Automatic detection and quantification of hippocampal atrophy on MRI in temporal lobe epilepsy: a proof-of-principle study

Current themes in neuroimaging of epilepsy: brain networks, dynamic phenomena, and clinical relevance

Brain scanning methods were first applied in patients with epilepsy more than 30years ago. A very substantial literature now exists in this field, which is exponentially increasing. Contemporary neuroimaging studies in epilepsy reflect new concepts in the epilepsies, as well as current methodological developments. In particular, this area is emphasising the role of networks in epileptogenicity, the existence of dynamic phenomena which can be captured by imaging, and is beginning to validate the implementation of neuroimaging in the clinic. Here, recent studies of the last 5years are reviewed, covering the full range of neuroimaging methods with SPECT, PET and MRI in epilepsy.

On the estimation of the location of the hippocampus in the context of hippocampal avoidance whole brain radiotherapy treatment planning

We demonstrate a technique for estimating the location of the hippocampus in MRI and CT images for use in radiotherapy treatment planning, using both rigid and contour based deformable image registration. The automatically generated contours can be subsequently modified for a given patient. By mapping the hippocampi from several patients into a template image set, a population-based average hippocampal atlas was generated. Approximate hippocampal contours can be automatically generated in a given image set by mapping this atlas onto it. The performance and accuracy of several atlases generated in different ways was tested on 10 MRI images and 7 CT images. Auto-contouring based on deformable registration significantly outperformed that based on rigid registration alone, with an average Dice similarity score of 0.62 (range .40-.76) for methods utilizing deformation. Comparable results were achieved in auto-contouring CT images when deformable registration was used, demonstrating that the methodology is robust with respect to imaging modality.

Atrophy patterns in Alzheimer's disease and semantic dementia: a comparison of FreeSurfer and manual volumetric measurements

Alzheimer's disease (AD) and semantic dementia (SD) are characterized by different patterns of global and temporal lobe atrophy which can be studied using magnetic resonance imaging (MRI). Manual delineation of regions of interest is time-consuming. FreeSurfer is a freely available automated technique which has a facility to label cortical and subcortical brain regions automatically. As with all automated techniques comparison with existing methods is important. Eight temporal lobe structures in each hemisphere were delineated using FreeSurfer and compared with manual segmentations in 10 control, 10 AD, and 10 SD subjects. The reproducibility errors for the manual segmentations ranged from 3% to 6%. Differences in protocols between the two methods led to differences in absolute volumes with the greatest differences between methods found bilaterally in the hippocampus, entorhinal cortex and fusiform gyrus (p<0.005). However, good correlations between the methods were found for most regions, with the highest correlations shown for the ventricles, whole brain and left medial-inferior temporal gyrus (r>0.9), followed by the bilateral amygdala and hippocampus, left superior temporal gyrus, right medial-inferior temporal gyrus and left temporal lobe (r>0.8). Overlap ratios differed between methods bilaterally in the amygdala, superior temporal gyrus, temporal lobe, left fusiform gyrus and right parahippocampal gyrus (p<0.01). Despite differences in protocol and volumes, both methods showed similar atrophy patterns in the patient groups compared with controls, and similar right-left differences, suggesting that both methods accurately distinguish between the three groups.

LEAP: learning embeddings for atlas propagation

We propose a novel framework for the automatic propagation of a set of manually labeled brain atlases to a diverse set of images of a population of subjects. A manifold is learned from a coordinate system embedding that allows the identification of neighborhoods which contain images that are similar based on a chosen criterion. Within the new coordinate system, the initial set of atlases is propagated to all images through a succession of multi-atlas segmentation steps. This breaks the problem of registering images that are very "dissimilar" down into a problem of registering a series of images that are "similar". At the same time, it allows the potentially large deformation between the images to be modeled as a sequence of several smaller deformations. We applied the proposed method to an exemplar region centered around the hippocampus from a set of 30 atlases based on images from young healthy subjects and a dataset of 796 images from elderly dementia patients and age-matched controls enrolled in the Alzheimer's Disease Neuroimaging Initiative (ADNI). We demonstrate an increasing gain in accuracy of the new method, compared to standard multi-atlas segmentation, with increasing distance between the target image and the initial set of atlases in the coordinate embedding, i.e., with a greater difference between atlas and image. For the segmentation of the hippocampus on 182 images for which a manual segmentation is available, we achieved an average overlap (Dice coefficient) of 0.85 with the manual reference.

