cortical sources involved in the spatial working memory

F. Babiloni¹, C. Babiloni^1,3, F. Carducci^1,2, T. Salvatore¹, P. Rossini ^2,3,  C. Del Percio¹, S. Salinari⁴ and F. Cincotti ^5
1Dip. Fisiologia umana e Farmacologia, Univ. "La Sapienza", Rome, ²“AFAR", Ospedale Isola Tiberina,Rome, ITALY, ³ IRCCS FBF “San Giovanni di Dio”, Brescia,  ITALY,  Dipartimento di Informatica e Sistemistica, Univ. “La Sapienza”, Rome, ITALY, ⁵IRCCS Fondazione Santa Lucia, Rome, ITALY

Abstract: In this study we addressed questions related to the cortical sources involved in the spatial working memory in humans. Such sources were estimated from non-invasive EEG recordings by using linear inverse procedure in a group of 9 normal healthy subjects. By using these procedures and head model, we were able to analyzed neural activation in regions of interest (ROIs) of few squared cm at the cortical level. Results showed an active partecipation of cortical prefrontal and parietal areas during the task phase in which an image have to be hold in the memory before an appropriate decision has to be made by the subjects.

INTRODUCTION

Human neuroimaging studies indicate that tasks requiring working memory activate a functional network linking regions of prefrontal cortex with posterior association cortices [1]. Recent studies have indicated that an EEG signal in the theta frequency band (4-7 Hz) largest over midline frontal region of the scalp is enhanced in tasks with greater working memory demands [2]. In contrast to the effect on the frontal midline theta signal, signals in the alpha range (8-12 Hz) are attenuated by increased task demands [2]. The magnitude of the alpha activity during cognitive tasks has been hypothesized to be inversely proportional to the number of cortical neurons recruited into a transient functional network for purposes of task performance [2].

In this study we address questions related to the cortical sources involved in the spatial working memory in humans. In particular, we were interested in which cortical sources are activated during a spatial working memory tasks. Furthermore, the sequences of activation of such cortical sources were also of interest for us. The cortical sources of spatial working memory tasks were estimated from non-invasive EEG recordings by using linear inverse estimation in a group of normal healthy subjects. A quasi-realistic head model, obtained as an average of 152 head models built with the use of magnetic resonance images (MRIs) was used for the source estimation. With this model a easy identification of the principal Broadmann cortical areas was obtained.  By using linear inverse estimation with quasi-realistic head model we were able to analyzed cortical activation at the level of region of interest (ROIs) of few squared cm at the cortical level. Results showed an active partecipation of cortical prefrontal and parietal areas during the task phase in which an image have to be hold in the memory before an appropriate decision has to be made by the subjects..

METHODS

Head models and cortical current estimation

The quasi-realistic head model used in this study was obtained by using the average head model available from the Montreal Neurological Institute. Such model was based on the MRIs head images of 152 normal subjects. The scalp, skull, dura mater and the cortical surfaces were then obtained with a countouring algorithm. Such surfaces were then used to build the Boundary Element Model of the head as volume conductor employed in the present study. Over than 3,000 current dipoles were located along the realistic reconstruction of cerebral cortex. Current density estimation of each dipole moment was obtained by solving weighted minimum norm problem associated with the EEG measures [3] Regularized cortical activity were then averaged within each one of the 44 considered region of interest (ROI) corresponding to the Broadmann areas.

Task and EEG recordings

After a warning stimulus, subjects were exposed to the vision of a cue on the computer screen. Cue visual stimulus consists of a couple of vertical bar that were first presented (trigger time) and then hidden for few seconds (delay time). During this time subjects were asked to hold the images in memory and after a go stimulus appear on the computer screen produce a motor performance in accordance to the image. The clicking of mouse buttons after a go signal represented the motor performance. In particular, the subjects have to push left bottom of mouse if higher left than right vertical bar, or viceversa. A brief on-line feed-back on the performance was automatically provided. Time course of the task was as follows: Working memory condition (WM): Pre-trigger time: duration 1 sec; warning visual stimulus (trigger time): duration 1 sec; cue stimulus: duration 5.5-7.5 sec; visual go stimulus: duration 1 sec; inter-trial interval: duration 5 sec. The control condition in which the spatial bar remains on the screen for all the task duration will be also developed and this was called in the following no working memory condition (NWM). Nine healthy subjects were considered in this study. Each EEG recording was made by using forty scalp electrodes, that were positioned according to an extension of the international 10-20 system. The EEG bands of interest were alpha1 (7-9 Hz) and alpha2 (9-11 Hz). Event Related Desynchronization (ERD) and Event-Related Synchronization (ERS) were computed from the estimated current density waveforms in the alpha1 and alpha2 frequency bands for each one of the ROIs analyzed

