《Theoretical evaluation of GSM UMTS electromagnetic fields on neuronal network response.pdf》由会员分享，可在线阅读，更多相关《Theoretical evaluation of GSM UMTS electromagnetic fields on neuronal network response.pdf（7页珍藏版）》请在三一文库上搜索。
1、IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 50, NO. 12, DECEMBER 20023029 Theoretical Evaluation of GSM/UMTS Electromagnetic Fields on Neuronal Network Response Francesca Apollonio, Micaela Liberti, and Guglielmo DInzeo, Member, IEEE AbstractUnderstanding the modalities of interaction
2、 of electromagnetic (EM) fields with biological material is a key point in the identification of possible induced effects. An integrated approach to model EM fields interaction with biological systems is proposed in this paper. Here, a neuronal network is identified as the biological target. In this
3、 paper, for the first time, the possible effects of EM signals related to universal mobile telecommunica- tion system and global system for mobile communication standard on the setup model are investigated. IndexTermsBioelectromagneticinteraction,microwaves (MWs), mobile phones, modeling, neuronal n
4、etwork. I. INTRODUCTION T HE study of possible electromagnetic (EM) field effects on biological systems can be rigorously faced by a preliminary investigation on biophysical mechanistic basis of the interaction between the fields themselves and the biological structures involved. Since the beginning
5、s of bioelectromagnetic studies, cellular membrane has been addressed as a primary site ofinteraction,leadingtodifferentmodelsinliterature.Inpartic- ular, authors have chosen to perform analysis of this biological system as the only effective way to understand interactions with EM fields, in order t
6、o explain effects at cellular and tissue level 1, 2. This approach is quite in accordance with a basic observation: the biophysical and biochemical physiological equilibriums are managed at cell and cell-membrane level. However, the cell membrane is not the basic biological unit for a bio-system, in
7、 fact, some other elementary structures exist with defined tasks and functional modalities. This leads to the determination of a biological scale of complexity that grows from the low biophysical level of ion-transport across cell membrane to the biological one of cellular cycles or signaling pathwa
8、ys. The structures and processes at each level of this scale, due to their electrical or polar nature, are identified as intrinsically sensitive to EM fields. Apollonio et al. recently proposed an integrated approach to model EM fields interaction with biological systems 2. This methodologyimplement
9、sthebiologicalscaleofcomplexityand evaluates the effects induced by the EM field on each compo- Manuscript received April 6, 2002; revised August 29, 2002. This work was supported in part by the Ministry of University, Technology and Scientific Re- search(AdPENEA-CRN/MURST)andbytheEuropeanCommunity,
10、VFrame- Work under the RAMP2001 Project. The authors are with the Center of Electromagnetic Fields and Biosystems and the Department of Electronic Engineering, University of Rome, “La Sapienza,” 00184 Rome, Italy (e-mail: dinzeouniroma1.it). Digital Object Identifier 10.1109/TMTT.2002.805287 nentoft
11、hemodeluptotheneuronalnetworklevel.Inthispaper, the authors investigate the possible effects of EM signals re- lated to universal mobile telecommunication system (UMTS) and global system for mobile communication (GSM) standard on the setup model. In the next two years, the new mobile telecommunicati
12、on standard UMTS is going to establish itself as the third-genera- tion technology, and services based on this standard will coexist with the current use of GSM second-generation technology. These standards greatly differ in frequencies and patterns used time-division multiple access (TDMA) versus c
13、ode-division multiple access (CDMA), therefore, it seems interesting to evaluate how RF fields associated with such wireless technology can interact with biological systems and how the differences in the signal physical layer could eventually modify physiological conditions. II. MATERIALS ANDMETHODS
14、 A. Neuronal Network Model A quasi-realistic neural network has been used in order to investigate possible modifications in the electrical responses under EM exposure. We mainly consider the pattern of action potentials and back-propagating action potentials, in particular, inter-spike intervals (IS
15、Is), spike coincidence, and synchroniza- tion of firing neurons, as macroscopic observables of possible metabolic changes within the neuron. The task of connecting neuronal models has been resolved in 2, approaching the problem of signal propagation inside the axon on the basis of the core-conductor
16、 model 3. The axon lengthcanvaryinarangebetween1100mm.Theoveralleffect of the neuronal axon can be taken into account by considering a pre-axon transmembrane voltageand a post-axon one , linked by the following relationship: (1) whereis the axon length, andis determined referring to the resistiveand
