Sensor device for electrical impedance tomography imaging, electrical impedance tomography imaging instrument and electrical impedance tomography method

Abstract

A sensor device for EIT imaging comprises an electrode array for measuring an impedance distribution, with at least one sensor for determining spatial orientation of the electrode array coupled to the electrode array. EIT imaging instrument is connectable to a sensor for determining spatial orientation of a test person, and optionally in addition connectable to a sensor for gathering information on electrical and/or acoustic activity and/or a sensor for gathering information on dilation. A computing device is connected or integrated for adjusting impedance data based on spatial data, which spatial data describe the spatial orientation of a test subject. An EIT imaging method for measuring an impedance distribution and adjusting said measured impedance distribution comprises measuring impedance distribution by using an impedance distribution measuring device comprising an electrode array, and transforming the measured impedance distribution into EIT images.

Claims

1. An electrical impedance tomography (EIT) imaging instrument, comprising: an electrode array having a plurality of electrodes for measuring an impedance distribution for investigating lung ventilation and perfusion, the plurality of electrodes of the electrode array being arranged in or on a belt structure defining an observation plane; a first sensor mechanically coupled to the electrode array for measuring spatial data representative of a spatial orientation of the observation plane with respect to a gravity vector simultaneously with the measurement of the impedance distribution; and a computing device for transforming the measured impedance distribution into an EIT image, for providing a temporal correlation of the spatial data and the impedance distribution and for creating an enhanced image by automatically rotating the EIT image with respect to the gravity vector, thereby providing information relating to an orientation of the observation plane to a user.

2. The electrical impedance tomography (EIT) imaging instrument of claim 1, wherein the first sensor for determining the spatial orientation comprises a three-dimensional acceleration sensor.

3. The electrical impedance tomography (EIT) imaging instrument as defined in claim 1, wherein at least one second sensor is coupled to the electrode array for gathering information on at least one of electrical activity, acoustic activity or dilation.

4. The electrical impedance tomography (EIT) imaging instrument as defined in claim 3, wherein the at least one second sensor comprises an electrocardiography sensor.

5. The electrical impedance tomography (EIT) imaging instrument of claim 3, wherein the at least one second sensor comprises a microphone or a phonocardiography sensor.

6. The electrical impedance tomography (EIT) imaging instrument of claim 3, wherein the at least one second sensor comprises a strain gauge.

7. The electrical impedance tomography (EIT) imaging instrument of claim 1, wherein the computing device is additionally configured for adjusting the impedance distribution based on at least one of dilation data, acoustic activity data or electrical activity data.

8. An electrical impedance tomography (EIT) imaging method for measuring and adjusting an impedance distribution, comprising the steps of: measuring an impedance distribution for investigating lung ventilation and perfusion, with an electrode array comprising a plurality of electrodes, the plurality of electrodes of the electrode array being arranged in or on a belt structure defining an observation plane; measuring spatial data describing a spatial orientation of the observation plane with respect to a gravity vector simultaneously with the measuring of the impedance distribution by a first sensor coupled to the electrode array; creating an EIT image; providing a temporal correlation of the spatial data and the impedance distribution; and creating an enhanced image by automatically rotating the EIT image with respect to the gravity vector, thereby providing information relating to the orientation of the observation plane to the user.

9. The EIT imaging method of claim 8, further comprising adjusting the measured impedance distribution according to at least one of: electrical activity; acoustic activity; and dilation of an observed body part within the observation plane.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The invention is hereinafter described with reference to the figures.

(2) FIG. 1 is a test person in random spatial position;

(3) FIG. 2 is a thoracic cut of a test person in three different orientations with respect to the gravity vector, (a) supine position, (b) back leaning position, (c) upright position, and position-dependent zone distribution;

(4) FIG. 3 is a thoracic cut of a test person in two different positions with respect to the gravity vector, (a) full supine position, (b) laterally turned supine position, and position-dependent zone distribution;

(5) FIG. 4 is a definition of u-h axis system and u′-h′ axis system based on the spatial orientation of the observation plane with respect to the gravity vector. Examples of filters are shown on the right-hand side of the picture;

(6) FIG. 5 is a measured and respective filtered strain gauge signal;

(7) FIG. 6 is a scheme of formation of a pixelized impulse image for one single node of a mesh;

(8) FIG. 7 is a scheme of formation of matrix P;

(9) FIG. 8 is a scheme of a filtering process in a spatial frequency domain; and

(10) FIG. 9 is a scheme of formation of matrix R.

