference in volume, we design a multi-scale input module (MSIM) to extract the hippocampus features of different receptive fields. Specifically, large scales are used to learn spatial dependencies, while smaller scales are used to learn more hippocampal details. Taking advantage of the fact that the features from different resolution layers offer semantically different information for segmentation, we develop a multi-resolution feature fusion module (MRFFM) to effectively aggregate complementary features and to simultaneously utilize the high-resolution features from the encoder to guide the low-resolution hippocampus edge segmentation in the decoder phase. In addition, many dense connections are introduced to further extend the feature fusion and contextual information of various resolutions between the encoder and decoder. The proposed approach achieves state-of-the-art performance, outperforming existing hippocampal segmentation approaches on the ADNI dataset.

Hippocampal atrophy measures from magnetic resonance imaging (MRI) are powerful tools for monitoring Alzheimer's disease (AD) progression. In this paper, we introduce a longitudinal image analysis framework based on robust registration and simultaneous hippocampal segmentation and longitudinal marker classification of brain MRI of an arbitrary number of time points. The framework comprises two innovative parts: a longitudinal segmentation and a longitudinal classification step. The results show that both steps of the longitudinal pipeline improved the reliability and the accuracy of the discrimination between clinical groups. We introduce a novel approach to the joint segmentation of the hippocampus across multiple time points; this approach is based on graph cuts of longitudinal MRI scans with constraints on hippocampal atrophy and supported by atlases. Furthermore, we use linear mixed effect (LME) modeling for differential diagnosis between clinical groups. The classifiers are trained from the average residue between the longitudinal marker of the subjects and the LME model. In our experiments, we analyzed MRI-derived longitudinal hippocampal markers from two publicly available datasets (Alzheimer's Disease Neuroimaging Initiative, ADNI and Minimal Interval Resonance Imaging in Alzheimer's Disease, MIRIAD). In test/retest reliability experiments, the proposed method yielded lower volume errors and significantly higher dice overlaps than the cross-sectional approach (volume errors: 1.55% vs 0.8%; dice overlaps: 0.945 vs 0.975). To diagnose AD, the discrimination ability of our proposal gave an area under the receiver operating characteristic (ROC) curve (AUC) [Formula: see text] 0.947 for the control vs AD, AUC [Formula: see text] 0.720 for mild cognitive impairment (MCI) vs AD, and AUC [Formula: see text] 0.805 for the control vs MCI.

Recently, a group of major international experts have completed a comprehensive effort to efficiently define a harmonized protocol for manual hippocampal segmentation that is optimized for Alzheimer's research (known as the EADC-ADNI Harmonized Protocol (the HarP)). This study compares the HarP with one of the widely used hippocampal segmentation protocols (Pruessner, 2000), based on a single automatic segmentation method trained separately with libraries made from each manual segmentation protocol. The automatic segmentation conformity with the corresponding manual segmentation and the ability to capture Alzheimer's disease related hippocampal atrophy on large datasets are measured to compare the manual protocols. In addition to the possibility of harmonizing different procedures of hippocampal segmentation, our results show that using the HarP, the automatic segmentation conformity with manual segmentation is also preserved (Dice's κ = 0.88 , κ = 0.87 for Pruessner and HarP respectively (p = 0.726 for common training library)). Furthermore, the results show that the HarP can capture the Alzheimer's disease related hippocampal volume differences in large datasets. The HarP-derived segmentation shows large effect size (Cohen's d = 1.5883) in separating Alzheimer's Disease patients versus normal controls (AD:NC) and medium effect size (Cohen's d = 0.5747) in separating stable versus progressive Mild Cognitively Impaired patients (sMCI:pMCI). Furthermore, the area under the ROC curve for a LDA classifier trained based on age, sex and HarP-derived hippocampal volume is 0.8858 for AD:NC, and for 0.6677 sMCI:pMCI. These results show that the harmonized protocol-derived labels can be widely used in clinic and research, as a sensitive and accurate way of delineating the hippocampus.