Comparison of AdaBoost and support vector machines for detecting Alzheimer's disease through automated hippocampal segmentation

We compared four automated methods for hippocampal segmentation using different machine learning algorithms: 1) hierarchical AdaBoost, 2) support vector machines (SVM) with manual feature selection, 3) hierarchical SVM with automated feature selection (Ada-SVM), and 4) a publicly available brain segmentation package (FreeSurfer). We trained our approaches using T1-weighted brain MRIs from 30 subjects [10 normal elderly, 10 mild cognitive impairment (MCI), and 10 Alzheimer's disease (AD)], and tested on an independent set of 40 subjects (20 normal, 20 AD). Manually segmented gold standard hippocampal tracings were available for all subjects (training and testing). We assessed each approach's accuracy relative to manual segmentations, and its power to map AD effects. We then converted the segmentations into parametric surfaces to map disease effects on anatomy. After surface reconstruction, we computed significance maps, and overall corrected p-values, for the 3-D profile of shape differences between AD and normal subjects. Our AdaBoost and Ada-SVM segmentations compared favorably with the manual segmentations and detected disease effects as well as FreeSurfer on the data tested. Cumulative p-value plots, in conjunction with the false discovery rate method, were used to examine the power of each method to detect correlations with diagnosis and cognitive scores. We also evaluated how segmentation accuracy depended on the size of the training set, providing practical information for future users of this technique.

The optimal template effect in hippocampus studies of diseased populations

We evaluate the impact of template choice on template-based segmentation of the hippocampus in epilepsy. Four dataset-specific strategies are quantitatively contrasted: the "closest to average" individual template, the average shape version of the closest to average template, a best appearance template and the best appearance and shape template proposed here and implemented in the open source toolkit Advanced Normalization Tools (ANTS). The cross-correlation similarity metric drives the correspondence model and is used consistently to determine the optimal appearance. Minimum shape distance in the diffeomorphic space determines optimal shape. Our evaluation results show that, with respect to gold-standard manual labeling of hippocampi in epilepsy, optimal shape and appearance template construction outperforms the other strategies for gaining data-derived templates. Our results also show the improvement is most significant on the diseased side and insignificant on the healthy side. Thus, the importance of the template increases when used to study pathology and may be less critical for normal control studies. Furthermore, explicit geometric optimization of the shape component of the unbiased template positively impacts the study of diseased hippocampi.

FLAIR signal and texture analysis for lateralizing mesial temporal lobe epilepsy

Standard magnetic resonance (MR) imaging analysis in several cases of mesial temporal lobe epilepsy (mTLE) either fail to show an identifiable hippocampal asymmetry or provide only subtle distinguishing features that remain inconclusive. A retrospective analysis of hippocampal fluid-attenuated inversion recovery (FLAIR) MR images was performed in cases of mTLE addressing, particularly, the mean and standard deviation of the signal and its texture. Preoperative T1-weighted and FLAIR MR images of 25 nonepileptic control subjects and 36 mTLE patients with Engel class Ia outcomes were analyzed. Patients requiring extraoperative electrocorticography (ECoG) with intracranial electrodes and thus judged to be more challenging were studied as a separate cohort. Hippocampi were manually segmented on T1-weighted images and their outlines were transposed onto FLAIR studies using an affine registration. Image intensity features including mean and standard deviation and wavelet-based texture features were determined for the hippocampal body. The right/left ratios of these features were used with a linear classifier to establish laterality. Whole hippocampal within-subject volume ratios were assessed for comparison. Mean and standard deviation of FLAIR signal intensities lateralized the site of epileptogenicity in 98% of all cases, whereas analysis of wavelet texture features and hippocampal volumetry each yielded correct lateralization in 94% and 83% of cases, respectively. Of patients requiring more intensive study with extraoperative ECoG, 17/18 were lateralized effectively by the combination of mean and standard deviation ratios despite a ratio of mean signal intensity near one in some. The analysis of mean and standard deviation of FLAIR signal intensities provides a highly sensitive method for lateralizing the epileptic focus in mTLE over that of volumetry or texture analysis of the hippocampal body.