DETAILED DESCRIPTION OF THE INVENTION

(11) While this invention is susceptible of embodiments in many different forms, certain embodiments of the invention are shown in the drawings and will herein be described in detail, with the understanding that the present disclosure is to be considered as an exemplification of the principles of the invention and is not intended to limit the broad aspects of the invention to the embodiments illustrated.

(12) As known in the art, EIT data is obtained by a plurality of electrodes that are placed around the chest of a test person 11, e.g. as depicted in FIG. 1. In the present invention, the electrodes may be mounted on a belt-like structure 13 which holds them in a geometrically defined position relative to a body part of a test person 11. In practice an observation plane 15 may be selected by placing a belt-like structure 13 assembled with an array of multiple electrodes around a body part. The exact arrangement of electrodes is not important for the present invention. An arrangement as shown in FIG. 1 is such that the positions of the electrodes define an observation plane 15 (also called electrode plane), for example around the circumference of the chest of a test person 11. In this arrangement the observation plane 15, as e.g. defined by the EIT belt 13, is perpendicular to the main cranio-caudal body axis 17 (anteroposterior axis or saggital axis of the trunk). Angle α (alpha) is defined by the position of the body axis with respect to the gravity vector g. Thus, α is the angle between the main body axis 17 and the gravity vector g. The vector g can be decomposed into two components with one of these components (g.sub.c) lying in a plane that is perpendicular to the body axis, for example the observation plane defined by the EIT belt, as shown in FIG. 1.

(13) A first pair of electrodes is used, as known in the art, to inject current or apply a voltage (excitation signals) to establish an electrical field. The voltages or currents resulting from the application of the excitation signal are subsequently measured at each remaining electrode. The excitation signal is then moved to a next pair of electrodes and the measuring sequence is repeated. In an arrangement of 32 electrodes, for example, 32×32 measurements result per measurement loop. The measurements of each loop are sometimes called “scan frame” or “EIT data vector”.

(14) According to present invention, spatial information on the orientation of a test person is gathered in addition to information about said established electrical field. For example a three-dimensional acceleration sensor (for example a Bosch BMA150 sensor Integrated Circuit) is mechanically coupled with the belt in order to enable the measurement of spatial information and electrical field information simultaneously. The acceleration sensor measures the gravity vector. Implicitly, the angle (i.e. angle α in FIG. 1) between the gravity vector g and the electrode plane 15 is measured since acceleration sensor and belt are mechanically coupled. The regularization, for example by means of a spatial filter F (see example below), is thereafter optimized to give weight to solutions that show homogeneity in the horizontal plane perpendicular to the gravity vector component g.sub.c. The weight, herein referred to as k, is also a function of the angle α of the gravity vector relative to the observation plane and user-defined factors c and d which determine the desired contrast. In one embodiment, the relationship between k and α is a sine function.

(15) Thus the relationship between the angle α, the weight k and the factors c and d can be expressed by the following equation. Hereby α is the angle between the body axis and the gravitation vector g (the definition of α is depicted in FIGS. 1 and 2):
k=c*γ(α)+d,

(16) where γ could be a linear function of α or a sine function of α to allow for smooth transitions.

(17) The acceleration sensor is mechanically connected to the belt-like structure 13 which contains the electrodes. It measures the orientation of the sensor belt relative to the gravity vector g, such as an angle α in space in degrees or radians.