Introduction: Chronic cannabis use is associated with neuroanatomical alterations in the hippocampus. While adverse impacts of cannabis use are generally attributed to Δ9-tetrahydrocannabinol, emerging naturalistic evidence suggests cannabidiol (CBD) is neuroprotective and may ameliorate brain harms associated with cannabis use, including protection from hippocampal volume loss. This study examined whether prolonged administration of CBD to regular cannabis users within the community could reverse or reduce the characteristic hippocampal harms associated with chronic cannabis use. Materials and Methods: Eighteen regular cannabis users participated in an ∼10-week open-label pragmatic trial involving daily oral administration of 200 mg CBD, with no change to their ongoing cannabis use requested. Participants were assessed at baseline and post-CBD treatment using structural magnetic resonance imaging. Automated longitudinal hippocampal segmentation was performed to assess volumetric change over the whole hippocampus and within 12 subfields. Results: No change was observed in left or right hippocampus as a whole. However, left subicular complex (parasubiculum, presubiculum, and subiculum) volume significantly increased from baseline to post-treatment (p=0.017 uncorrected) by 1.58% (Cohen's d=0.63; 2.83% in parasubiculum). Heavy cannabis users demonstrated marked growth in the left subicular complex, predominantly within the presubiculum, and right cornu ammonis (CA)1 compared to lighter users. Associations between greater right subicular complex and total hippocampal volume and higher plasma CBD concentration were evident, particularly in heavy users. Conclusions: Our findings suggest a restorative effect of CBD on the subicular and CA1 subfields in current cannabis users, especially those with greater lifetime exposure to cannabis. While replication is required in a larger, placebo-controlled trial, these findings support a protective role of CBD against brain structural harms conferred by chronic cannabis use. Furthermore, these outcomes suggest that CBD may be a useful adjunct in treatments for cannabis dependence and may be therapeutic for a range of clinical disorders characterized by hippocampal pathology (e.g., schizophrenia, Alzheimer's disease, and major depressive disorder).

Background To evaluate a new semi-automated segmentation method for calculating hippocampal volumes and to compare results with standard software tools in a cohort of people with subjective memory complaints (SMC) and mild cognitive impairment (MCI). Methods Data from 58 participants, 39 with SMC (17 male, 22 female, mean age 72.6) and 19 with MCI (6 male, 13 female, mean age 74.3), were analyzed. For each participant, T1-weighted images were acquired using an MPRAGE sequence on a 3 Tesla MRI system. Hippocampal volumes (left, right, and total) were calculated with a new, age appropriate registration template, based on older people and using the advanced software tool ANTs (Advanced Normalization Tools). The results were compared with manual tracing (seen as the reference standard) and two widely accepted automated software tools (FSL, FreeSurfer). Results The hippocampal volumes, calculated by using the age appropriate registration template were significantly (P

Background Hippocampal atrophy is a supportive feature for the diagnosis of probable Alzheimer’s disease (AD). However, even for an expert neuroradiologist, tracing the hippocampus and measuring its volume is a time consuming and extremely challenging task. Accordingly, the development of reliable fully-automated segmentation algorithms is of paramount importance. Materials and methods The present study evaluates (i) the precision and the robustness of the novel Hippocampal Unified Multi-Atlas Network (HUMAN) segmentation algorithm and (ii) its clinical reliability for AD diagnosis. For these purposes, we used a mixed cohort of 456 subjects and their T1 weighted magnetic resonance imaging (MRI) brain scans. The cohort included 145 controls (CTRL), 217 mild cognitive impairment (MCI) subjects and 94 AD patients from Alzheimer’s Disease Neuroimaging Initiative (ADNI). For each subject the baseline, repeat, 12 and 24 month follow-up scans were available. Results HUMAN provides hippocampal volumes with a 3% precision; volume measurements effectively reveal AD, with an area under the curve (AUC) AUC1 = 0.08 ± 0.02. Segmented volumes can also reveal the subtler effects present in MCI subjects, AUC2 = 0.76 ± 0.05. The algorithm is stable and reproducible over time, even for 24 month follow-up scans. Conclusions The experimental results demonstrate HUMAN is a precise segmentation algorithm, besides hippocampal volumes, provided by HUMAN, can effectively support the diagnosis of Alzheimer’s disease and become a useful tool for other neuroimaging applications.

