System for investigating biological structures

Abstract

A system 1 for investigating biological structures, the system 1 comprising: a probe 9 for application to a subject under investigation, said probe comprising a plurality of electrodes 11 arranged in an array, the electrodes of said array being operable to apply electrical signals to said subject and to receive signals that are indicative of electrical properties of said subject; a probe controller 5 responsive to control signals from a controller to select electrodes of said plurality to form a sensor, said sensor being operable to apply electrical signals to a region of said subject and to output received signals that are indicative of electrical properties of said region of said subject; a measurement module 7 coupled to the probe controller for the receipt of said received signals, said measurement module being operable to determine from said received signals the electrical properties of said region of the subject; and a system controller 3 operable to output control signals to said probe controller, the control signals specifying the electrodes of the array that are to be selected to form said sensor

Description

[0001] SYSTEM FOR INVESTIGATING BIOLOGICAL STRUCTURES

[0002] Field

[0003] This disclosure relates to systems, particularly but not exclusively, to systems for investigating biological structures. One implementation of the teachings of this disclosure provides a system that may be operated to provide an indication of potential vascular occlusion. This disclosure also relates in one aspect to sensors for use in such systems, and in another aspect to interpretation modules for such systems. Another aspect of this disclosure relates to a dressing for accelerating wound healing.

[0004] In aspects of this document, the term "occlusion" is not intended to refer to a total occlusion of a vessel, rather it is intended that the term should be interpreted to mean any partial or total blockage of a vessel that affects the rate at which fluid can flow through that vessel.

[0005] Background

[0006] Bioimpedance measurements using a tetrapolar (4-electrode) sensor on top of the skin, or on the surface of a biological structure or organ have previously been contemplated in a range of healthcare applications including: Cardiovascular Monitoring through impedance plethysmography for Blood Pressure Monitoring; cardiac output measurements across the chest and implantable devices to monitor heart condition; body composition analysis; Fluid Status Monitoring in heart failure management and dialysis; Respiratory monitoring through impedance pneumography for sleep apnoea detection and electrical impedance tomography (EIT) imaging of lungs; Brain Activity Monitoring; Hydration and Electrolyte Metabolic Studies; Tumour Detection; Foetal Monitoring; Continuous Glucose Monitoring through implantable devices close to arteries; Gastrointestinal and Oesophageal Monitoring; and others.

[0007] Tetrapolar sensors are considered superior in performance when the four electrodes are placed exactly over the area or biological feature that need to be monitored. However, this poses issues, as in many cases this area or feature may be dynamically varying in size and / or position, requiring a specialised operator to have means to know its exact location at the time of measurement and then to reposition, resize and re-space the electrodes accordingly.

[0008] In consequence, whilst impedimetric sensing is rapid and appropriate for real time continuous and chronic monitoring, the requirement for continuously manually shifting the electrodes with precise knowledge of biological feature targeted, even if these are located below the skin and not visible, tends to make this technique more useful in experimental or research settings.

[0009] Aspects of this disclosure are concerned with the foregoing.

[0010] In general terms, in one aspect the present disclosure relates to a system for investigating biological structures, the system comprising: a probe for application to a subject under investigation, said probe comprising a plurality of electrodes arranged in an array, the electrodes of said array being operable to apply electrical signals to said subject and to receive signals that are indicative of electrical properties of said subject; a probe controller responsive to control signals from a controller to select electrodes of said plurality to form a sensor, said sensor being operable to apply electrical signals to a region of said subject and to output received signals that are indicative of electrical properties of said region of said subject; a measurement module coupled to the probe controller for the receipt of said received signals, said measurement module being operable to determine from said received signals the electrical properties of said region of the subject; and a system controller operable to output control signals to said probe controller, the control signals specifying the electrodes of the array that are to be selected to form said sensor.

[0011] In a preferred arrangement the sensor may comprise first and second electrodes for the application of an electrical signal (such as a current, e.g. an AC current, or a voltage) to said region of the subject. The sensor may comprise third and fourth electrodes for the receipt of signals that are indicative of an electrical property of the region of the subject. The sensor may comprise a tetrapole.

[0012] In one arrangement the electrodes of said tetrapole may be linearly arranged. The linearly arranged electrodes of said tetrapole may comprise a line of four electrodes that are immediately adjacent to one another. The linearly arranged electrodes of said tetrapole may comprise a line of four electrodes that are each spaced from one another by one or more electrodes. Varying the tetrapole sensor dimensions (e.g. the distance between the middle electrodes, electrode size etc) allows for the interrogation of different areas, depths and biological feature sizes from one side of access on the surface of a biological structure (e.g. the skin). Varying the interelectrode distance (particularly between the electrical signal sensing electrodes) can be used to vary the penetration depth and the volume of the material or tissue being measured.