(18) In FIG. 2, thoracic cuts are shown in different orientations with respect to the gravity vector g. Within the thorax 21, the two lungs 23 and 25 are located on the right and left sides. For illustration, where applicable the zones of a 3-zone model (as introduced above) are shown in three different shades in each lung. According to the model the zones extend usually in layers that are perpendicular to the gravitational vector. The first zone 31 is marked in light gray, the second zone 32 in gray, and the third zone 33 in dark gray. According to the model the aeration of the alveoli is reduced in zone 2 and further reduced or even absent in zone 3 due to collapse of alveoli. If the gravitational vector is within (i.e. parallel to) the observation plane, then the zones 1, 2 and 3 are maximally expressed (FIG. 2a). In other words, the three zones are clearly distinguishable if the gravitational vector g is perpendicular to the main body axis 17. FIG. 2a shows this situation by means of the supine position, where the main body axis (shown as black dot in white circle) is perpendicular to the drawing plane. If the body is inclined, the gravitational vector moves outside the observation plane and the zones become much less pronounced (FIG. 2b). In the upright position (FIG. 2c), the angle α is about zero and the zones disappear almost completely. From FIGS. 2a to 2c, the zone expression becomes weaker as the observation plane is inclined further relative to the gravitational vector.

(19) If the gravitational vector g is within the observation plane 15, then the zones of a three-zone model are maximally expressed, as shown in FIG. 2a. The zones remain also expressed—although to a lesser extent—for any body positioning, where the gravitational vector g is outside the observation plane, i.e. where α is different from 90 degrees (FIGS. 2a and 2b), because there is always a component g.sub.c of the gravitational vector g within the observation plane, except when a equals zero (α=0). With decreasing angle α, starting from 90 degrees, the zones become less pronounced and disappear completely at about 0 degree (FIG. 2c). Typically, this relationship is expressed as a sine function. However, other functions may be used instead.

(20) In upright patients (FIG. 2c), the gravitational vector follows the body axis of a person and zones 1 to 3 follow from the tip of the lungs (apex) to the diaphragm. In this position, the zones 1 to 3 will not manifest themselves on the cross-sectional EIT image. If the patient is in supine position, zone 3 is located on the back (dependent region) while zone 1 is located towards the chest (independent region). In supine position (FIG. 2a), the zones 1 to 3 are within the plane of observation and thus impact the EIT images. If a patient is positioned in the prone position, the zones 1 to 3 are still in the plane of observation but in reversed order.

(21) Above depicted body positions include all positions from supine to upright position. However, positions where the body is turned or twisted sideways are not considered yet. Turned body positions may be described by the dorsoventral axis 19 penetrating the plane defined by the main body axis 17 and the gravity vector g. The dorsoventral axis 19 is defined to be perpendicular to the main body axis 17 (as illustrated in FIG. 1). Turned positions may be relevant or important even when doing short examinations, since the patient may be advised to take a certain position during the EIT measurement. Therefore, a maximum of possible positions should be accounted for, including but not limited to sideways turned positions.

(22) In FIGS. 3a and 3b is shown the zone distribution of a thorax in varied lying positions. In said Figures, the zone distribution of a thorax in neutral supine position (FIG. 3a) is compared with the zone distribution of a thorax in sideways turned supine position (FIG. 3b). In sideways turned supine position, according to FIG. 3b, angle β is defined by the position of the dorsoventral axis 19 with respect to the gravity vector g. As depicted above and in FIG. 1c, in lying position gravity vector g equals to gravity vector component g.sub.c. Thus in general, in lying position, e.g. in supine or in sideways turned supine position, β is the angle between the dorsoventral saggital axis and the gravity vector component g.sub.c. Also when the test person takes a position between supine position and upright position and is additionally turned to one side, β is the angle between the dorsoventral axis 19 and the gravity vector component g.sub.c. The dependency of the zone formation on the rotation around the longitudinal axis 17 of the patient is reflected in an asymmetry of the zone distribution in the two lungs. The zones are stacked in layers that are perpendicular to the gravitational vector g. If the body is turned to one side by the angle β, the zones still follow gravity as shown in FIG. 3b. In this example, the angle α of the gravity vector g relative to the body axis remains unchanged at 90 degrees. In other words, the gravitational vector lies within the observation plane and the zones are maximally expressed. In the sideways turned position, the mechanically ventilated lung situated lower (i.e. the right lung in FIG. 3b) may experience extensive aeration reduction, while the other lung (i.e. left lung) is affected barely.