Summary Objective To compare manual and automated preoperative and postoperative hippocampal volume measurements in patients with intractable epilepsy. Methods We studied 34 patients referred to the Clinical Epilepsy Section, National Institute of Neurological Disorders and Stroke (NINDS), National Institutes of Health (NIH) for evaluation of intractable epilepsy and 21 normal volunteers who received 1.5 or 3 T GE Signa magnetic resonance imaging (MRI) scans. Hippocampal volumes were traced manually on each slice and assembled into three-dimensional volumes by investigators who were blinded to other data. Automated volumetric measurements were obtained using FreeSurfer. Statistical analysis was performed with GraphPad Prism. Results Automated hippocampal volumes were larger than manual volumes in both patients and normal volunteers (p

The hippocampus is located within the medial temporal lobe and plays a key role in learning and episodic, semantic, and spatial memory. Dysfunction has been reported in neurologic and psychiatric disorders including epilepsy (Wu et al., 2005), Alzheimer's disease (Apostolova et al., 2006), schizophrenia (Tanskanen et al., 2005), and depression (Bremner et al., 2000). Temporal lobe epilepsy (TLE) is the most common drug-resistant focal epilepsy, with seizures frequently arising from the hippocampus. In surgical series of TLE, the pathology is often hippocampal sclerosis (HS) comprising neuronal loss and gliosis and marked by atrophy and signal change on magnetic resonance imaging (Van Paesschen, 2004). Atrophy of the hippocampus through HS provides a good biomarker for the laterality of the seizure focus (Bernasconi et al., 2003), and combined with concordant neurophysiology and neuropsychological data can be sufficient to recommend surgery. Hippocampal atrophy is associated with a favorable surgical outcome (Schramm & Clusmann, 2008). Visual assessment of hippocampal volumes is unreliable, as it may be compromised by head position and primarily detects hippocampal asymmetry rather than volume loss, making bilateral atrophy difficult to identify. Hippocampal segmentation and volumetry are thus important for diagnosis and surgical planning (Watson et al., 1997). The gold standard for hippocampal segmentation is manual delineation by trained raters. This is accurate, reproducible, and sensitive but is time-consuming, requires anatomic knowledge, and is subject to interrater and intrarater variability. The hippocampus is challenging to delineate as it is small and highly variable with ill-defined margins. Many protocols exist for manual segmentation depending on which structures are included and the boundary definition (Konrad et al., 2009). Automated segmentation techniques aim to ensure operator independence, high reproducibility, and reduced demand for human time and expertise. The strongest drive for automation has come from researchers working with large cohorts of patients with Alzheimer's disease patients. Hippocampal volumes are an early marker for the disease, are related to cognitive status, and may reflect disease progression in clinical trials (Frisoni & Jack, 2011). In atlas-based segmentation approaches, a template and associated manual labels are registered (matched) to the new image (Carmichael et al., 2005). Commonly used methods, including FreeSurfer (Fischl et al., 2002), rely on a single template so that subjects that differ significantly from the template, for example HS, are poorly segmented. Segmentation of hippocampi that are sclerotic is more challenging than segmenting hippocampi in Alzheimer's disease, as the latter is associated with more prominent cerebrospinal fluid (CSF)–hippocampal boundaries, whereas the former is associated with signal change. The use of an atlas with multiple template images is more effective than a single template (Heckemann et al., 2006) and depends on the quality of registration and template selection strategy. Most previous atlas-based segmentation studies used small template databases of healthy subjects. Results obtained in TLE are significantly worse than in healthy subjects or Alzheimer's disease (Kim et al., 2012), as aside from atrophy, approximately 40% of patients with TLE demonstrate an atypical shape or position of the hippocampus (Bernasconi et al., 2005). In this study, we adapted our published method developed for use in Alzheimer's disease (Cardoso et al., 2013) to a large cohort of adult patients with epilepsy by employing accurate nonlinear registration (Modat et al., 2010) and a large template database that encompasses the range of pathology observed in epilepsy at a tertiary referral center. Manual segmentations of the most similar images from the template database are combined using a label fusion strategy based on local similarity to ensure accurate segmentation regardless of pathology. We demonstrate that this technique achieves reliable segmentation with no more variability than that seen between different expert raters. The algorithm is made freely available via an online Web-based service (https://hipposeg.cs.ucl.ac.uk). In addition, the software, scripts, and an anonymized version of the template database are available from this website.