17、 dielectric characteristics of the axon. More specifi- cally,referringtoFig.1(a),itispossibletoapproximatetheaxon with a cylinder and, ifis the core resistance per unit length, is the resistance of extracellular fluid per unit length, and is the resistance across a unit length of passive membrane 3,
18、 it is possible to interconnect neuronal cells one another. The model for the neuronal cell has to take into account the characteristicdoublelayerofthecellularmembraneresponsible 0018-9480/02$17.00 2002 IEEE 3030IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 50, NO. 12, DECEMBER 2002 Fig
19、. 1.Interconnection of neuronal models. (a) Network connection of two neurons. (b) Stochastic neuron model. for the insulation of the cytoplasm from the external environ- ment, as well as several ionic channels, i.e., macromolecular structures through which ions can pass from the extracellular mediu
20、m to the intracellular and vice versa. The electrical inter- pretation of the model gives rise to a lumped-element circuit for the neuron (the deterministic neuron) in which the double layer isrepresentedbymeansofacapacitance,andionicchannelscan be simulated by means of nonlinear resistances (ions m
21、oving through the membrane) and fixed potentials (different ion con- centration inside and outside the cell) 4. Results obtained with the lumped model give a good matching with experimental data 5.Connectingseverallumped-elementmodelsgivesrisetothe deterministic neuronal network. The response of the
22、 determin- istic neuronal network to extremely low frequency (ELF) EM fieldshasbeeninvestigated in2usingthelumped-elementcir- cuit able to simulate the behavior of silent and firing neuronal membranes. Nevertheless, such a model is not properly usable at RF frequencies due to the presence of the sho
23、rt-circuiting ca- pacitance of the membrane. In order to overcome this limit, a different way to model ionic channels has been introduced based on a Markov model approach 68. The channels behavior is fully defined by a set ofstates, a transition rate matrix, whose elements are the transition rates r
24、egulating the kinetic of the process, and a vector, whoseelements are the occupancies for each state at the starting time. Each state in the model represents a possible conformation for the channel (open or closed), and transitions among states represent structural modifications associated to energe
25、tic changes. If the channel is considered ohmic (experimentations confirm this hypothesis with reasonable approximation 9), the current flux through the channel at any instant is proportional to the probability for the system of being in an open state. A scheme reporting Fig. 2.State machine model f
26、or sodium voltage-dependent channel. Model topology: eight states, two different kinetics, transition rates. the different states and the matrixelements are reported in Fig. 2 for sodium, a voltage-dependent ionic channel used in the following. The channel is composed of four molecular subunits, who
27、se structural conformations are responsible for channel opening or closing. As a consequence, the model has eight states with only one considered “open,” which means conductive. Three subunits follow an-like kinetics regulating fluctuations between openand closedstates, while one subunit follow an-l
28、ike kinetics for the inactivating process. The transition rates depend on membrane voltage and on temperature. In order to evaluate the time course of channel gating, the channel is commonly supposed to be a zeroth-order Markov chain, stationary and ergodic, and the channel activity is rep- resented
29、 by a random process where the aleatory variable is the dwell time in a certain state 10. Under such hypotheses, it is possible to quantify the current flowing through the channel, as a function of the time , by the evaluation of the probability of finding the system in the open state 1012. The mode
30、l has been implemented and solved by means of a Monte Carlo tech- nique, following the procedure summarized below: 1) single process simulation identification of the current state of the channel; evaluationofthedwell-timeinthecurrentstate by stochastic technique 10; evaluation of the state that will
31、 be occupied after a time; 2) statistical averaging realization ofsingle-process simulation; evaluation of the open probability temporal evolu- tion. Once the dwell time in a certain statehas been determined asafunctionoftransmembranevoltageandtemperature,itisas- sumed that the channel is insensitiv
32、e to an external stimulation for a time, after which, returning to be sensitive, is ready for the evaluation of the next state. For this reason, providing that an appropriate time step is used in the numerical algorithm, the APOLLONIO et al.: THEORETICAL EVALUATION OF GSM/UMTS EM FIELDS ON NEURONAL
33、NETWORK RESPONSE3031 Fig. 3.Membrane voltage behavior. The parameters observed are ISIs, the distance between two subsequent spikes, and?membrane threshold, membrane value at which the spike starts. Fig.4.IoniccurrentscalculatedviaMarkovmodelswithadifferentnumberof realizations?.(a)Sodiumcurrentwith