(23) In FIG. 4, left side, various positions of the patient (schematic of the thorax) are depicted with respect to the gravitational vector g. On the right side are shown respective angles α and β by which the two coordinate systems u/h and u′/h′ are related to each other. Both coordinate systems have the same point of origin. Again, α represents the angle between the main body axis 17 and the gravity vector g, and β, represents the angle between the dorsoventral axis 19 and the gravity vector component g.sub.c. In the orientations given in the first three positions of FIG. 4, gravitational vector component g.sub.c equals gravitational vector (g.sub.c=g) and angle β defines the rotation angle between the u/h axis system and the u′/h′ axis system. In the bottom example (upright position), angle β is not defined (g.sub.c=0), but it does not matter since there is hardly any zone expression. Vectors ω.sub.u′ and ω.sub.h′ result from the respective rotation of the u′/h′ axis system versus the u/h axis system and define a rotated variant of the spatial filtering (see below).

(24) In further embodiments the invention combines spatial sensor data (which may be measured with a tri-axial acceleration sensor, for determining the orientation) and temporal sensor data (which are for example measured with a dilation sensor, a further acceleration sensor and/or an electrical and/or acoustic activity sensor) to improve the images and related information obtained by EIT.

(25) In one embodiment, the excursion (i.e. dilation or expansion) of the belt-like structure, or parts of the belt-like structure is measured, for example by a strain gauge. A signal processor calculates the deviation from a local minimum and subsequently projects the onset of inhalation back in time. The data vector that was measured at that back-projected point in time is taken as the “breath reference vector”. Since this procedure involves a certain delay in image processing, the goal of implementation is to find the balance between accuracy and timing. Since breathing in adults is done at frequencies well above three breaths per minute and well below 50 breaths per minute, a delay of a few hundred milliseconds can be tolerated and should be sufficient to reliably generate the reference point in time. In fact, pulmonary function testing often relies on a test called “occlusion test” to measure the respiratory drive of a patient. Such a test is done in 100 milliseconds in adults and in children. Finally, the “breath reference vector” is subtracted from each EIT data vector, making visible impedance changes that are caused by breathing activities and ventilation.

(26) In one embodiment, the dilation sensor (also called breath sensor) is measuring the force on the belt-like structure by means of a strain-gauge and the body position by means of a 3D acceleration sensor. The user may input two parameters c and d to adjust the image with respect to zones 1, 2, and 3.

(27) As the patient inhales or a ventilator delivers a breath to the patient, the belt-like structure expands and the strain-gauge measures this expansion. The sensor signal is converted to digital format, typically at 50 to 100 samples per second, and may be analyzed digitally. Analysis may include a simple low-pass filter and subsequent determination of minimal force. Alternatively, analysis may be done as in FIG. 5, illustrating the analysis of a strain gauge signal to find the EIT reference vector in presence of significant signal drift. Arbitrary units of a strain gauge (solid line) and its derivative (for example high-pass filter at RC=0.1 sec, dotted line) are plotted during breathing while the lung volume is being continuously changed. The local minimum of the strain gauge (solid arrows) is around 0.1 second before every zero crossing of its high-pass filtered derivative (dotted arrows).

(28) The sensor data, typically from a strain gauge, is first filtered by a low-pass filter with a cut-off frequency of 20 Hz. Thereafter, a high-pass filter is employed with a cut-off frequency of 0.1 Hz. If the so filtered curve crosses the zero-line going from low force to high force (dotted arrows), the chest is starting to expand and thus this point is taken as “start of inhalation”. The EIT data vector measured at this point in time, or a predefined lead time earlier, typically one time constant of the high-pass filter earlier, is taken as the reference vector for subsequent differential EIT imaging. This procedure will introduce a slight delay in the image sequence which, however, is clinically irrelevant.

(29) Alternatively, the breathing activity is measured by a second acceleration sensor (e.g. replacing above mentioned strain gauge sensor). Since the chest is moving with every breath, the second acceleration sensor can sense this movement and turn it into a signal that can be used to indicate the onset of inhalation and to determine the reference vector as described above.

(30) Further alternatively, the same acceleration sensor measures both, the breathing activity and the direction (i.e. orientation) of the gravity vector relative to the observation plane. For this purpose a 3D acceleration sensor may be used.

(31) In another embodiment, the temporal data sensor is an electrical activity sensor, for example a sensor as used in electrocardiography (ECG), i.e. an electrocardiography sensor. Such a sensor can be used to create reference images related to the onset of the heart contraction thereby creating a “heart reference data vector”. Thereafter, the “heart reference data vector” is subtracted from each EIT data, making visible impedance changes that are caused by heart activity.