Hippocampal segmentation accuracy of out-of-the-box software tools (FreeSurfer, AHEAD, BrainParser) is analysed wrt. potential variability in populations with different pathologies. Findings confirm variabilities wrt. different pathologies but also human rater ground truth and single pathologies exhibit significant variability as well.

In automated hippocampus segmentation, issues related to ground truth rater variability, subject variability and variability of software segmentation accuracy are investigated in the context of 3 publicly available, out-of-the-box software packages. Ground truth variability among three manual raters is controlled using a majority voting based label fusion scheme and observed subject variability underpins the importance of availability of large scale ground truth.

We describe a fully automated method for hippocampal segmentation. The method uses SPM5 (http://www.fil.ion.ucl.ac.uk/spm/) software to segment the brain into grey/white matter, and spatially normalize the images to standard space. Grey matter pixels within a predefined hippocampal region in standard space are identified to segment the hippocampi. The method was validated on 36 subjects (9 each of Alzheimer's disease, dementia with Lewy bodies, vascular dementia, and healthy controls). The mean absolute difference in volume compared with manual segmentation was 11% (SD 9%). Linear regression between manual and automated volume gave V(auto) = V(manual) × 0.83 + 401 ml. The method provides an acceptable automated alternative to manual segmentation which may be of value in large studies. Hum Brain Mapp 2008. © 2007 Wiley‐Liss, Inc.

ObjectiveThe pathologic validation of European Alzheimer's Disease Consortium Alzheimer's Disease Neuroimaging Initiative Center Harmonized Hippocampal Segmentation Protocol (HarP).MethodsTemporal lobes of nine Alzheimer's disease (AD) and seven cognitively normal subjects were scanned post-mortem at 7 Tesla. Hippocampal volumes were obtained with HarP. Six-micrometer-thick hippocampal slices were stained for amyloid beta (Aβ), tau, and cresyl violet. Hippocampal subfields were manually traced. Neuronal counts, Aβ, and tau burden for each hippocampal subfield were obtained.ResultsWe found significant correlations between hippocampal volume and Braak and Braak staging (ρ = -0.75, P = .001), tau (ρ = -0.53, P = .034), Aβ burden (ρ = -0.61, P = .012), and neuronal count (ρ = 0.77, P < .001). Exploratory subfield-wise significant associations were found for Aβ in Cornu Ammonis (CA)1 (ρ = -0.58, P = .019) and subiculum (ρ = -0.75, P = .001), tau in CA2 (ρ = -0.59, P = .016), and CA3 (ρ = -0.5, P = .047), and neuronal count in CA1 (ρ = 0.55, P = .028), CA3 (ρ = 0.65, P = .006), and CA4 (ρ = 0.76, P = .001).ConclusionsThe observed associations provide pathological confirmation of hippocampal morphometry as a valid biomarker for AD and pathologic validation of HarP.

