A biomimetic active electrolocation sensor for detection ... - IEEE Xplore

6 downloads 0 Views 2MB Size Report
Abstract—Weakly electric fish sense their surroundings in complete darkness by active electrolocation. In a biomimetic approach we designed catheter-based ...
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

1

> Sensors-4423-201.R1
Sensors-4423-201.R1 < the entire body surface of the fish [5]. In G. petersii, a typical electric image has a centre-surround (or Mexican hat) spatial profile (Fig. 1). A good conductor projects an image with a center region where the local EOD amplitude is increased, surrounded by a rim area where the amplitude decreases compared to the amplitude in absence of an object. Images of non-conductors are of opposite arrangement: local EOD amplitude decreases centrally and slightly increases in the surrounding rim area [6]. Electric images are always blurred, or “out of focus”, since no focusing mechanisms comparable to the lens of an eye exist. Hence when projecting 3-D objects on the 2-D sensory surface, there is no one-to-one relationship between spatial object properties and image shape (Fig. 1). In addition to shape, size, and geometrical properties of the object, electric images depend on object distance and location along the fish’s body, the fish’s body proportions, bending movements of the fish’s body, the presence of additional objects, the background and many more [7]. During active electrolocation weakly electric fish detect and quantitatively analyze several object properties based on the electric images. Amongst these is the electrical resistance of objects determined by measuring the amplitude change imposed on the local EOD. Mormyrids can also perceive capacitive object properties (‘capacitance detection’). In addition, they can localize objects in 3-dimensional space and thus have a true sense of depth perception [8]. Even though the electric sense lacks focusing mechanisms, weakly electric fish also perceive an object’s three-dimensional shape [9]. As mentioned above, the electric images of objects located close to each other near the fish fuse in a nonlinear way leading to complex electric images. In spite of this effect, Mormyrids are able to perceive the shape of an object even when it is positioned right in front of a large background. This ability even persists if the object and the background are made from the same material, e.g. a metal object in front of a metal background. These findings show that weakly electric fish have a remarkable ability to analyze complex 3-dimensional scenes of objects and can identify single object’s properties even in a natural setting containing many objects of various sizes, shapes and material [10]. Inspired by the remarkably capabilities of weakly electric fishes in detecting, recognizing and analyzing objects, we designed technical sensor systems that can solve similar problems of remote object sensing. We applied the principles of active electrolocation to technical systems by building devices that like the fish produce electrical signals in a conducting medium (water or blood) and simultaneously sense local voltages. Here we report results about catheter based medical sensors, which are intended to be used for the detection and analysis of pathological changes of the walls of blood vessels. One of the major causes of heart attack is the rupture of so called vulnerable plaques of coronary arteries. These wall inclusions have a diameter of between 1 and 22 mm [11] and a mean volume of 8.38 ± 6.72 mm³ [12, 13]. They consist of a lipid filled core covered by a thin membrane

[14]. When this tenuous cap bursts, the lipid content gets in contact with the blood, and finally causes a thrombosis leading to heart attack. Diagnostic detection of vulnerable plaques and discrimination from less dangerous stable plaques therefore is a major challenge for medical diagnostics. We try to meet this challenge by designing a biologically inspired, catheter based sensor system, which uses active electrolocation for plaque identification. Here we report first experimental results that test in principal whether plaque analysis is possible using an active electrolocation method inspired by weakly electric fish. II. PLAQUE DETECTION IN ARTIFICIAL BLOOD VESSELS The principle of active electrolocation was adapted for a new catheter-based method to detect and distinguish coronary atherosclerotic plaques. The measuring setup consisted of a commercial four electrode catheter (Finder 6F; Dr. Osypka GmbH; Woxx 6F; BIOTRONIK GmbH & Co. KG), being connected to a special designed LabVIEW®-program. The software was used to generate a pre-recorded EOD signal (5V) displayed by the first electrode of the catheter (sender). The EOD signal was a biphasic pulse similar to a singleperiod sine wave signal and had a duration of 0.5 ms. It was received by the three remaining electrodes (E1-E3) and the peak-to-peak amplitude of the signal was calculated by the program (see also figure 9).

Fig. 2. Modulation of the peak-to-peak amplitude measured by the 3 electrodes (E1-E3) of the catheter for a small plaque (P1, 1 mm³, black lines) and a large plaque (P5, 268 mm³, grey lines). The first negative peak (black arrows) occurred, when the sender was beneath the plaque, while the second peak (white arrows) occurred when the receiver was beneath the plaque. Depending on the distance between the sender and the receiving electrode, the electric images had different expansions. SD = standard deviation.