(32) In another embodiment, the temporal data sensor is an acoustic activity sensor or microphone, for example a sensor as used in phonocardiography (PCG), i.e. a phonocardiography sensor. Such a sensor can be used to create reference images related to the onset of the heart contraction thereby creating a “heart reference data vector”. Thereafter, the “heart reference data vector” is subtracted from each EIT data, making visible impedance changes that are caused by heart activity.

(33) In another embodiment, at least two temporal data sensors are combined with a spatial data sensor. For example, an electrical or acoustic activity sensor, e.g. a sensor as used in electrocardiography or phonocardiography, respectively, and a dilation sensor, e.g. a strain gauge sensor, are combined with a 3-D acceleration sensor (spatial data sensor). This allows to measure orientation of the observation plane with respect to the gravity and at the same time dilation of the electrode belt due to breathing and electrical activity due to cardiac activity.

DESCRIPTION OF USE OF THE INVENTION

(34) The present invention can be used to enhance the image quality of EIT devices in stand-alone monitors and in mechanical ventilators and anaesthesia machines. Such improvement can be done by either creating the enhanced images or by plotting the gravity vector directly on the image, or automatically rotating the image with respect to the gravitational vector thereby providing orientation to the user. A particular use of such improved EIT images is to initiate specific therapies such as recruitment manoeuvres, physiotherapy, or changes in posture and to measure the effectiveness of the therapeutic interventions.

(35) A typical application of the sensor device for EIT imaging, the EIT imaging instrument and EIT imaging method according to present invention is in mechanically ventilated intensive care patients. About 15% of these patients suffer from acute lung injury and more than 30% of these die. It is estimated that about half of these patients could be saved by adequate treatment. Such treatment involves lung recruitment to effectively minimize zones 2 and 3. However, lung recruitment manoeuvres entail risks. Clinicians therefore often use lung recruitments only when lung damage has already become obvious. Unfortunately, this is often too late. With the disclosed invention, a care provider would have the means to judge the need and the success of lung recruitment manoeuvres early in disease, save lives, and reduce cost of care.

(36) In another use, the context sensitive EIT can be used to optimize the body position of a patient with respect to lung function.

EXAMPLE

(37) Below is depicted an example for reconstructing an EIT image by using the finite element method and adjusting raw EIT data with respect to position and orientation of the patient according to present invention.

(38) The reconstruction problem is solved using the finite-element method (FEM). The FEM uses a mesh of triangular elements, defined by nodes, to discretize the space or surface of interest. Then the physics of the problem is applied to the mesh and the problem is solved using given boundary conditions.

(39) Doing this for EIT, one gets
Y(σ)*V=C,

(40) where Y is the conductance matrix, depending on the conductivity σ, V is a set of voltage distribution and C is a set of applied currents.

(41) Given that one can only measure the voltage at the medium's boundary at given locations (i.e. at the electrodes) the operator D is introduced. It returns a vector v of voltage measurements corresponding to a given system and scanning pattern,
v=D(V)=D(Y.sup.−1*C).

(42) The above equation is then linearized with respect to σ using Taylor expansion,
Δv=S Δσ,

(43) where S is the sensitivity matrix

(44) $\frac{\delta v}{\delta \sigma }{|}_{{\sigma }_0},$

(45) Δσ is σ−σ.sub.0.

(46) and Δv is v−v.sub.0.

(47) For a given change in the measurements Δv, we thus obtain a change in conductivity Δσ. In the image reconstruction process, the idea is to find the change in conductivity Δσ from a given change in the measurements Δv.

(48) To compute Δσ one has to invert the matrix S. This operation is in general non-trivial and cannot be performed using the classical inverse of a matrix. This category of problem is known in the literature as inverse ill-posed problems. This means that the problem has more unknowns than equations. A way to calculate solutions, despite the ill-posed nature, is to use a regularization technique which implies that some assumptions are made about the medium of interest. The idea in EIT is essentially to find a least-square solution Δσ of the problem (∥SΔσ−Δv∥.sup.2). Since the problem is ill-posed, a regularization term is added yielding the following cost function (see, for example, Adler A, Guardo R, Electrical impedance tomography: regularised imaging and contrast detection, IEEE Trans Med Imaging, 1996, 15 170-9):

(49) $\begin{array}{cc}\Phi =\frac{1}{2}{.Math.S\mathrm{\Delta \sigma }-\Delta v.Math.}^2+\frac{\lambda }{2}.Math.F\Delta \sigma .Math.,& \left(I\right)\end{array}$

(50) where λ is the weighting term of the regularization term and F is a spatial high-pass filter matrix.