Background: Hippocampal volumetry derived from structural magnetic resonance imaging (MRI) has been endorsed by the Alzheimer’s disease (AD) diagnostic guidelines as a radiological marker of disease progression. Among the top performing automated hippocampal segmentation methods are multi-atlas segmentation methods, which rely on manual annotations. In this study, we investigate a combination of such method with annotations from a new Harmonized Hippocampal Protocol (HHP). We compare its capabilities to a FreeSurfer method and verify its impact on segmentation and diagnostic group separation capabilities. Methods: 40 manual HHP hippocampal annotations (12 normal control (NC), 11 mild cognitive impairment (MCI), 17 AD) were transformed to a common segmentation space. The corresponding 1.5T MRIs were preprocessed using FreeSurfer. An automated Non-Local Patch-based segmentation technique (N-L Patch) was used to segment the leftand right hippocampus, separately. All 40 HHP annotations were used as atlases during pre-selection, but only the 9 most similar contributed to the final segmentation. Leaveone-out cross-validation was performed on the 40 atlases, and the corresponding DICE-scores with the manual annotations were calculated. A standardized ADNI dataset containing 1.5T MRIs from 504 subjects (169 NC, 234 MCI, 101 AD) at baseline and month 12 was segmented using the method described above and atrophy rate calculated as percentage volume change was estimated. Results: Mean (6sd) cross-validation DICE-scores of the 40 atlases segmented using N-L Patch and crosssectional FreeSurfer were 0.868 (60.019) and 0.781 (60.031), respectively. A paired t-test between N-L Patch and FreeSurfer DICE-scores showed significance (p

IC-P-058 AUTOMATED HIPPOCAMPAL SEGMENTATION USING NEW STANDARDIZED MANUAL SEGMENTATIONS FROM THE HARMONIZED HIPPOCAMPAL PROTOCOL Cecilie Benedicte Anker, Akshay Pai, Lauge Sorensen, Mark Lyksborg, Rasmus Larsen, Knut Conradsen, Mads Nielsen, Biomediq A/S, Copenhagen, Denmark; University of Copenhagen, Copenhagen, Denmark; Biomediq A/S, Copenhagen, Denmark; Technical University of Denmark, Kongens Lyngby, Denmark; Biomediq, Copenhagen, Denmark. Contact e-mail: ceciliea@biomediq.com

The application of voxel-level three-dimensional registration to serial magnetic resonance imaging (MRI) is described. This fluid registration determines deformation fields modeling brain change, which are consistent with a model describing a viscous fluid. The objective was to validate the measurement of hippocampal volumetric change by fluid registration in Alzheimer's disease (AD) against current methodologies. The hippocampus was chosen for this study because it is difficult to measure reproducibly by manual segmentation and is widely studied; however, the technique is applicable to any structure which can be delineated on a scan. First, suitable values for the viscosity-body-force-ratio, alpha (0.01), and the number of iterations (300), were established and the convergence, repeatability, linearity, and accuracy investigated and compared with expert manual segmentation. A simple model of hippocampal atrophy was used to compare simulated volumetric change against that obtained by fluid registration. Finally the serial segmentation was compared with the current gold standard technique-expert human labeling with a volume repeatability of approximately 4%-in 27 subjects (15 normal controls, 12 clinically diagnosed with Alzheimer's disease). The scan-rescan volumetric consistency of serial segmentation by fluid-registration was shown to be superior to human serial segmentors ( approximately 2%). The mean absolute volume difference between fluid and manual segmentation was 0.7%. Fluid registration has potential importance for tracking longitudinal structural changes in brain particularly in the context of the clinical trial where large numbers of subjects may have multiple MR scans.

