Non-linear registration improves statistical power to detect hippocampal atrophy in aging and dementia

OBJECTIVE

To compare the performance of different methods for determining hippocampal atrophy rates using longitudinal MRI scans in aging and Alzheimer's disease (AD).

BACKGROUND

Quantifying hippocampal atrophy caused by neurodegenerative diseases is important to follow the course of the disease. In dementia, the efficacy of new therapies can be partially assessed by measuring their effect on hippocampal atrophy. In radiotherapy, the quantification of radiation-induced hippocampal volume loss is of interest to quantify radiation damage. We evaluated plausibility, reproducibility and sensitivity of eight commonly used methods to determine hippocampal atrophy rates using test-retest scans.

MATERIALS AND METHODS

Manual, FSL-FIRST, FreeSurfer, multi-atlas segmentation (MALF) and non-linear registration methods (Elastix, NiftyReg, ANTs and MIRTK) were used to determine hippocampal atrophy rates on longitudinal T1-weighted MRI from the ADNI database. Appropriate parameters for the non-linear registration methods were determined using a small training dataset (N = 16) in which two-year hippocampal atrophy was measured using test-retest scans of 8 subjects with low and 8 subjects with high atrophy rates. On a larger dataset of 20 controls, 40 mild cognitive impairment (MCI) and 20  AD patients, one-year hippocampal atrophy rates were measured. A repeated measures ANOVA analysis was performed to determine differences between controls, MCI and AD patients. For each method we calculated effect sizes and the required sample sizes to detect one-year volume change between controls and MCI (NCTRL_MCI) and between controls and AD (NCTRL_AD). Finally, reproducibility of hippocampal atrophy rates was assessed using within-session rescans and expressed as an average distance measure DAve, which expresses the difference in atrophy rate, averaged over all subjects. The same DAve was used to determine the agreement between different methods.

RESULTS

Except for MALF, all methods detected a significant group difference between CTRL and AD, but none could find a significant difference between the CTRL and MCI. FreeSurfer and MIRTK required the lowest sample sizes (FreeSurfer: NCTRL_MCI = 115, NCTRL_AD = 17 with DAve = 3.26%; MIRTK: NCTRL_MCI = 97, NCTRL_AD = 11 with DAve = 3.76%), while ANTs was most reproducible (NCTRL_MCI = 162, NCTRL_AD = 37 with DAve = 1.06%), followed by Elastix (NCTRL_MCI = 226, NCTRL_AD = 15 with DAve = 1.78%) and NiftyReg (NCTRL_MCI = 193, NCTRL_AD = 14 with DAve = 2.11%). Manually measured hippocampal atrophy rates required largest sample sizes to detect volume change and were poorly reproduced (NCTRL_MCI = 452, NCTRL_AD = 87 with DAve = 12.39%). Atrophy rates of non-linear registration methods also agreed best with each other.

DISCUSSION AND CONCLUSION

Non-linear registration methods were most consistent in determining hippocampal atrophy and because of their better reproducibility, methods, such as ANTs, Elastix and NiftyReg, are preferred for determining hippocampal atrophy rates on longitudinal MRI. Since performances of non-linear registration methods are well comparable, the preferred method would mostly depend on computational efficiency.

Inter-rater agreement in glioma segmentations on longitudinal MRI

Background

Tumor segmentation of glioma on MRI is a technique to monitor, quantify and report disease progression. Manual MRI segmentation is the gold standard but very labor intensive. At present the quality of this gold standard is not known for different stages of the disease, and prior work has mainly focused on treatment-naive glioblastoma. In this paper we studied the inter-rater agreement of manual MRI segmentation of glioblastoma and WHO grade II-III glioma for novices and experts at three stages of disease. We also studied the impact of inter-observer variation on extent of resection and growth rate.

Methods

In 20 patients with WHO grade IV glioblastoma and 20 patients with WHO grade II-III glioma (defined as non-glioblastoma) both the enhancing and non-enhancing tumor elements were segmented on MRI, using specialized software, by four novices and four experts before surgery, after surgery and at time of tumor progression. We used the generalized conformity index (GCI) and the intra-class correlation coefficient (ICC) of tumor volume as main outcome measures for inter-rater agreement.

Results

For glioblastoma, segmentations by experts and novices were comparable. The inter-rater agreement of enhancing tumor elements was excellent before surgery (GCI 0.79, ICC 0.99) poor after surgery (GCI 0.32, ICC 0.92), and good at progression (GCI 0.65, ICC 0.91). For non-glioblastoma, the inter-rater agreement was generally higher between experts than between novices. The inter-rater agreement was excellent between experts before surgery (GCI 0.77, ICC 0.92), was reasonable after surgery (GCI 0.48, ICC 0.84), and good at progression (GCI 0.60, ICC 0.80). The inter-rater agreement was good between novices before surgery (GCI 0.66, ICC 0.73), was poor after surgery (GCI 0.33, ICC 0.55), and poor at progression (GCI 0.36, ICC 0.73). Further analysis showed that the lower inter-rater agreement of segmentation on postoperative MRI could only partly be explained by the smaller volumes and fragmentation of residual tumor. The median interquartile range of extent of resection between raters was 8.3% and of growth rate was 0.22 mm/year.

Conclusion

Manual tumor segmentations on MRI have reasonable agreement for use in spatial and volumetric analysis. Agreement in spatial overlap is of concern with segmentation after surgery for glioblastoma and with segmentation of non-glioblastoma by non-experts.

•

Inter-rater agreement for longitudinal glioma segmentation was determined.

