Structure

Source Localization of Brain Pathological Activity by Means of Magnetic Encephalography


Initial data for analysis were obtained by means of 148-channel magnetometer equipped with superconducting coils (SQUIDs) in the Medical school of New-York University.

The main purpose of this work is in detecting and localization of the pathological signal sources in the cases of wide-spread diseases (Parkinsonism and its variations). Measured signal represents space-temporary structure: 148-dimension vector of measurements in 148 points on the head surface, which is measured with the frequency 500 Hz. The whole work on the analysis of obtained signal may be divided into following stages:

  1. Extracting of the useful signal - that reveals the proper type of brain activity (e.g., evoked signal, caused by periodic stimulus - auditory, visual, tactile etc; the signal connected with the tremor generation or auditory hallucinations in parkinsonian cases).
  2. The moments of the time choice for which an inverse problem of current source localization will be solved.
  3. An inverse problem solving to localize the sources for pathological or normal activity. Extracted signal values in chosen time instants are applied as initial data for analysis. To precise results NMR-tomogram of the patient is used to take into account specific physiological restrictions.

Here demonstration clip is presented. It illustrates functioning of computer system for brain space-time activity analysis. Algorithms and software were developed in the Institute of Mathematical Problems of Biology Russian Academy of Sciences. An experiment with the auditory stimulation by the 7 Hz signal is shown. It is well known that auditory signal leads to excitation of auditory cortex, placed in temporal parts of the brain. Areas of excitation are sufficiently compact that brings to sources simulation as current dipoles. Because the separate area of excitation exists for each ear, we have a case with two current dipoles.

An inverse problem solution is presented for each time instant by means of two current dipoles with variable moments which are localized main part of time in the auditory cortex. Once in a while sources, optimally approximating the field on the head surface, fall into deeper area of brain. This phenomenon probably is explained by the episodic thalamic activity.

In the left part of the screen three mutually perpendicular NMR-tomogram sections are presented. All plane sections pass through one of the sources. In the fourth window section, perpendicular to the source radius-vector is imaged. Circles in the same windows points dipoles locations and their colors describe dipoles moments magnitude, changing from maximum (white circle) to zero (black circle). The lines coming out of the circles indicate a direction of dipole moment.

In the right part of the screen the picture of magnetic field distribution over head within measuring helmet is presented. Different colors correspond to different values of the field, herewith positive values of magnetic induction flux (field lines cross crane outward) and negative values correspond red and blue parts of the spectrum. Experimental data are presented on the upper cadre and on the lower we have calculated field from two localized sources.

Multisequencing of computing procedures is realized on the separate measuring channels (from several parallel processes up to 128). Here the most informative channels are chosen of 148 initial time series. Computing difficulty of the problem is conditioned by the great arrays volumes (for separate experiment about 350 эт) and by complexity of optimization procedure realization in multidimensional parameters space. Using the powerful transputers is the single possibility to solve the problem in the complex cases with 3D-distributed current sources.

Presented clip was demonstrated on the Annual Industrial Fair Hannover Messe-2001 (23-28 April 2001, Hanover, Germany); it is included in the catalogue of fair section IT/Software, subsections: "Complete solutions for continuous computer simulation/ process simulation/ computational fluid dynamics" and "Complete solutions for discrete computer simulation".

Calculations were completed on the computers of Joint Supercomputer Center and Center of Bioinformatics.

The research presented was made posible in part by Award No. RB1-2027 of US Civilian Research and Development Foundation for the Independent States of the former Soviet Union (CRDF). Any opinions, findings and conclusions or recomendations expressed in this matirial are those of the authors and do not necessary reflect those of the CRDF.