BackgroundThis study aimed to have international experts converge on a harmonized definition of whole hippocampus boundaries and segmentation procedures, to define standard operating procedures for magnetic resonance (MR)-based manual hippocampal segmentation.MethodsThe panel received a questionnaire regarding whole hippocampus boundaries and segmentation procedures. Quantitative information was supplied to allow evidence-based answers. A recursive and anonymous Delphi procedure was used to achieve convergence. Significance of agreement among panelists was assessed by exact probability on Fisher's and binomial tests.ResultsAgreement was significant on the inclusion of alveus/fimbria (P = .021), whole hippocampal tail (P = .013), medial border of the body according to visible morphology (P = .0006), and on this combined set of features (P = .001). This definition captures 100% of hippocampal tissue, 100% of Alzheimer’s disease-related atrophy, and demonstrated good reliability on preliminary intrarater (0.98) and inter-rater (0.94) estimates.DiscussionConsensus was achieved among international experts with respect to hippocampal segmentation using MR resulting in a harmonized segmentation protocol. published

We present the rationale, design, description and early user cases for a web-enabled platform for certification of manual segmentation of the hippocampus from MRIs in the context of Alzheimer's disease. Preliminary thresholds serving as qualification criteria for volumetric correlation, spatial overlap and spatial distance are introduced.

The development of tools for a fully automatic segmentation of the relevant brain structures, such as the hippocampus, is potentially very useful for pathologies detection. In this paper, a method for the automated hippocampal segmentation, based on virtual ant colonies, is proposed. The algorithm used, the Channeler Ant Model (CAM), represents an effective way to segment 3D objects with a complex shape in a noisy background. The CAM was modified by inserting a shape knowledge that is crucial to face the hippocampus segmentation. The algorithm was trained and tested using a database of 56 T1 weighted MRI images with a known manual segmentation of the hippocampus volume. The results are comparable to other methods: an average Dice Index of 0.74 and 0.72 is obtained over the left and right hippocampi, respectively. The lack of a heavy training procedure, because all the model parameters are fixed, and the speed make this approach very effective.

In drug-resistant temporal lobe epilepsy (TLE), detecting hippocampal atrophy on MRI is crucial as it allows defining the surgical target. In addition to atrophy, about 40% of patients present with malrotation, a developmental anomaly characterized by atypical morphologies of the hippocampus and collateral sulcus. We have recently shown that both atrophy and malrotation impact negatively the performance of volume-based techniques. Here, we propose a novel hippocampal segmentation algorithm (SurfMulti) that integrates deformable parametric surfaces, vertex-wise modeling of locoregional texture and shape, and multiple templates in a unified framework. To account for inter-subject variability, including shape variants, we used a library derived from a large database of healthy (n = 80) and diseased (n = 288) hippocampi. To quantify malrotation, we generated 3D models from manual hippocampal labels and automatically extracted collateral sulci. The accuracy of SurfMulti was evaluated relative to manual labeling and segmentation obtained through a single atlas-based algorithm (FreeSurfer) and a volume-based multi-template approach (Vol-multi) using the Dice similarity index and surface-based shape mapping, for which we computed vertex-wise displacement vectors between automated and manual segmentations. We then correlated segmentation accuracy with malrotation features and atrophy. SurfMulti outperformed FreeSurfer and Vol-multi, and achieved a level of accuracy in TLE patients (Dice = 86.9%) virtually identical to healthy controls (Dice = 87.5%). Vertex-wise shape mapping showed that SurfMulti had an excellent overlap with manual labels, with sub-millimeter precision. Its performance was not influenced by atrophy or malrotation (|r| 0.2), while FreeSurfer (|r| > 0.35, p 0.28, p < 0.05) were hampered by both anomalies. The magnitude of atrophy detected using SurfMulti was the closest to manual volumetry (Cohen’s d: manual = 1.71, t = 7.6; SurfMulti = 1.60, t = 7.0; Vol-multi = 1.38, t = 6.1; FreeSurfer = 0.91, t = 3.9). The high performance of SurfMulti regardless of cohort, atrophy and shape variants identifies this algorithm as a robust segmentation tool for hippocampal volumetry.