•

Agreement between 4 experts was higher than between 4 novices.

•

Three time-points of glioblastoma (WHO IV) and diffuse glioma (WHO II-III) are studied.

•

Impact on extent of resection and growth rate measurements was determined.

How does upper extremity Fugl-Meyer motor score relate to resting-state EEG in chronic stroke? A power spectral density analysis

OBJECTIVE

We investigated the potential added value of high-density resting-state EEG by addressing differences with healthy individuals and associations with Fugl-Meyer motor assessment of the upper extremity (FM-UE) scores in chronic stroke.

METHODS

Twenty-one chronic stroke survivors with initial upper limb paresis and eleven matched controls were included. Group differences regarding resting-state EEG parameters (Delta Alpha ratio (DAR) and pairwise-derived Brain Symmetry Index (BSI)) and associations with FM-UE were investigated, as well as lateralization of BSI and the value of different frequency bands.

RESULTS

Chronic stroke survivors showed higher BSI compared to controls (p < 0.001), most pronounced in delta and theta frequency bands (p < 0.0001; p < 0.001). In the delta and theta band, BSI was significantly negatively associated with FM-UE (both p = 0.008) corrected for confounding factors. DAR showed no differences between groups nor association with FM-UE. Directional BSI showed increased power in the affected versus the unaffected hemisphere.

CONCLUSIONS

Asymmetry in spectral power between hemispheres was present in chronic stroke, most pronounced in low frequencies and related to upper extremity motor function deficit.

SIGNIFICANCE

BSI is related to motor impairment and higher in chronic stroke patients compared to healthy controls, suggesting that BSI may be a marker of selective motor control.

Inter-observer variation of hippocampus delineation in hippocampal avoidance prophylactic cranial irradiation

BACKGROUND

Hippocampal avoidance prophylactic cranial irradiation (HA-PCI) techniques have been developed to reduce radiation damage to the hippocampus. An inter-observer hippocampus delineation analysis was performed and the influence of the delineation variability on dose to the hippocampus was studied.

MATERIALS AND METHODS

For five patients, seven observers delineated both hippocampi on brain MRI. The intra-class correlation (ICC) with absolute agreement and the generalized conformity index (CI_gen) were computed. Median surfaces over all observers' delineations were created for each patient and regional outlining differences were analysed. HA-PCI dose plans were made from the median surfaces and we investigated whether dose constraints in the hippocampus could be met for all delineations.

RESULTS

The ICC for the left and right hippocampus was 0.56 and 0.69, respectively, while the CI_gen ranged from 0.55 to 0.70. The posterior and anterior-medial hippocampal regions had most variation with SDs ranging from approximately 1 to 2.5 mm. The mean dose (D_mean) constraint was met for all delineations, but for the dose received by 1% of the hippocampal volume (D_1%) violations were observed.

CONCLUSION

The relatively low ICC and CI_gen indicate that delineation variability among observers for both left and right hippocampus was large. The posterior and anterior-medial border have the largest delineation inaccuracy. The hippocampus D_mean constraint was not violated.

Interactions between different EEG frequency bands and their effect on alpha-fMRI correlations

In EEG/fMRI correlation studies it is common to consider the fMRI BOLD as filtered version of the EEG alpha power. Here the question is addressed whether other EEG frequency components may affect the correlation between alpha and BOLD. This was done comparing the statistical parametric maps (SPMs) of three different filter models wherein either the free or the standard hemodynamic response functions (HRF) were used in combination with the full spectral bandwidth of the EEG. EEG and fMRI were co-registered in a 30 min resting state condition in 15 healthy young subjects. Power variations in the delta, theta, alpha, beta and gamma bands were extracted from the EEG and used as regressors in a general linear model. Statistical parametric maps (SPMs) were computed using three different filter models, wherein either the free or the standard hemodynamic response functions (HRF) were used in combination with the full spectral bandwidth of the EEG. Results show that the SPMs of different EEG frequency bands, when significant, are very similar to that of the alpha rhythm. This is true in particular for the beta band, despite the fact that the alpha harmonics were discarded. It is shown that inclusion of EEG frequency bands as confounder in the fMRI-alpha correlation model has a large effect on the resulting SPM, in particular when for each frequency band the HRF is extracted from the data. We conclude that power fluctuations of different EEG frequency bands are mutually highly correlated, and that a multi frequency model is required to extract the SPM of the frequency of interest from EEG/fMRI data. When no constraints are put on the shapes of the HRFs of the nuisance frequencies, the correlation model looses so much statistical power that no correlations can be detected.

A study of the brain's resting state based on alpha band power, heart rate and fMRI

Considering that there are several theoretical reasons why fMRI data is correlated to variations in heart rate, these correlations are explored using experimental resting state data. In particular, the possibility is discussed that the "default network", being a brain area that deactivates during non-specific general tasks, is a hemodynamic effect caused by heart rate variations. Of fifteen healthy controls ECG, EEG and fMRI were co-registered. Slice time dependent heart rate regressors were derived from the ECG data and correlated to fMRI using a linear correlation analysis where the impulse response is estimated from the data. It was found that in most subjects substantial correlations between heart rate variations and fMRI exist, both within the brain and at the ventricles. The brain areas with high correlation to heart rate are different from the "default network" and the response functions deviate from the canonical hemodynamic response function. Furthermore, a general negative correlation was found between heart beat intervals (reverse of heart rate) and alpha power. We interpret this finding by assuming that subject's state varies between drowsiness and wakefulness. Finally, given this large correlation, we re-examined the contribution of heart rate variations to earlier reported fMRI/alpha band correlations, by adding heart rate regressors as confounders. It was found that inclusion of these confounders most often had a negligible effect. From its strong correlation to alpha power, we conclude that the heart rate variations contain important physiological information about subject's resting state. However, it does not provide a full explanation of the behaviour of the "default network". Its application as confounder in fMRI experiments is a relatively small computational effort, but may have a substantial impact in paradigms where heart rate is controlled by the stimulus.