The measurements were performed in artificial blood vessels containing artificial plaques of different sizes (P1 to P5). As artificial vessels, we used agarose tubes with a lumendiameter of 5 mm and a wall-diameter of 2 mm. The conductivity of the vessel wall was set to 0.57 S*m-1, which is similar to a natural blood vessel [15]. The artificial plaques consisted of cores of different volumes (1, 5, 33.51, 113.51 and 268.08 mm³) and diameters (1.87, 3.78, 4, 6, 8 mm).

Copyright (c) 2010 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

3

> Sensors-4423-201.R1 < During measurements, the tube was fixed on a polystyrenebase in an experimental tank filled with a special solution of a conductivity (0.7 S*m-1) resembling that of blood [15]. The catheter was moved stepwise through the vessel, measuring an average of 30 EODs at each position. Based on the signal traces of each electrode (E1-E3), the electric image was calculated as the modulation of the object-related voltage drop in relation to expectancy values of the undistorted signal. The dimensionless modulation values were calculated by dividing the measured voltage in the presence of the plaque by the expected voltage without the plaque at each position of the catheter (Fig. 2). These images showed different profiles depending on two parameters: the distance between sender and receiver electrodes (E1-E3) and the volume of the plaque. A small plaque (Fig. 2, black lines) produced smaller modulations and narrower electric images than a large plaque (Fig. 2, grey lines). Image widths also increased with larger electrode distances (Figs. 2, 3).

arrows) emerged when the receiver (E1-E3) was beneath the target. Both peaks are clearly related to the target, allowing a reliable detection of the artificial plaque. The absolute value of the maximum negative peak in the centre of the electric image is equal to the maximum voltage change in the obtained signal, caused by the presence of the plaque. This parameter is referred to as the peak-amplitude of the electric image. Both image-parameters (width and amplitude) were used to differentiate between the different plaque-types (P1-P5). All of the tested plaque types could be detected by calculating the modulation of the peak-to-peak amplitude caused by the different object volumes. However, the plaques could be significantly separated by their mean calculated image widths up to a certain plaque-volume depending on the distance between the sender and the receiver (Fig. 3). By fitting the image widths of the five measured plaques with the Boltzmann equation and setting the threshold to 12.5% below the asymptote, we could determine the upper linear detection range for each measuring electrode (E1-E3). The limits were 32.62 mm³ for the first electrode (E1), 57 mm³ for the second electrode (E2) and 59.25 mm³ for the third electrode (E3) as illustrated in Fig. 3, left side by the vertical dashed lines. III. PHYSICAL MEASUREMENTS A. Experimental setup In addition to the biological inspired measurements with a manually operated catheter in an artificial vessel with a fat target, measurements were performed with a physical setup. The objective of these experiments was to find out about the influence of uneven movements of the catheter. In addition, we wanted to measure how different target properties, such as size, shape and physical properties, effect the measurements.

Fig. 3. Boltzmann fit of the calculated image widths caused by different plaque volumes (left side) and comparison of the mean image widths of the tested plaque types (right side, Anova: level of significance: *≤0.05; **≤0.01; *** Sensors-4423-201.R1 < target was moved automatically with stepping motors in two dimensions (X- and Y-positions) at a fixed distance D along the catheter (Figs. 4, 5). The catheter was a commercial available design by VascoMed (VascoMed, Institut für Kathetertechnolgie GmbH, Weil am Rhein, Germany; Type VascoStim 6/4F S) with 6 platinum-iridium (Pt-90%, Ir-10%) electrodes and with a diameter of about 1.3 mm (see figure 5). The targets used were cubes of aluminum and plastic with side lengths of 14 mm. The water had a conductivity of 18.2 µS*cm-1 at room temperature. The target was attached to the moving unit by a thin plastic rod of 4 mm diameter. The moving unit consisted of three stepping motors and controllers from Physik Instrumente (PI). For the measurements, the target was moved in steps parallel to the catheter in a plane with a fixed distance D, see figure 5. The step size of the movement was 1 mm and the distance between two measurement lines in the measurement plane was 0.7 mm. The target movement and the signal acquisition were controlled by LabVIEW® (release 2010, National Instruments, Austin, Texas, USA). We measured the change in the amplitude of the electric field created by the catheter’s electrodes due to the target. For measuring, the electrode at the tip of the catheter was used as input (sender), and the next electrode 6.7 mm apart was used as the sensing electrode. A lock-in amplifier SR830 (Stanford Research Systems, Sunnyvale, California, USA) was used to produce the input signal (sine wave of 1V at 1 KHz) and as read-out for the sensing electrode. B. Measurements In order to demonstrate the influence of a conducting respectively a non-conducting target and the measuring distance, 2-dimensional measurements were performed with the aluminum and the plastic cube positioned at a distance D (1 mm and 4 mm) from the measurement plane to the catheter (see figure 5). The resulting field distribution in the measurement plane can be interpreted in principle as a distribution that would be measured with a stationary target and an array of point electrodes distributed over the surface of the catheter.