Hippocampal atrophy is a marker of disease state and progression in Alzheimer's disease. The gold standard to measure hippocampal volume is through manual segmentation. A number of protocols to measure hippocampal volume through manual segmentation have been developed, but the marked heterogeneity of anatomical landmarks has given rise to wide variability of volume estimates. With the aim of fostering the use of hippocampal volume in routine clinical settings, an international task force is currently working on developing a harmonized protocol that will resolve and reduce the present heterogeneity. The task force will then validate the harmonized protocol, develop harmonized probabilistic hippocampal maps, and develop illustrative and educational material on the use of the harmonized protocol and maps.

Hippocampal volume is reduced in Alzheimer Disease (AD) and has been proposed as a possible surrogate biomarker to aid early diagnosis. Whilst automated methods to segment the hippocampus from magnetic resonance images are available, manual segmentation, in spite of being time-consuming and unsuitable for large samples, is still the standard. In order to study the validity of FreeSurfer's automated method, we compared hippocampal automated measures with manual tracing in a sample composed of healthy elderly (N=41), Mild Cognitive Impairment (MCI) (N=23), and AD (N=25) subjects. Percent volume overlap, percent volume difference, correlations, and Bland-Altman plots were studied. Automated measures were slightly larger than hand tracing ones (mean difference 10%). Percent volume overlap showed good results, but was far from perfect (78%). Manual and automated volume correlations were approximately 0.84 and the Bland-Altman analysis showed acceptable interchangeability of methods. Within-group analysis demonstrated that patient samples obtained smaller values in validity indexes than controls. Globally, FreeSurfer's automated hippocampal volumetry showed adequate validity when compared to manual tracing, with a tendency to overestimation. Nevertheless, the greater difference between automated and manual segmentation in atrophic brains suggests that studies in AD based on this software could be more likely to produce false negatives. (C) 2009 Elsevier Ireland Ltd. All rights reserved.

Voxel-level three-dimensional fluid registration was used to propagate hippocampal segmentation from baseline MR scans onto follow-up scans. In this way automated measurements of hippocampal volume and volume changes were derived. The objective was to compare this technique with the current gold standard of manual segmentation. Serial measurements were performed on 15 normal controls and 12 Alzheimer's disease (AD) subjects. Hippocampal regions on the rigidly-registered repeat scans were generated both by manual segmentation and fluid registration. There was no significant difference between the repeat volumes generated by the two methods (p