Artifact removal in co-registered EEG/fMRI by selective average subtraction

OBJECTIVE

Co-registration of EEG (electroencephalogram) and fMRI (functional magnetic resonance imaging) remains a challenge due to the large artifacts induced on the EEG by the MR (magnetic resonance) sequence magnetic fields. Thus, we present an algorithm, based on the average-subtraction method, which is able to correct EEG data for gradient and pulse artifacts.

METHODS

MR sequence timing parameters are estimated from the EEG data and both slice and volume artifact templates are subtracted from the data. A clustering algorithm is proposed to account for the variability of the pulse artifact.

RESULTS

The algorithm is able to keep the spontaneous EEG as well as visual evoked potentials (VEPs), while removing gradient and pulse artifacts with only a subtraction of selectively averaged data. In the frequency domain, the artifact frequencies are strongly attenuated. Estimated MR sequence time parameters showed that the correction is extremely sensitive to the slice time value. Pulse artifact clustering showed that most of the variability is due to the time jitter of the pulse artifact markers.

CONCLUSIONS

Selective subtraction of averages in combination with proper time alignment is enough to remove most of the MR-induced artifacts.

SIGNIFICANCE

Clean EEG can be obtained from raw signals that are corrupted by MR-induced artifacts during simultaneous EEG-fMRI scanning without using dedicated hardware to synchronize EEG and fMRI clocks.

Magnetoencephalographic evaluation of resting-state functional connectivity in Alzheimer's disease

Statistical interdependencies between magnetoencephalographic signals recorded over different brain regions may reflect the functional connectivity of the resting-state networks. We investigated topographic characteristics of disturbed resting-state networks in Alzheimer's disease patients in different frequency bands. Whole-head 151-channel MEG was recorded in 18 Alzheimer patients (mean age 72.1 years, SD 5.6; 11 males) and 18 healthy controls (mean age 69.1 years, SD 6.8; 7 males) during a no-task eyes-closed resting state. Pair-wise interdependencies of MEG signals were computed in six frequency bands (delta, theta, alpha1, alpha2, beta and gamma) with the synchronization likelihood (a nonlinear measure) and coherence and grouped into long distance (intra- and interhemispheric) and short distance interactions. In the alpha1 and beta band, Alzheimer patients showed a loss of long distance intrahemispheric interactions, with a focus on left fronto-temporal/parietal connections. Functional connectivity was increased in Alzheimer patients locally in the theta band (centro-parietal regions) and the beta and gamma band (occipito-parietal regions). In the Alzheimer group, positive correlations were found between alpha1, alpha2 and beta band synchronization likelihood and MMSE score. Resting-state functional connectivity in Alzheimer's disease is characterized by specific changes of long and short distance interactions in the theta, alpha1, beta and gamma bands. These changes may reflect loss of anatomical connections and/or reduced central cholinergic activity and could underlie part of the cognitive impairment.

Correlating the alpha rhythm to BOLD using simultaneous EEG/fMRI: Inter-subject variability

Simultaneous recording of electroencephalogram/functional magnetic resonance images (EEG/fMRI) was applied to identify blood oxygenation level-dependent (BOLD) changes associated with spontaneous variations of the alpha rhythm, which is considered the hallmark of the brain resting state. The analysis was focused on inter-subject variability associated with the resting state. Data from 7 normal subjects are presented. Confirming earlier findings, three subjects showed a negative correlation between the BOLD signal and the average power time series within the alpha band (8--12 Hz) in extensive areas of the occipital, parietal and frontal lobes. In small thalamic areas, the BOLD signal was positively correlated with the alpha power. For subjects 3 and 4, who displayed two different states during the data acquisition time, it was shown that the corresponding correlation patterns were different, thus demonstrating the state dependency of the results. In subject 5, the changes in BOLD were observed mainly in the frontal and temporal lobes. Subject 6 only showed positive correlations, thus contradicting the negative BOLD alpha power cortical correlations that were found in most subjects. Results suggest that the resting state varies over subjects and, sometimes, even within one subject. As the resting state plays an important role in many fMRI experiments, the inter-subject variability of this state should be addressed when comparing fMRI results from different subjects.

The influence of brain tumor treatment on pathological delta activity in MEG

The goal of the MEG study was to investigate the influence of tumor treatment on pathological delta activity (1-4 Hz). The treatment consisted of neurosurgery, and in some of the patients, additional radiotherapy. MEG and MR recordings were made both before and after the treatment in 17 patients. The signal power in the delta frequency band was determined for each recording. The malignant tumors were associated with large tumor volumes. Furthermore, both malignant tumors and tumor volume were associated with high signal powers in the delta band, indicating a correlation of delta power with the severity of the lesions. In all patients with high grade tumors, the delta power was lower after the treatment. The sources underlying the delta signals were estimated with an automatic single dipole analysis method. Estimated sources were projected onto MR scans. Preoperatively 14 clusters of equivalent sources describing focal activity were found in 12 out of 17 patients. Thirteen of these clusters were located near the tumor, and one cluster near an edema border. The locations near tumors are plausible and suggest that in general the source estimation was reliable. After the operation, 13 such clusters were found in 12 patients. Eleven clusters were located near the lesion border and one cluster near the edema border. Furthermore a cluster contralateral to the lesion in the other hemisphere indicated that brain lesions can affect the functioning of more distant brain areas than just the peritumoral brain tissue. Of the 12 patients who had preoperatively peritumoral clusters, 11 patients had postoperatively perilesional sources. In these cases the shift in source locations was in general considerably smaller than the dimension of the preoperative tumors. This finding indicates that similar areas generate the pre- and postoperative delta activity. Furthermore, focal delta sources were found in a case without tumor recurrence, and also in cases that most tumor tissue was removed. These findings suggest that the pathology underlying the slow waves is not the presence of the tumor bulk but the structural damage done by the tumors on the surrounding white/gray matter.