Hippocampal volume assessment in temporal lobe epilepsy: How good is automated segmentation?

PURPOSE

Quantitative measurement of hippocampal volume using structural magnetic resonance imaging (MRI) is a valuable tool for detection and lateralization of mesial temporal lobe epilepsy with hippocampal sclerosis (mTLE). We compare two automated hippocampal volume methodologies and manual hippocampal volumetry to determine which technique is most sensitive for the detection of hippocampal atrophy in mTLE.

METHODS

We acquired a three-dimensional (3D) volumetric sequence in 10 patients with left-lateralized mTLE and 10 age-matched controls. Hippocampal volumes were measured manually, and using the software packages Freesurfer and FSL-FIRST. The sensitivities of the techniques were compared by determining the effect size for average volume reduction in patients with mTLE compared to controls. The volumes and spatial overlap of the automated and manual segmentations were also compared.

RESULTS

Significant volume reduction in affected hippocampi in mTLE compared to controls was detected by manual hippocampal volume measurement (p < 0.01, effect size 33.2%), Freesurfer (p < 0.01, effect size 20.8%), and FSL-FIRST (p < 0.01, effect size 13.6%) after correction for brain volume. Freesurfer correlated reasonably (r = 0.74, p << 0.01) with this manual segmentation and FSL-FIRST relatively poorly (r = 0.47, p << 0.01). The spatial overlap between manual and automated segmentation was reduced in affected hippocampi, suggesting the accuracy of automated segmentation is reduced in pathologic brains.

DISCUSSION

Expert manual hippocampal volumetry is more sensitive than both automated methods for the detection of hippocampal atrophy associated with mTLE. In our study Freesurfer was the most sensitive to hippocampal atrophy in mTLE and could be used if expert manual segmentation is not available.

Appearance and incomplete label matching for diffeomorphic template based hippocampus segmentation

We present a robust, high-throughput, semiautomated template-based protocol for segmenting the hippocampus in temporal lobe epilepsy. The semiautomated component of this approach, which minimizes user effort while maximizing the benefit of human input to the algorithm, relies on "incomplete labeling." Incomplete labeling requires the user to quickly and approximately segment a few key regions of the hippocampus through a user-interface. Subsequently, this partial labeling of the hippocampus is combined with image similarity terms to guide volumetric diffeomorphic normalization between an individual brain and an unbiased disease-specific template, with fully labeled hippocampi. We solve this many-to-few and few-to-many matching problem, and gain robustness to inter and intrarater variability and small errors in user labeling, by embedding the template-based normalization within a probabilistic framework that examines both label geometry and appearance data at each label. We evaluate the reliability of this framework with respect to manual labeling and show that it increases minimum performance levels relative to fully automated approaches and provides high inter-rater reliability. Thus, this approach does not require expert neuroanatomical training and is viable for high-throughput studies of both the normal and the highly atrophic hippocampus.

Defining the human hippocampus in cerebral magnetic resonance images--an overview of current segmentation protocols

Due to its crucial role for memory processes and its relevance in neurological and psychiatric disorders, the hippocampus has been the focus of neuroimaging research for several decades. In vivo measurement of human hippocampal volume and shape with magnetic resonance imaging has become an important element of neuroimaging research. Nevertheless, volumetric findings are still inconsistent and controversial for many psychiatric conditions including affective disorders. Here we review the wealth of anatomical protocols for the delineation of the hippocampus in MR images, taking into consideration 71 different published protocols from the neuroimaging literature, with an emphasis on studies of affective disorders. We identified large variations between protocols in five major areas. 1) The inclusion/exclusion of hippocampal white matter (alveus and fimbria), 2) the definition of the anterior hippocampal-amygdala border, 3) the definition of the posterior border and the extent to which the hippocampal tail is included, 4) the definition of the inferior medial border of the hippocampus, and 5) the use of varying arbitrary lines. These are major sources of variance between different protocols. In contrast, the definitions of the lateral, superior, and inferior borders are less disputed. Directing resources to replication studies that incorporate characteristics of the segmentation protocols presented herein may help resolve seemingly contradictory volumetric results between prior neuroimaging studies and facilitate the appropriate selection of protocols for manual or automated delineation of the hippocampus for future research purposes.