[0013] In one implementation, the system controller may be operable to output a plurality of control signals each specifying a different selection of sensors for investigating different regions of said subject, the arrangement being such that the control signals define a sensor that tracks across the subject to investigate multiple regions of the subject.

[0014] The control signals may each specify a different selection of sensors for investigating different linearly arranged regions of said subject. The control signals may be configured to track the sensor over the subject in different directions to investigate multiple linearly arranged regions of said subject. An advantage of this arrangement is that sweeping over the subject from different directions enhances the investigation by improving the resolution.

[0015] In another arrangement the electrodes of said tetrapole may be arranged in a polygon, such as a square or rectangle. The system controller may be operable to output a plurality of control signals each specifying a different polygonally arranged selection of sensors for investigating different regions of said subject, the arrangement being such that the control signals define a sensor that tracks across the subject in concentric polygons to investigate multiple regions of the subject.

[0016] In one implementation the received signals may be indicative of the electrical impedance of the region of the subject investigated by the sensor. The sensor may be operable to apply electrical signals at multiple frequencies to said region of said subject. The measurement module may be operable to determine tissue type from received signals.

[0017] Another aspect of this disclosure relates to a system for providing an indication of potential vascular occlusion, the system comprising: a sensor comprising an array of a plurality of electrodes; means operable to energise groups of electrodes to generate signals representative of the impedance of an area of a subject to whom the sensor has been attached, and an interpretation module operable to receive said signals and determine therefrom an indication of potential vascular occlusion.

[0018] In one implementation the interpretation module may be configured to generate a vascular image of the area of the subject to which the sensor has been attached. In others, the interpretation module may simply be configured to warn the operator in the event that a potential vascular occlusion is indicated.

[0019] In one envisaged arrangement, a sensor is provided that comprises an array of electrodes for applying a voltage to a subject and an array of electrodes for measuring the impedance of the subject. One or both of the arrays may be dynamically reconfigurable so that the pattern of electrodes in use at any one point in time can be varied. In one envisaged implementation a single array of electrodes may be employed, with some of said electrodes being operable to apply a voltage or other electrical signal and others being operable to receive signals indicative of impedance. The electrodes may be operated in groups, with a subset employed to apply an electrical signal to a subject and another subset employed to receive electrical signals.

[0020] A control system may be provided to successively engage adjacent groups of electrodes one after the other in a first direction, and then to successively engage adjacent groups of electrodes one after the other in a second direction different to the first. This arrangement allows the control system to sweep over the sensor in multiple directions to enhance detection of features in the subject.

[0021] An additional implementation of the teachings of the invention provides a method for operating a system of the type disclosed herein. In one implementation the method may be employed to create an electrical map or image of a subject, or a region of a subject. In another implementation the method may be employed as a means for monitoring a subject or region of a subject, or a biological process occurring in said subject.

[0022] One further implementation of the teachings of this disclosure provides any novel combination of features disclosed herein (regardless of whether they are disclosed in the “field”, “background”, “summary”, or “detailed description” sections or any of Paper 1 , Paper 2 and / or Paper 3).

[0023] Other advantages and aspects of the system disclosed herein will be apparent from the detailed description provided below.

[0024] Brief Description of the Drawings

[0025] The teachings of this disclosure, and arrangements embodying those teachings, will hereafter be described by way of illustrative example with reference to the accompanying drawings, in which:

[0026] Fig. 1 is an illustrative diagram of a system that implements the teachings of this disclosure; and

[0027] Fig. 2 is a schematic representation of a probe that may be employed with the system of Fig. 1.

[0028] Detailed

[0029] Details of illustrative implementations of the teachings of this disclosure are provided below by way of example only. Additional illustrative details are provided in the attached documents: Paper 1 , Paper 2, and Paper 3. All of these papers are to be treated as being incorporated in this disclosure as if they were reproduced in full herein. It is also intended for subject matter disclosed in the accompanying papers to be freely portable between them and any other part of this disclosure - that is to say that any feature from one of the accompanying papers may be combined with any other feature from the same paper and / or a different paper and / or from anywhere else in this disclosure without adding subject matter to the present disclosure.

[0030] In general terms, Paper 1 is concerned with the fact that combining multiple smaller electrodes to create an 'Effective' electrode yields superior sensitivity in impedance measurements using a co-planar electrode system, compared to larger electrodes. A disadvantage of the arrangement disclosed is that the user must select the electrode position based on prior knowledge of the sample geometry.

[0031] Paper 2, in general, demonstrates a practical approach whereby splitting a conventional strip electrode into multiple smaller electrodes (sub-electrodes) enables the selection of a preferred sub-electrode from each group by comparing the measured impedances from different combinations. This technique enables us to choose the most appropriate set of electrodes closest to an artery or a structure, enhancing the practicality of the technique.