A whole head MEG study of the amplitude-modulation-following response: phase coherence, group delay and dipole source analysis

OBJECTIVE

The amplitude-modulation-following response (AMFR) is the frequency component detectable in the electroencephalogram (EEG) or magnetoencephalography (MEG) corresponding to the modulation frequency of an amplitude modulated tone used as a continuous acoustic stimulus. Various properties of the AMFR depend on modulation frequency, suggesting that different generators along the auditory pathway are involved. The present study addresses these issues on the basis of a whole head MEG experiment.

METHODS

AM tones with modulators in the 40 Hz and 80 Hz range were presented unilaterally to 10 normal hearing subjects. Biomagnetic responses were recorded with a 151 channel MEG system. The data analysis concentrated on the phase coherence of the responses, group delays and the estimated location of underlying equivalent dipole sources.

RESULTS

MEG AMFR is more reliably detected in the 40 Hz than in the 80 Hz range. Both response amplitude and phase coherence indicate clear bilateral activation over the parietal/temporal region. Dipole source analysis confirms that sources are located in or near the auditory cortex. Group delays at 80 Hz are shorter than at 40 Hz.

CONCLUSIONS

In both modulation frequency ranges MEG responses are dominated by activity in the auditory cortex, in apparent contrast with EEG data in the literature, pointing to dominant contributions of thalamic sources to the 80 Hz AMFR.

Spike cluster analysis in neocortical localization related epilepsy yields clinically significant equivalent source localization results in magnetoencephalogram (MEG)

OBJECTIVE

In magnetoencephalogram (MEG) recordings of patients with epilepsy several types of sharp transients with different spatiotemporal distributions are commonly present. Our objective was to develop a computer based method to identify and classify groups of epileptiform spikes, as well as other transients, in order to improve the characterization of irritative areas in the brain of epileptic patients.

METHODS

MEG data centered on selected spikes were stored in signal matrices of C channels by T time samples. The matrices were normalized and euclidean distances between spike representations in vector space R(CxT) were input to a Ward's hierarchical clustering algorithm.

RESULTS

The method was applied to MEG data from 4 patients with localization-related epilepsy. For each patient, distinct spike subpopulations were found with clearly different topographical field maps. Inverse computations to selected spike subaverages yielded source solutions in agreement with seizure classification and location of structural lesions, if present, on magnetic resonance images.

CONCLUSIONS

With the proposed method a reliable categorization of epileptiform spikes is obtained, that can be applied in an automatic way. Computation of subaverages of similar spikes enhances the signal-to-noise ratio of spike field maps and allows for more accurate reconstruction of sources generating the epileptiform discharges.

In vivo measurement of the brain and skull resistivities using an EIT-based method and the combined analysis of SEF/SEP data

Results of "in vivo" measurements of the skull and brain resistivities are presented for six subjects. Results are obtained using two different methods, based on spherical head models. The first method uses the principles of electrical impedance tomography (EIT) to estimate the equivalent electrical resistivities of brain (rhobrain), skull (rhoskull) and skin (rhoskin) according to. The second one estimates the same parameters through a combined analysis of the evoked somatosensory cortical response, recorded simultaneously using magnetoencephalography (MEG) and electroencephalography (EEG). The EIT results, obtained with the same relative skull thickness (0.05) for all subjects, show a wide variation of the ratio rhoskull/rhobrain among subjects (average = 72, SD = 48%). However, the rhoskull/rhobrain ratios of the individual subjects are well reproduced by combined analysis of somatosensory evoked fields (SEF) and somatosensory evoked potentials (SEP). These preliminary results suggest that the rhoskull/rhobrain variations over subjects cannot be disregarded in the EEG inverse problem (IP) when a spherical model is used. The agreement between EIT and SEF/SEP points to the fact that whatever the source of variability, the proposed EIT-based method <Au: Addition of "method" O.K? appears to have the potential to reduce systematic errors in EEG IP associated to the misspecification of rhoskull/rhobrain, rhobrain, rhoskull and rhoskin.

Localization of fast MEG waves in patients with brain tumors and epilepsy

It was investigated if single dipole analysis of spontaneous fast waves (>8 Hz) can be used to determine the location of the epileptic focus. Automatic dipole analysis was applied to MEG data of 25 patients with intracranial tumors and epilepsy. The frequency range of 8-50 Hz was divided into standard EEG bands. MEG results were overlaid on the MRI scans of the patients. Dipoles describing fast wave fields were located in the parietal/occipital cortex, and not at tumor border zones. In the cases that the dipoles were lateralized there was no clear preference to be located ipsi or contralateral to the tumor. However the generators of epileptic activity in these patients are thought to be located in the border areas of the tumors. Therefore it seems unlikely that the dipole locations describing fast waves are related to the epileptic zones in patients with brain tumors and epilepsy. A remarkable finding is that lateralized dipoles tend to be located in the right hemisphere and not in the left hemisphere. This appears to reflect an asymmetry of possibly normal background activity.