Automatic segmentation of the hippocampus and the amygdala driven by hybrid constraints: method and validation

The segmentation from MRI of macroscopically ill-defined and highly variable structures, such as the hippocampus (Hc) and the amygdala (Am), requires the use of specific constraints. Here, we describe and evaluate a fast fully automatic hybrid segmentation that uses knowledge derived from probabilistic atlases and anatomical landmarks, adapted from a semi-automatic method. The algorithm was designed at the outset for application on images from healthy subjects and patients with hippocampal sclerosis. Probabilistic atlases were built from 16 healthy subjects, registered using SPM5. Local mismatch in the atlas registration step was automatically detected and corrected. Quantitative evaluation with respect to manual segmentations was performed on the 16 young subjects, with a leave-one-out strategy, a mixed cohort of 8 controls and 15 patients with epilepsy with variable degrees of hippocampal sclerosis, and 8 healthy subjects acquired on a 3 T scanner. Seven performance indices were computed, among which error on volumes RV and Dice overlap K. The method proved to be fast, robust and accurate. For Hc, results with the new method were: 16 young subjects {RV = 5%, K = 87%}; mixed cohort {RV = 8%, K = 84%}; 3 T cohort {RV = 9%, K = 85%}. Results were better than with atlas-based (thresholded probability map) or semi-automatic segmentations. Atlas mismatch detection and correction proved efficient for the most sclerotic Hc. For Am, results were: 16 young controls {RV = 7%, K = 85%}; mixed cohort {RV = 19%, K = 78%}; 3 T cohort {RV = 10%, K = 77%}. Results were better than with the semi-automatic segmentation, and were also better than atlas-based segmentations for the 16 young subjects.

Multivariate analysis of structural and diffusion imaging in traumatic brain injury

RATIONALE AND OBJECTIVES

Diffusion tensor (DT) and T1 structural magnetic resonance images provide unique and complementary tools for quantifying the living brain. We leverage both modalities in a diffeomorphic normalization method that unifies analysis of clinical datasets in a consistent and inherently multivariate (MV) statistical framework. We use this technique to study MV effects of traumatic brain injury (TBI).

MATERIALS AND METHODS

We contrast T1 and DT image-based measurements in the thalamus and hippocampus of 12 TBI survivors and nine matched controls normalized to a combined DT and T1 template space. The normalization method uses maps that are topology-preserving and unbiased. Normalization is based on the full tensor of information at each voxel and, simultaneously, the similarity between high-resolution features derived from T1 data. The technique is termed symmetric normalization for MV neuroanatomy (SyNMN). Voxel-wise MV statistics on the local volume and mean diffusion are assessed with Hotelling's T(2) test with correction for multiple comparisons.

RESULTS

TBI significantly (false discovery rate P < .05) reduces volume and increases mean diffusion at coincident locations in the mediodorsal thalamus and anterior hippocampus.

CONCLUSIONS

SyNMN reveals evidence that TBI compromises the limbic system. This TBI morphometry study and an additional performance evaluation contrasting SyNMN with other methods suggest that the DT component may aid normalization quality.

Hippocampus segmentation in MR images using atlas registration, voxel classification, and graph cuts