Statistical analysis

The obtained results were subjected to separate statistical tests. First of all, after the computation of relevant ERD and ERS features for both WN and NWM tasks, a t-paired test on the mean activation within each considered ROI was performed. Analysis of Variance (ANOVA), in a full within design was also performed to elucidate the neural sources that perform differently in terms of EEG rhythm during the two tasks (WM and NWM). In such ANOVA, the considered Broadmann area is one of the factor considered (factor BROADMANN, with 44 levels), while another factor is the type of task analyzed (factor TASK with 2 levels; WM and NWM). The processed dependent variable was the amplitude of ERD or ERS. The correction of Greenhouse-Gasser for the violation of the spherical hypothesis in all the employed ANOVAs was used. The post hoc analysis was performed with the Scheffe’s test at the p = 0.05 statistical level.

RESULTS

Significant cortical activity was estimated during WM task with respect to the NWM task in all subjects in the analyzed frequency bands. Fig, 2 shows the cortical sites (dark gray) where there are statistical significant decreases of the alpha power (ERD) during the WM task with respect to the NWM task. Such statistical map is relative to the grand average data of nine subjects.

Figure 1. Cortical sites where there is statistical significant decrease of the alpha power during the WM task with respect to the NWM task (dark gray).

The decrease of alpha power was associated to an increased activity of cortical neurons. Note the segmentation of the cortical areas in ROIs, roughly corresponding to the Broadmann cortical areas. In this frequency band (9-11 Hz) there is a strong statistical activation of the fronto-parietal cortical networks during the spatial WM task with respect to the spatial NWM task. This network was active during the period in which the subjects have to take into their memory the information about the greatest between the two bars on the screen. Furthermore, the activation of such network is simultaneous. The ANOVA returns a statistical significant interaction of the factor TASK x BROADMANN in reducing the variance of the ERD computed for all the subjects (p < 0.001). Scheffe’s post-hoc test performed at the 0.05 level of significance demonstrated the significance of the alpha desynchronization during the WM task with respect to the NWM task.

DISCUSSION

In this study the cortical activity has been estimated by using a quasi-realistic head model, linear inverse procedure and non invasive EEG recordings. The head model used allows to us to express brain activity  in terms of Broadmann and Tailarach coordinates, that improves the capability to replicate results between different research groups. Furthermore, it alleviates the need to use individual MRIs of the experimental subjects, which are not always available. The results here obtained are in line with those obtained by neuroimaging fMRI studies, highlighting the fronto-parietal cortical network that was activated during the working memory task. With respect to the previous neuroimaging techniques in literature, it must be noted that the employed EEG technology allowed to us to easily determine the temporal relationship between the activation of fronto-parietal network, with a temporal resolution of less of 1 sec. Such temporal resolution is higher than the temporal resolution available from the raw EEG. This loss of temporal resolution was mainly due to the time period used to compute stable spectral estimates. However, results obtained suggested that the activation of the fronto-parietal network is simultaneous within the time window of 1 second adopted. It is worth of note that such temporal relationship was not possible to be determined with other neuroimaging techniques such as fMRI or PET.

In conclusion, the present study highlights the benefit of the use of linear inverse methodology coupled to the use of quasi-realistic head models for the study of cortical sources of human brain.

REFERENCES

[1] A. Belger, A. Puce, J.H. Krystal, J.C. Gore, P. Goldman-Rakic, G.McCarthy, Dissociation of mnemonic and perceptual processes spatial and nonspatial working memory using fMRI, Hum .Brain Mapp. 6 (1998) 14–32.

 [2] A. Gevins, M.E. Smith, L. McEvoy, D. Yu, High resolution EEG mapping of cortical activation related to working memory: effects of difficulty, type of processing, and practice, Cereb. Cortex 7, (1997) 374–385.

 [3] Babiloni F., Babiloni C., L. Locche, F. Cincotti, P.M. Rossini and F. Carducci, High resolution EEG: source estimates of Laplacian-transformed somatosensory-evoked potentials using a realistic subject head model constructed from magnetic resonance images, Medical & Biological Engineering & Computing, 2000, 38:512-519