Magnetic source imaging contributes to the presurgical identification of sensorimotor cortex in patients with frontal lobe epilepsy

OBJECTIVE

One of the primary goals of preoperative evaluation of patients considered to be candidates for epilepsy surgery is the delineation of eloquent cortex adjacent to the area of resection. The aim of this study is the functional localization of the sensorimotor cortex in relation to an epileptogenic frontal lobe lesion, thus enabling a more complete resection in these patients while minimizing the risk of postoperative neurological deficits.

METHODS

Participating in this study were patients with epilepsy, diagnosed as being related to a left or right frontal lobe lesion. Magnetoencephalographic responses evoked by electrical stimulation of the left and right hand median nerve were localized using single time-point equivalent dipole (ED) modeling, taking into account the realistic shape of the head. Instead of relying on the primary component (N/P 20) of the somatosensory evoked magnetic fields (SEFs) in this study ED fits were obtained for each time-point of the somatosensory evoked responses. On a cortical rendering, the reconstructed dipoles were depicted relative to the anatomy obtained from 3D-magnetic resonance imaging.

RESULTS

The results of single time-point ED analysis including all the components of the responses indicated that the sources underlying the SEFs are located at the borders of the central sulcus (CS). The opposite direction of the sources underlying, respectively, the primary and subsequent late component of the SEFs indicated distinct sources located at the opposite banks of the CS. These sources, therefore, might correspond to the sensory hand projection area and the primary motor area of the sensorimotor cortex. It appeared that the location of the EDs obtained for the SEFs of 4 of the 7 patients studied were asymmetric for the left and right hemisphere, probably because of a displacement of the sensorimotor areas relative to the CS. The systematic assessment of the dipole fits compared to brain anatomy confirmed that volume conduction changes due to the lesion were not responsible for these observed deviations, thus leaving as explanation space-occupying and neurophysiological changes due to the lesion.

A fast method to derive realistic BEM models for E/MEG source reconstruction

A fast method for segmentation of a subject's skin, skull or brain compartment for electroencephalogram (EEG)/magnetoencephalogram (MEG) (E/MEG) source localization is proposed. The method is based on a description of volumes with spherical harmonics and a database of exact surfaces. Using the spherical harmonic coefficients, sets of basis surfaces are obtained for each compartment. New segmentations can be acquired by combining the appropriate basis surfaces to describe a delineation of the volume in a limited number of magnetic resonance (MR) slices. Alternatively, a representation of the skin can be derived from digitized head shape. Skull and brain then can be predicted from the skin representation with a prediction model also obtained from the segmentation database. Database segmentations were recomputed with the proposed method. Mean deviations from the originals were about 2 and 3 mm for compartments derived from MR and head shape. Dipole simulations with original surfaces for forward and computed segmentations for inverse calculations showed average dipole mislocalizations of 1.6 and 3.3 mm, respectively. With the proposed method highly accurate segmentation can be performed with much less effort and in much less time compared with other techniques. The method also is applicable when MR data is unavailable but a digitization of the head is.

The use of an MEG device as 3D digitizer and motion monitoring system

An algorithm is described that localizes a set of simultaneously activated coils using MEG detectors. These coil positions are used for continuous or intermittent head position registration during long MEG sessions, to coregistrate MR and MEG data and to localize EEG electrodes attached to the scalp, when EEG and MEG are recorded simultaneously. The algorithm is based on a mathematical model in which the coils are described as stationary magnetic dipoles with known source time functions. This knowledge makes it possible to detect and remove bad channels automatically. It is also assumed that the source time functions are orthogonal. Therefore, the localization problem splits into independent localization problems. for each coil. The method is validated in a phantom experiment, where the relative coil positions were known. From this experiment it is found that the average error is 0.25 cm. An error of 0.23 cm was found in an experiment where 64 electrode positions were measured four times independently. Examples of the applications of the method are presented. Our method eliminates the use of an external 3D digitizer and maps the MEG directly onto other modalities. This is not only a practical advantage, but it also reduces the gross registration error. Furthermore, head motions can be monitored and MEG data can be corrected for these motions.

Event-related desynchronization during anticipatory attention for an upcoming stimulus: a comparative EEG/MEG study

OBJECTIVES

Our neurophysiological model of anticipatory behaviour (e.g. Acta Psychol 101 (1999) 213; Bastiaansen et al., 1999a) predicts an activation of (primary) sensory cortex during anticipatory attention for an upcoming stimulus. In this paper we attempt to demonstrate this by means of event-related desynchronization (ERD).

METHODS

Five subjects performed a time estimation task, and were informed about the quality of their time estimation by either visual or auditory stimuli providing Knowledge of Results (KR). EEG and MEG were recorded in separate sessions, and ERD was computed in the 8-10 and 10-12 Hz frequency bands for both datasets.

RESULTS

Both in the EEG and the MEG we found an occipitally maximal ERD preceding the visual KR for all subjects. Preceding the auditory KR, no ERD was present in the EEG, whereas in the MEG we found an ERD over the temporal cortex in two of the 5 subjects. These subjects were also found to have higher levels of absolute power over temporal recording sites in the MEG than the other subjects, which we consider to be an indication of the presence of a 'tau' rhythm (e.g. Neurosci Lett 222 (1997) 111).