(51) One can note that the use of the Euclidian norm (squared) is not mandatory; another norm can also be used.

(52) In the art (Adler and Guardo, 1996), F.sub.freq can be modeled as a high-pass Gaussian spatial filter of the form:
F.sub.freq(u, h)=1−e.sup.−ω.sup.0.sup.2.sup.(u.sup.2.sup.+h.sup.),

(53) where ω.sub.0 is the cutoff frequency, and u, h variables are the ordinate, abscissa directions in spatial frequency space, respectively (FIG. 4). The application of this filter results in an image with filtered high spatial frequencies, i.e. a smoother image. It is possible to extend the same idea in order to filter more or less the high frequencies in a particular direction:

(54) $F_{\mathrm{freq}}\left(u,h\right)=1-e^{-\left({\omega }_{u^{\prime }}^2{u}^{\mathrm{\prime 2}}+{\omega }_{h^{\prime }}^2{h}^{\mathrm{\prime 2}}\right)},$

(55) where u′=u cos(β)−h sin(β) and h′=u sin(β)+h cos(β) are the axis direction rotated by an angle β (see FIG. 4) around (u.sub.0, h.sub.0)=(0,0). The new cutoff frequencies are ω.sub.u′ and ω.sub.h′ linked to the u′ and h′ axis respectively.

(56) The way to calculate F is given in the following paragraph. First of all it is important to note that F as given in equation (I) is the filter for the finite-element mesh space, so that in order to calculate F the following workflow is used:

(57) FIG. 6 depicts schematically the process used to form the pixelized impulse image for one single node of a mesh. For each mesh node, an impulse image is generated. In other words, for each node, one creates a new mesh where all the node values are set to zero except for the considered node, which is set to one. Then, the meshes obtained are pixilized using a sufficiently fine grid; during this process the impulse images of the meshes are linearly interpolated on their corresponding pixel grids between nodes.

(58) FIG. 7 depicts the formation of the matrix P. Each image is transformed into a column vector, and all image column vectors are then appended to form a matrix P. This matrix P will be used later on.

(59) FIG. 8 depicts the filtering process in the spatial frequency domain. At this stage, one also applies a 2D-FFT (two-dimensional Fast Fourier Transform) on the impulse images to go into the spatial frequency domain. The filter F.sub.freq is applied and a 2D-IFFT (two-dimensional Inverse Fast Fourier Transform) is performed to go back into the spatial domain. One thus obtains a set of impulse response images.

(60) FIG. 9 depicts the formation of matrix R. Each impulse response image is transformed into a column vector, and all image column vectors are appended to form a matrix R.

(61) The spatial-domain filter matrix F is obtained from the following expression:
PF=R.

(62) One uses the pseudo-inverse to extract the matrix F:
F=[P.sup.T P].sup.−1P.sup.TR.

(63) The ratio

(64) $\frac{{\omega }_{u^{\prime }}}{{\omega }_{h^{\prime }}}=c*\gamma \left(\alpha \right)+d$
(for example, γ(α)=sin(α), d=c=1) is one of the main filter parameter and is in direct relationship with the angle α.

(65) The invention proposes a sine function for the gamma function but it is understood that any other function could be used. For example, γ(α)=α/90, d=1, c=0, where α is between 0 and 90 degrees, could be used. The gamma function could also be a mathematical model that represents the physiological zones (1 to 3) expression described above. The same notice is also valid for the F function, because other spatial filtering shapes could be used, for example a rectangle or an ellipse.

(66) While the invention has been described above with reference to specific embodiments and examples thereof, it is apparent that many changes, modifications, and variations can be made without departing from their inventive concept disclosed herein. Accordingly, it is intended to embrace all such changes, modifications and variations that fall within the spirit and broad scope of the appended claims.