Objective: To define a harmonized protocol for manual hippocampal segmentation from magnetic resonance scans. Background Heterogeneity of landmarks among protocols leads to different volume estimates, hampering comparison of studies and clinical use. Landmark differences among the 12 most common protocols were extracted, operationalized, and quantitatively investigated. The results were presented to the Delphi panel, consisting of seventeen researchers with substantial expertise in hippocampal segmentation, to reach an evidence-based consensus on segmentation landmarks. Design/Methods: The Delphi panel participated in recursive anonymous voting sessions where feedback from previous rounds was utilized to progressively facilitate panelists9 agreement. Panelists were presented with tracing alternatives, each associated with quantitative data relating: (i) reliability, (ii) impact on whole hippocampal volume, and (iii) correlation with AD pathology. Panelists were asked to choose among alternatives and provide justification and comments. Anonymous votes and comments, and voting statistics of the first round were fed into the second Delphi round. Exact probability on binomial test of the choice was computed. At the time of writing, two Delphi rounds were completed. Results: Sixteen panelists completed both Delphi rounds. Agreement was significant on inclusion of alveus and fimbria (81%, p=0.021), of the whole visible hippocampal tail (75%, p=0.035), and approached significance for the segmentation of the medial border of the body following visible morphology (69%, p=0.057). Agreement on including the minimum hippocampus, alveus, morphological subiculum, whole hippocampal tail in the final hippocampal tracing protocol was significant (88%, p=0.001). Based on the previous quantitative investigation of landmark differences, the hippocampus so defined covers 100% of hippocampal tissue, captures 100% of AD-related atrophy, and has good intra-rater (0.99) and inter-rater (0.94) reliability measured on volume statistics. Conclusions: The most “inclusive” landmarks were preferred by Delphi panelists. Further rounds will be run to seek statistically significant agreement for all aspects. Supported by: Alzheimer9s Association. Lilly International. Wyeth International (Pfizer Group). Disclosure: Dr. Boccardi has nothing to disclose. Dr. Bocchetta has nothing to disclose. Dr. Apostolova has nothing to disclose. Dr. Barnes has nothing to disclose. Dr. Bartzokis has received personal compensation for activities with Bristol-Myers Squibb Company and Novartis. Dr. Bartzokis has received research support from Novartis and Janssen Pharmaceutica. Dr. Corbetta has nothing to disclose. Dr. DeCarli has received personal compensation for activities wtih Merck Pharmaceutical, Avanir, Bayer Pharmaceuticals, and Pfizer as a consultant.Dr. DeCarli has received personal compensation in an editorial capacity for Alzheimer9s Disease and Associated Disorders.Dr. DeCarli has received research support from Merck. Dr. DeToledo-Morrell has nothing to disclose. Dr. Firbank has nothing to disclose. Dr. Ganzola has nothing to disclose. Dr. Gerritsen has nothing to disclose. Dr. Henneman has nothing to disclose. Dr. Killiany has received personal compensation for activities with Sunovion Pharmaceuticals Inc as a consultant. Dr. Killiany has received research support from The MRI Centers of New England and a clinical affiliation with Essex Neurological Associates. Dr. Malykhin has nothing to disclose. Dr. Pasqualetti has nothing to disclose. Dr. Pruessner has nothing to disclose. Dr. Redolfi has nothing to disclose. Dr. Robitaille has nothing to disclose. Dr. Soininen has received personal compensation for activities with ACImmune. Dr. Tolomeo has nothing to disclose. Dr. Wang has nothing to disclose. Dr. Watson has nothing to disclose. Dr. Wolf has nothing to disclose. Dr. Duchesne has received (royalty or license fee or contractual rights) payments from Screen Inc.Dr. Duchesne has received research support from Afga Healthcare Inc. Dr. Jack has received personal compensation for activities with Janssen Pharmaceutica, Eisai Inc, General Electric, Johnson and Johnson, and Eli Lilly & Company as a consultant.Dr. Jack has received research support from Pfizer Inc, Allon, and Baxter, Inc. Dr. Frisoni has received personal compensation for activities Eli Lilly & Company, Bristol-Myers Squibb Company, Bayer Healthcare, Lundbeck Research USA, Inc., Elan Corporation, AstraZeneca Pharmaceuticals, Pfizer Inc, Baxter, Taurx, and Wyeth Pharmaceuticals. Dr. Frisoni has received personal compensation in an editorial capacity for Lancet Neurology, Aging Clinical & Experimental Research, Neurobiology of Aging, Alzheimer9s Diseases and Associated Disorders, and Neurogenerative Diseases. Dr. Frisoni has received research support from Wyeth Corporation, Eli Lilly & Company, Lundbeck Research USA, Inc.

Alzheimer's Disease, MCI, and Control Subjects J H Morra1, Z Tu1, L G Apostolova1,2, A E Green1,2, C Avedissian1, S K Madsen1, N Parikshak1, A W Toga1, C R Jack3, N Schuff4, M W Weiner4,5, P M Thompson1 Laboratory of Neuro Imaging, UCLA, Los Angeles, CA, United States/Dept. of Neurology, UCLA, Los Angeles, CA, United States/Mayo Clinic College of Medicine, Rochester, MN, United States/Dept. of Radiology, UCSF, San Francisco, CA, United States/Dept. of Medicine and Psychiatry, UCSF, San Francisco, CA, United States