CONCLUSIONS

It is concluded that the results are in line with the predictions of our neurophysiological model.

The localization of spontaneous brain activity: first results in patients with cerebral tumors

OBJECTIVE

From EEG studies, it is known that structural brain lesions are accompanied by abnormal rhythmic electric activity. With the better spatial resolution of MEG, MEG dipole analysis can extend the knowledge based on EEG power spectra. This study presents the first results of a completely automatic analysis method applied to spontaneous MEG.

METHODS

Spontaneous MEG data of 5 patients with cerebral brain tumors and 4 controls were collected using a whole-head MEG system. Signals were bandpass-filtered with cut-off frequencies according to standard EEG bands. A moving dipole model was fitted to samples with at least twice the average sample power. Dipoles explaining 90% or more of the magnetic variance were projected onto a matched MR scan.

RESULTS

In controls, dipole distributions are symmetrical with respect to the mid-sagittal plane whereas distributions in patients often are asymmetrical to it. Dipoles describing gamma activity were located contralateral, and dipoles describing delta and theta activity were located ipsilateral to lesions.

CONCLUSIONS

The automatic method gives plausible 3-dimensional information about generator foci of abnormal slow waves and other rhythms with respect to lesion foci and thereby adds physiological knowledge to that derived from EEG power spectra.

The accuracy of image registration for the brain and the nasopharynx using external anatomical landmarks

We investigated the accuracy of 3D image registration using markers that are repeatedly applied to external anatomical landmarks on the head. The purpose of this study is to establish a lower limit of the errors that would occur in, for instance, MRI-SPECT matching, which in some situations can only be achieved using external landmarks. Marker matching was compared with (single-modality) volume matching for 20 MRI scans. The results were compared with a published expression for the target registration error (TRE) which gives the 3D distribution of the mismatch between both scans. It was found that the main error source is reapplying the external markers on the anatomical landmarks. The published expression describes the relative distribution of the TRE in space well, but tends to underestimate the actual registration error. This deviation is due to anisotropy in the error distribution of the marker position (errors in the direction perpendicular to the skin surface are in general much smaller than errors in other directions). A simulation of marker matching with anisotropy in the errors confirmed this finding. With four reapplied markers, the TRE is 6 mm or smaller in most regions of the head.

The application of electrical impedance tomography to reduce systematic errors in the EEG inverse problem--a simulation study

In this paper we propose a new method, using the principles of electrical impedance tomography (EIT), to correct for the systematic errors in the inverse problem (IP) of electroencephalography (EEG) that arise from the wrong specification of the electrical conductivities of the head compartments. By injecting known currents into pairs of electrodes and measuring the resulting potential differences recorded from the other electrodes, the equivalent conductivities of brain (sigma3), skull (sigma2) and scalp (sigma1) can be estimated. Since the geometry of the head is assumed to be known, the electrical conductivities remain as the only unknown parameters to be estimated. These conductivities can then be used in the inverse problem of EEG. The simulations performed in this study, using a three-layer sphere to model the head, prove the feasibility of the method, theoretically. Even in the presence of simulated noise with a value of signal-to-noise ratio (SNR) equal to 10, estimations of the electrical conductivities within 5% of the true values were obtained. Simulations showed the existence of a strong relation between errors in the skull thickness and the EIT estimated conductivities. If the skull thickness is wrongly specified, for example overestimated by a factor of two, the conductivity determined by EIT is also overestimated by a factor of two. Simulations showed that this compensation effect also works in the inverse problem of EEG. Application of the proposed method reduces systematic errors in the dipole localization, up to an amount of 1 cm. However it proved to be ineffective to decrease the dipole strength error.

The boundary element method in the forward and inverse problem of electrical impedance tomography

In this paper, a new formulation of the reconstruction problem of electrical impedance tomography (EIT) is proposed. Instead of reconstructing a complete two-dimensional picture, a parameter representation of the gross anatomy is formulated, of which the optimal parameters are determined by minimizing a cost function. The two great advantages of this method are that the number of unknown parameters of the inverse problem is drastically reduced and that quantitative information of interest (e.g., lung volume) is estimated directly from the data, without image segmentation steps. The forward problem of EIT is to compute the potentials at the voltage measuring electrodes, for a given set of current injection electrodes and a given conductivity geometry. In this paper, it is proposed to use an improved boundary element method (BEM) technique to solve the forward problem, in which flat boundary elements are replaced by polygonal ones. From a comparison with the analytical solution of the concentric circle model, it appears that the use of polygonal elements greatly improves the accuracy of the BEM, without increasing the computation time. In this formulation, the inverse problem is a nonlinear parameter estimation problem with a limited number of parameters. Variants of Powell's and the simplex method are used to minimize the cost function. The applicability of this solution of the EIT problem was tested in a series of simulation studies. In these studies, EIT data were simulated using a standard conductor geometry and it was attempted to find back this geometry from random starting values. In the inverse algorithm, different current injection and voltage measurement schemes and different cost functions were compared. In a simulation study, it was demonstrated that a systematic error in the assumed lung conductivity results in a proportional error in the lung cross sectional area. It appears that our parametric formulation of the inverse problem leads to a stable minimization problem, with a high reliability, provided that the signal-to-noise ratio is about ten or higher.