Fig. 5. Top view of the catheter with its dimensions (above) and scheme of the movement of the target along the catheter (below). D is the distance between the object and the catheter, which was either 1 or 4 mm.

4 Figure 6 shows the results of these measurements. It can be seen that the field distributions due to the metal and the plastic target have opposite appearances. This can be explained by the different distribution of the field lines: with a metal target, no electrical field is present inside the target and the surface of the target is equipotential. With a plastic target, the surface is not equipotential because the potential lines continue inside the target.

Fig. 6. The change in amplitude of the electric field caused by an aluminum cube (14 mm side length) with a measurement distance of (a) 1 mm and (b) 4 mm; and caused by a plastic cube (14 mm side length) with a measurement distance of (c) 1 mm and (d) 4 mm. Scan areas 46 mm [scan width y (78-32)] x 18.2 mm [scan width x (36-10)].

In real measurements of blood vessels, the targets (e.g. plaques) might have no homogeneous material properties. For example, the core of a target could have a different conductivity than the shell of the target. To investigate this effect, we covered the plastic target with a 0.02 mm thick layer of an aluminum sheet. Figure 7 shows a line scan of both targets at a measurement distance of 1 mm and a fixed yposition of 1 mm. It is obvious that the plastic target covered with an aluminum sheet behaved like a complete conductive target. We observed the same principal effect when we covered an aluminum cube with a plastic sheet. Under these conditions, the aluminum cube appeared like a non conductive cube. A similar result was obtained when a bright metallic

Copyright (c) 2010 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

5

> Sensors-4423-201.R1 < copper target was compared with the same copper target, but with a naturally grown thin corrosion layer. The nonconductive corrosion layer changed the electric image of the target from a conductive to a typical non-conductive object image. For the detection of different plaque types it can be expected that the composition of the outer layers of the plaque might influence the measurements. Due to the fact that no secured material properties of the plaque, consisting of a lipid core with more or less calcification and a fibrous cap, are known so far, a simulation of the effect by Finite Elements (FEM) will yield no reliable results. To overcome this problem, in future studies the material properties of the plaque will be measured thoroughly, e. g. by impedance spectroscopy.

E0 =Electric Field Vector (V/mm) a = Radius of target (mm) r = Position vector relative to centre of the target (mm)

ε 1 =Dielectric constant of the medium  ε 2 =Dielectric constant of the target ρ1 =Resistivity of the medium ρ 2 =Resistivity of the target

On the right hand side of the formula the term in the bigger bracket can be called the “electrical contrast factor”, which varies from X = -1/2 for a perfect insulator (with ρ2 >> ρ1 and ε2 ≤ ε1) to X = +1 for a perfect conductor (ρ2 = 0). If the resistivity and the dielectric constant of the target are equal to the surrounding medium, X will be 0 and the target becomes electrically invisible [16].

Fig. 7. Plastic cube with and without the aluminum sheet showing the shielding effect. Measurement distance D is 1 mm; scan width of y-position is constant at 1 mm.

IV. MODELING AND SIMULATIONS A. Theory As mentioned above, an electric fish produces an electric field in the water around its body by activating an electric organ in its tail (current source). If an object with a dielectric constant other than that of water comes in the vicinity the fish, the electric field changes and results in changing input patterns to the electroreceptors of the skin as shown in Fig 1. Using Poisson’s equation the potential induced surrounding the fish can be described as a dipole filed. Solving the equation for spherical objects, the change in potential can be calculated by the following formula: 3

⎛ a ⎞ ⎛ ρ − ρ 2 + iωρ1ρ 2 (ε 2 − ε1 ) ⎞ ⎟⎟ δΦ (r ) = E0 ⋅ r ⎜ ⎟ ⎜⎜ 1 ⎝ r ⎠ ⎝ 2 ρ 2 + ρ1 + iωρ1ρ 2 (2ε1 + ε 2 ) ⎠

Fig. 8. Simulated electrical images generated by a catheter similar to the one shown in Figs. 4 and 5. for conducting and insulating cubes with a side length of 14 mm (see inset). The distances D of both objects were set to 1, 2 or 4 mm. Electric images generated for the conductor and for the insulator show opposite shapes because the electric contrast factor changes from positive to negative. These results are similar to those shown in Fig. 6.

We used the electrical contrast factor as an indicator to distinguish between different kinds of plaques in arterial blood vessels. In these cases, the weakly electric fish is represented by a technical sensor. The medium around the sensor is blood, and the target is either a stenotic plaque or vulnerable plaque with a lipid core [17].

where:

δΦ(r ) =Change in potential at position r (mV)  Copyright (c) 2010 IEEE. Personal use is permitted. For any other purposes, Permission must be obtained from the IEEE by emailing [email protected].

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication.

6

> Sensors-4423-201.R1