Since hippocampal volume has been found to be an early biomarker for Alzheimer's disease, there is large interest in automated methods to accurately, robustly, and reproducibly extract the hippocampus from MRI data. In this work we present a segmentation method based on the minimization of an energy functional with intensity and prior terms, which are derived from manually labelled training images. The intensity energy is based on a statistical intensity model that is learned from the training images. The prior energy consists of a spatial and regularity term. The spatial prior is obtained from a probabilistic atlas created by registering the training images to the unlabelled target image, and deforming and averaging the training labels. The regularity prior energy encourages smooth segmentations. The resulting energy functional is globally minimized using graph cuts. The method was evaluated using image data from a population-based study on diseases among the elderly. Two set of images were used: a small set of 20 manually labelled MR images and a larger set of 498 images, for which manual volume measurements were available, but no segmentations. This data was previously used in a volumetry study that found significant associations between hippocampal volume and cognitive decline and incidence of dementia. Cross-validation experiments with the labelled set showed similarity indices of 0.852 and 0.864 and mean surface distances of 0.40 and 0.36 mm for the left and right hippocampus. 83% of the automated segmentations of the large set were rated as 'good' by a trained observer. Also, the proposed method was used to repeat the manual hippocampal volumetry study. The automatically obtained hippocampal volumes showed significant associations with cognitive decline and dementia, similar to the manually measured volumes. Finally, direct quantitative and qualitative comparisons showed that the proposed method outperforms a multi-atlas based segmentation method.

Automatic volumetry on MR brain images can support diagnostic decision making

Background

Diagnostic decisions in clinical imaging currently rely almost exclusively on visual image interpretation. This can lead to uncertainty, for example in dementia disease, where some of the changes resemble those of normal ageing. We hypothesized that extracting volumetric data from patients' MR brain images, relating them to reference data and presenting the results as a colour overlay on the grey scale data would aid diagnostic readers in classifying dementia disease versus normal ageing.

Methods

A proof-of-concept forced-choice reader study was designed using MR brain images from 36 subjects. Images were segmented into 43 regions using an automatic atlas registration-based label propagation procedure. Seven subjects had clinically probable AD, the remaining 29 of a similar age range were used as controls. Seven of the control subject data sets were selected at random to be presented along with the seven AD datasets to two readers, who were blinded to all clinical and demographic information except age and gender. Readers were asked to review the grey scale MR images and to record their choice of diagnosis (AD or non-AD) along with their confidence in this decision. Afterwards, readers were given the option to switch on a false-colour overlay representing the relative size of the segmented structures. Colorization was based on the size rank of the test subject when compared with a reference group consisting of the 22 control subjects who were not used as review subjects. The readers were then asked to record whether and how the additional information had an impact on their diagnostic confidence.

Results

The size rank colour overlays were useful in 18 of 28 diagnoses, as determined by their impact on readers' diagnostic confidence. A not useful result was found in 6 of 28 cases. The impact of the additional information on diagnostic confidence was significant (p < 0.02).

Conclusion

Volumetric anatomical information extracted from brain images using automatic segmentation and presented as colour overlays can support diagnostic decision making.

FreeSurfer-initiated fully-automated subcortical brain segmentation in MRI using Large Deformation Diffeomorphic Metric Mapping

Fully-automated brain segmentation methods have not been widely adopted for clinical use because of issues related to reliability, accuracy, and limitations of delineation protocol. By combining the probabilistic-based FreeSurfer (FS) method with the Large Deformation Diffeomorphic Metric Mapping (LDDMM)-based label-propagation method, we are able to increase reliability and accuracy, and allow for flexibility in template choice. Our method uses the automated FreeSurfer subcortical labeling to provide a coarse-to-fine introduction of information in the LDDMM template-based segmentation resulting in a fully-automated subcortical brain segmentation method (FS+LDDMM). One major advantage of the FS+LDDMM-based approach is that the automatically generated segmentations generated are inherently smooth, thus subsequent steps in shape analysis can directly follow without manual post-processing or loss of detail. We have evaluated our new FS+LDDMM method on several databases containing a total of 50 subjects with different pathologies, scan sequences and manual delineation protocols for labeling the basal ganglia, thalamus, and hippocampus. In healthy controls we report Dice overlap measures of 0.81, 0.83, 0.74, 0.86 and 0.75 for the right caudate nucleus, putamen, pallidum, thalamus and hippocampus respectively. We also find statistically significant improvement of accuracy in FS+LDDMM over FreeSurfer for the caudate nucleus and putamen of Huntington's disease and Tourette's syndrome subjects, and the right hippocampus of Schizophrenia subjects.

Subcortical and cerebellar atrophy in mesial temporal lobe epilepsy revealed by automatic segmentation

PURPOSE

To determine the validity and utility of using automated subcortical segmentation to identify atrophy of the hippocampus and other subcortical and cerebellar structures in patients with mesial temporal lobe epilepsy (MTLE).