Magnetic source imaging in fixation-off sensitivity: relationship with alpha rhythm

A patient in whom a variety of abnormal EEG findings can be elicited by elimination of central vision and fixation demonstrates fixation-off sensitivity. The underlying mechanisms of fixation-off sensitivity and its relationship with alpha rhythm remain unclear. To obtain a better understanding of this issue, we used a whole-head magnetoencephalograph to study an epileptic child with fixation-off sensitivity resulting in a 3-Hz, large-amplitude oscillation (300 microV) over the occipital regions on the EEG. Magnetic source localization revealed alpha activity around the calcarine fissure and surrounding parieto-occipital areas. Magnetic sources of abnormalities relating to fixation-off sensitivity, however, usually were located deeper in the brain, suggesting more extensively distributed sources, with involvement of the cingulate gyrus and the basomesial occipitotemporal region. Distributions of the sources of both types of activities show independent clusters but also an appreciable domain of overlap. Our findings indicate that abnormalities related to fixation-off sensitivity can emerge in thalamocortical networks, with larger and more anterior cortical distribution than those that generate alpha rhythm. Transition in the type of oscillation appears not only to depend on a change in cellular dynamics but also to be reflected in a different spatial distribution of the underlying neuronal networks.

The electric resistivity of human tissues (100 Hz-10 MHz): a meta-analysis of review studies

The electric resistivity of various human tissues has been reported in many studies, but on comparison large differences appear between these studies. The aim of this study was to investigate systematically the resistivities of human tissues as published in review studies (100 Hz-10 MHz). A data set of 103 resistivities for 21 different human tissues was compiled from six review studies. For each kind of tissue the mean and its 95% confidence interval were calculated. Moreover, an analysis of covariance showed that the calculated means were not statistically different for most tissues, namely skeletal (171 omega cm) and cardiac (175 omega cm) muscle, kidney (211 omega cm), liver (342 omega cm), lung (157 omega cm) and spleen (405 omega cm), with bone (> 17,583 omega cm), fat (3,850 omega cm) and, most likely, the stratum corneum of the skin having higher resistivities. The insignificance of differences between various tissue means could imply an equality of their resistivities, or, alternatively, could be the result of the large confidence intervals which obscured real existing differences. In either case, however, the large 95% confidence intervals reflected large uncertainties in our knowledge of resistivities of human tissues. Applications based on these resistivities in bioimpedance methods, EEG and EKG, should be developed and evaluated with these uncertainties in mind.

Automatic registration of pelvic computed tomography data and magnetic resonance scans including a full circle method for quantitative accuracy evaluation

The purpose of this study is to develop a method for registration of CT and MR scans of the pelvis with minimal user interaction and to obtain a means for objective quantification of the registration accuracy of clinical data without markers. CT scans were registered with proton density MR scans using chamfer matching on automatically segmented bone. A fixed threshold was used to segment CT, while morphological filters were used to segment MR. The method was tested with transverse and coronal MR scans of 18 patients and sagittal MR scans of 8 patients. The registration accuracy was estimated by comparing (triangulating) registrations of a single CT scan with MR in different orientations in a "full circle." For example, CT is first matched on transverse MR, next transverse MR is matched independently on coronal MR, and finally coronal MR is matched independently on CT. The product of the three transformations is the identity if all matching steps are perfect. Deviations from identity occur both due to random errors and due to some types of systematic errors. MR was registered on MR (to close the "circle") by minimization of rms voxel value differences. CT-MR registration takes about 1 min, including user interaction. The random error for CT-MR registration with transverse or coronal MR was 0.5 mm in translation and 0.4 degree in rotation (standard deviation) for each axis. A systematic registration error of about 1 mm was demonstrated along the MR frequency encoding direction, which is attributed to the chemical shift. In conclusion, the presented algorithm efficiently and accurately registers pelvic CT and MR scans on bone. The "full circle" method provides an estimate of the registration accuracy on clinical data.

Registration of MR and SPECT without using external fiducial markers

The aim of our work is to present, test and validate an automated registration method used for matching brain SPECT scans with corresponding MR scans. The method was applied on a data set consisting of ten brain IDEX SPECT scans and ten T1- and T2-weighted MR scans of the same subjects. Of two subjects a CT scan was also made. (Semi-) automated algorithms were used to extract the brain from the MR, CT and SPECT images. Next, a surface registration technique called chamfer matching was used to match the segmented brains. A perturbation study was performed to determine the sensitivity of the matching results to the choice of the starting values. Furthermore, the SPECT segmentation threshold was varied to study its effect on the resulting parameters and a comparison between the use of MR T1- and T2-weighted images was made. Finally, the two sets of CT scans were used to estimate the accuracy by matching MR to CT and comparing the MR-SPECT match to the SPECT-CT match. The perturbation study showed that for initial perturbations up to 6 cm the algorithm fails in less than 4% of the cases. A variation of the SPECT segmentation threshold over a realistic range (25%) caused an average variation in the optimal match of 0.28 cm vector length. When T2 is used instead of T1 the stability of the algorithm is comparable but the results are less realistic due the large deformations. Finally, a comparison of the direct SPECT-MR match and the indirect match with CT as intermediate yields a discrepancy of 0.4 cm vector length. We conclude that the accuracy of our automatic matching algorithm for SPECT and MR, in which no external markers were used, is comparable to the accuracies reported in the literature for non-automatic methods or methods based on external markers. The proposed method is efficient and insensitive to small variations in SPECT segmentation.