34、? ? ?and? ? ?realizations. (b) Calcium current with? ? ?and? ? ?realizations. channel can be investigated even with high-frequency compo- nents. This explains how EM exposure even in the microwave (MW) range can be considered. Ionic channels simulated through stochastic state machines havebeenintrod
35、ucedinthecircuitalmodelforasingleneuronin 2. Fig. 3 reports on the behavior of membrane voltagein phys- iological conditions. The integrated model obtained (stochastic TABLE I ISIAND?OF THESTOCHASTICNEURON FORDIFFERENTNUMBER OF REALIZATIONS a b Fig. 5.Example of UMTS and GSM signals used for simulat
36、ions. (a)?is 0.546 ms (carrier on) and?is 4.6 ms. (b)?is 0.66 ms, in each timeslot? power varies randomly of 1 dB and frequency varies of 5 MHz over 1940 MHz. neuron) is composed by cascading protein channel models, as sketched in Fig. 1(b), and can respond to RF fields like mobile telecommunication
37、 signals. In this paper, the contribution of activated synapses to link several stochastic neurons has been taken into account, following the core-conductor approach previously introduced. The final result is a tissue network: the stochastic neuronal network. The more the ionic currents simulated wi
38、th Markov models are similar to the currents generated in the deterministic model, the more the entire neuronal network works fine, thus giving realistic data. To this regard, a crucial point is played by the statistical averaging process of the ionic current Monte Carlo solution, in particular, the
39、 number of realizationof a single gating process. Fig. 4 reports on the sodium and calcium currents simulated in the stochastic neuron through Markov models with different values of. It is possible to observe that increasingfrom 1024 to 2048 realizations determines a reduction of “noise” in the curr
40、ent waveforms, particularly for the sodium current and, hence, a better approximation of the analytical curve. The resulting statistical parameters (mean value and standard deviation) of the ISI and threshold voltage are reported in Table I for bothand realizations. The value ofhas been chosen for t
41、his study. In order to assess the connection, three neurons have 3032IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 50, NO. 12, DECEMBER 2002 TABLE II OUTPUTVARIABLES FOR THEDIFFERENTMODELS been considered and placed at a distance of 1 mm. Interest on stochastic models of nervous structu
42、res increased recently 13 and a complete simulation of the involved channels seems to be crucial 13, 14. Such results 13 open the possibility of applying our approach to other kinds of nervous tissues. Once defined, the physiological situation of a neuronal tissue with its own pattern of receptors a
43、ndsignaling transduction sys- tems, the response of the network to high-frequency fields like UMTS and GSM signals have been investigated, comparing the results with a continuous wave (CW) signal. B. EM Coupling When dealing with a cell exposed to an EM field, in good ap- proximation, the electric f
44、ield value present on the membrane can be related to the transmembrane voltage, as shown in 15. Following this assumption, the cell has been considered as a sphere of radiusof 40m covered by a membrane of very small thickness and with a capacitanceF/m , im- plyingthatthefrequenciesofinterestforthiss
45、tudy,i.e.,theMW range, are above therelaxation frequency for the cell. At these frequencies, the membrane potential can be calcu- lated approximately as follows: (2) whereandare, respectively, the conductivity of the cyto- plasm, equal to 1 S/m, and the conductivity of the extracellular fluid of 2 S
46、/m 2, 16. As stated previously, ionic channels coefficientsand depend on the transmembrane voltage, which can be considered as the sum of two contributions: the first related to the physiological membrane values and the second related to a perturbing component due to the EM field acting on the cell
47、16. Therefore, the EM field can be thought of as a perturbation of the equilibrium state that modifies ionic currents. In this context, ionic channels, as the elementary biological units, state the sensitivity of the model to the external perturbation. In order to establish which value of the membra
48、ne potential due to the EM field can be considered significant, a statistical methodology has been considered. The T-Student test,1commonly used to test whether the mean (or median) of a physical variable differs between two population groups, has been applied to the mean open probability of ionic c
49、hannels, 1GraphPadSoftwareInc.,SanDiego,CA.19921998.Online.Available: http:/ calculated with Markov models. The T-Student test gives as a result the-value; ifis large, there is no evidence of a difference comparing the two different groups. In particular, considering as an example the potassium channel, the physiological and exposed situation have been compared for different values of induced membrane potential. Extremely significant values ofhave been obtained for a 20V of EM-induced volta