METHODS

Volumetric MRIs were obtained on 21 patients with MTLE (11 right, 10 left) and 21 age- and gender-matched healthy controls. Labeling of subcortical and cerebellar structures was accomplished using automated reconstruction software (FreeSurfer). Multivariate analysis of covariance (MANCOVA) was used to explore group differences in intracranial-normalized, age-adjusted volumes and structural asymmetries. Step-wise discriminant function analysis was used to identify the linear combination of volumes that optimized classification of individual subjects.

RESULTS

Results revealed the expected reduction in hippocampal volume on the side ipsilateral to the seizure focus, as well as bilateral reductions in thalamic and cerebellar gray matter volume. Analysis of structural asymmetries revealed significant asymmetry in the hippocampus and putamen in patients compared to controls. The discriminant function analysis revealed that patients with right and left MTLE were best distinguished from one another using a combination of subcortical volumes that included the right and left hippocampus and left thalamus (91-100% correct classification using cross-validation).

DISCUSSION

Volumetric data obtained with automated segmentation of subcortical and cerebellar structures approximate data from previous studies based on manual tracings. Our data suggest that automated segmentation can provide a clinically useful means of evaluating the nature and extent of structural damage in patients with MTLE and may increase diagnostic classification of patients, especially when hippocampal atrophy is mild.

Fully automatic segmentation of the hippocampus and the amygdala from MRI using hybrid prior knowledge

The segmentation of macroscopically ill-defined and highly variable structures, such as the hippocampus Hc and the amygdala Am, from MRI requires specific constraints. Here, we describe and evaluate a hybrid segmentation method that uses knowledge derived from a probabilistic atlas and from anatomical landmarks based on stable anatomical characteristics of the structures. Combined in a previously published semi-automatic segmentation method, they lead to a fast, robust and accurate fully automatic segmentation of Hc and Am. The probabilistic atlas was built from 16 young controls and registered with the "unified segmentation" of SPM5. The algorithm was quantitatively evaluated with respect to manual segmentation on two MRI datasets: the 16 young controls, with a leave-one-out strategy, and a mixed cohort of 8 controls and 15 subjects with epilepsy with variable hippocampal sclerosis. The segmentation driven by hybrid knowledge leads to greatly improved results compared to that obtained by registration of the thresholded atlas alone: mean overlap for Hc on the 16 young controls increased from 78% to 87% (p < 0.001) and on the mixed cohort from 73% to 82% (p < 0.001) while the error on volumes decreased from 10% to 7% (p < 0.005) and from 18% to 8% (p < 0.001), respectively. Automatic results were better than the semi-automatic results: for the 16 young controls, average overlap increased from 84% to 87% (p < 0.001) for Hc and from 81% to 84% (p < 0.002) for Am, with equivalent improvements in volume error.

Voxel-based morphometry of temporal lobe epilepsy: an introduction and review of the literature

We review the applications and results of voxel-based morphometry (VBM) studies that have reported brain changes associated with temporal lobe epilepsy (TLE). A PubMed search yielded 18 applications of VBM to study brain abnormalities in patients with TLE up to May 2007. Across studies, 26 brain regions were found to be significantly reduced in volume relative to healthy controls. There was a strong asymmetrical distribution of temporal lobe abnormalities preferentially observed ipsilateral to the seizure focus, particularly of the hippocampus (82.35% of all studies), parahippocampal gyrus (47.06%), and entorhinal (23.52%) cortex. The contralateral hippocampus was reported as abnormal in 17.65% of studies. There was a much more bilateral distribution of extratemporal lobe atrophy, preferentially affecting the thalamus (ipsilateral = 61.11%, contralateral = 50%) and parietal lobe (ipsilateral = 47.06%, contralateral = 52.94%). VBM generally reveals a distribution of brain abnormalities in patients with TLE consistent with the region-of-interest neuroimaging and postmortem literature. It is unlikely that VBM has any clinical utility given the lack of robustness for individual comparisons. However, VBM may help elucidate some unresolved important research questions such as how recurrent temporal lobe seizures affect hippocampal and extrahippocampal morphology using serial imaging acquisitions.