Demonstration of a reduction in muscarinic receptor binding in early Alzheimer's disease using iodine-123 dexetimide single-photon emission tomography

Decreased muscarinic receptor binding has been suggested in single-photon emission tomography (SPET) studies of Alzheimer's disease. However, it remains unclear whether these changes are present in mildly demented patients, and the role of cortical atrophy in receptor binding assessment has not been investigated. We studied muscarinic receptor binding normalized to neostriatum with SPET using [123I]4-iododexetimide in five mildly affected patients with probable Alzheimer's disease and in five age-matched control subjects. Region of interest (ROI) analysis was performed in a consensus procedure blind to clinical diagnosis using matched magnetic resonance (MRI) images. Cortical atrophy was assessed by calculating percentages of cerebrospinal fluid in each ROI. An observer study with three observers was conducted to validate this method. Alzheimer patients showed statistically significantly less [123I]4-iododexetimide binding in left temporal and right temporo-parietal cortex compared with controls, independent of age, sex and cortical atrophy. Mean intra-observer variability was 3.6% and inter-observer results showed consistent differences in [123I]4-iododexetimide binding between observers. However, differences between patients and controls were comparable among observers and statistically significant in the same regions as in the consensus procedure. Using an MRI-SPET matching technique, we conclude that [123I]4-iododexetimide binding is reduced in patients with mild probable Alzheimer's disease in areas of temporal and temporo-parietal cortex.

The computation of MR image distortions caused by tissue susceptibility using the boundary element method

Static field inhomogeneity in magnetic resonance (MR) imaging produces geometrical distortions which restrict the clinical applicability of MR images, e.g., for planning of precision radiotherapy. The authors describe a method to compute distortions which are caused by the difference in magnetic susceptibility between the scanned object and the surrounding air. Such a method is useful for understanding how the distortions depend on the object geometry, and for correcting for geometrical distortions, and thereby improving MR/CT registration algorithms. The geometric distortions in MR can be directly computed from the magnetic field inhomogeneity and the applied gradients. The boundary value problem of computing the magnetic field inhomogeneity caused by susceptibility differences is analyzed. It is shown that the boundary element method (BEM) has several advantages over previously applied methods to compute the magnetic field. Starting from the BEM and the assumption that the susceptibilities are very small (typically O(10(-5)) or less), a formula is derived to compute the magnetic field directly, without the need to solve a large system of equations. The method is computationally very efficient when the magnetic field is needed at a limited number of points, e.g., to compute geometrical distortions of a set of markers or a single surface. In addition to its computational advantage the method proves to be efficient to correct for the lack of data outside the scan which normally causes large artifacts in the computed magnetic field. These artifacts can be reduced by assuming that at the scan boundary the object extends to infinity in the form of a generalized cylinder. With the adaptation of the BEM this assumption is equivalent to simply omitting the scan boundary from the computations. To the authors' knowledge, no such simple correction method exists for other computation methods. The accuracy of the algorithm was tested by comparing the BEM solution with the analytical solution for a sphere. When the applied homogeneous field is 1.5 T the agreement between both methods was within 0.11.10(-6) T. As an example, the method was applied to compute the displacement vector field of the surface of a human head, derived from an MR imaging data set. This example demonstrates that the distortions can be as large as 3 mm for points just outside the head when a gradient strength of 3 mT/m is used. It was also observed that distortion within the head can be described accurately as a linear scaling in the axial direction.

A fast method to compute the potential in the multisphere model

A series expansion is derived for the potential distribution, caused by a dipole source in a multilayered sphere with piecewise constant conductivity. When the radial coordinate of the source approaches the radial coordinate of the field point the spherical harmonics expansion converges only very slowly. It is shown how the convergence can be improved by first calculating an asymptotic approximation of the potential and using the so-called addition-subtraction method. Since the asymptotic solution is an approximation of the true solution, it gives some insight on the dependence of the potential on the conductivities. The formulas will be given in Cartesian coordinates, so that difficulties with coordinate transformations are avoided. Attention will be paid to the (fast) computation of the partial derivatives of the potential, which is useful for inverse algorithms.

A linear discretization of the volume conductor boundary integral equation using analytically integrated elements

A method is presented to compute the potential distribution on the surface of a homogeneous isolated conductor of arbitrary shape. The method is based on an approximation of a boundary integral equation as a set linear algebraic equations. The potential is described as a piecewise linear or quadratic function. The matrix elements of the discretized equation are expressed as analytical formulas.

A random dipole model for spontaneous brain activity

The statistical properties of the EEG and the MEG are described mathematically as the result of randomly distributed dipoles. These dipoles represent the interactions of cortical neurons. For certain dipole distributions, the first- and second-order moments of the electric and magnetic fields are derived analytically. If the dipoles are in a spherical volume conductor and have no preference for any direction, the variance of a differentially measured EEG-signal is only a function of the electrode distance. In this paper, the theoretically derived variance function will be compared with EEG- and MEG-measurements. It is shown that a dipole with a fixed position and a randomly fluctuating amplitude is an adequate model for the alpha-rhythm. An expression for the covariance between the magnetic field and a differentially measured EEG-signal is derived. This covariance is considered as a function of the magnetometer position, and is compared with the measurements of Chapman et al. [23]. The theory can be used to obtain a (spatial) covariance matrix of the background noise, which occurs in evoked potential measurements. Such a covariance matrix can be used to obtain a maximum likelihood estimator of the dipole parameters in evoked potential studies, to evaluate the merits of the so-called "Laplacian derivation," and for the interpolation of electromagnetic data.

The use of the asymptotic expansion to speed up the computation of a series of spherical harmonics

When a function is expressed as an infinite series of spherical harmonics the convergence can be accelerated by subtracting its asymptotic expansion and adding it in analytically closed form. In the present article this technique is applied to two biophysical cases: to the potential distribution in a spherically symmetric volume conductor and to the covariance matrix of biomagnetic measurements.