[0032] Paper 3 is concerned with a multi-nodal electrode array that has a fixed interelectrode distance and an in-line tetrapolar topology, and the use of that array for mapping of the impedance of a superficial structure. Combining the four directional scanning to obtain a two-dimensional mapping of the impedances enabled the location of a structure on or below the skin with greater precision than can be obtained from single directional scanning. The ability to locate a subcutaneous structure using this four-directional scanning could be used to detect vessel edges more accurately and the flow of liquids (e.g. Blood) that can potentially be used in impedance plethysmography.

[0033] In contrast to Paper 1 and 2, which are concerned in general with the division of larger electrodes into smaller sub-electrodes for enhanced sensitivity in impedance measurements, Paper 3 discloses a different implementation. In general terms, the arrangement disclosed therein uses the directional sensitivity distribution of in-line tetrapolar electrode topology and combined multidirectional scanning to achieve a higher resolution, and hence provide an improved location identification system.

[0034] In implementations of the disclosure, tetrapolar sensors of the type disclosed herein consist of four electrodes: two for injecting current (outer electrodes) and two for measuring voltage (inner electrodes). Desirable tetrapolar electrode impedimetric sensor properties that conventionally require electrode replacing, shifting, resizing and / or moving include electrode spacing, geometry, size and shape.

[0035] With such sensors, the interelectrode distance (particularly between the voltagesensing electrodes) affects the penetration depth and the volume of the material or tissue being measured, and hence varying the interelectrode distance allows the penetration depth and volume measured to be varied.

[0036] That notwithstanding, having the current injection electrodes close to their neighbouring voltage-sensing electrodes tends to reduce nonlinearities (sensor systematic errors due to negative sensitivity). Traditional EIT scans volumes contained between the electrodes with the current injection electrodes remaining fixed in place while shifting the measuring pairs, while shifting the whole tetrapole would introduce fewer sensor-related errors.

[0037] Electrode geometry (e.g., in-line, square, or concentric polygon arrangements) alters the focal area, affecting the specific region of the material or tissue that is probed.

[0038] The size of the electrodes impacts the amount of noise and interference in the measurement; larger electrodes may reduce noise but reduce resolution, particularly in sensitive measurements. Smaller electrodes can improve localization but may introduce more variability.

[0039] In some applications, electrode shape affects so-called edge effects, e.g. square electrodes introduce edge-effect nonlinearities relative to round or polygonal electrodes.

[0040] There are two modes of operation of the systems disclosed herein: (i) scanning, where an area is swept by shifting a set of four electrodes sequentially to capture I map information about the area or volume or features below the sensor area; and (ii) chronic monitoring of a target (e.g. vasculature, tumour, wound, etc) to observe variations. Scanning may be accomplished through the use of tetrapolar combinations of the electrodes (or electrode pixels) of the array in their original size. Monitoring can be done through the use of larger “effective” electrodes formed through grouping the smaller electrodes accordingly to form a tetrapole comprised of four groups of electrode pixels that can carry out tetrapolar sensing.

[0041] Advantages of the arrangements disclosed herein are as follows:

[0042] • Scanning with in-line tetrapoles formed of adjacent electrode pixels in multiple different directions enhances spatial resolution. If the matrix I array of electrode pixels is orthogonal there can be four scans (horizontal, vertical, diagonal left and diagonal right) characterized by the direction of shifting of the active scanning tetrapole. If the matrix I array of electrode pixels is hexagonal, or forms any other polygonal combination, more scanning directions can be incorporated offering superior resolution. The reason that such a combination of scans offers superior resolution to a single direction scan has to do with the distribution and alignment of different types of cells across different layers of tissue.

[0043] • Scanning with in-line tetrapoles as above, where the tetrapoles are formed by two electrode pairs for each injection and neighbouring voltage measuring electrode with a larger interelectrode distance between measuring electrodes (achieved through omitting an electrode in the array for example) can achieve a larger penetration depth. In-line tetrapole sensors with larger interelectrode distances between the measuring electrodes used for scanning in multiple directions tend to achieve enhanced spatial resolution at a different depth than smaller tetrapole sensors.

[0044] • Scanning with different topology geometries, eg with concentric polygons or with squares (formed by four electrode pixels at the four corners of a square geometry) can achieve high resolution mapping with different focal areas, i.e. different spatial resolution outcomes. Combining different geometries together with scans as the aforementioned ones can achieve even more superior resolution. Increasing interelectrode distances can interrogate greater volumes and depths.

[0045] • Once a structure is detected through a “first pass” it is possible to re-scan the area of interest with a combination of tetrapoles as described above to achieve higher localised resolution while reducing computational load that would be required to apply the refined scan to the overall area.

[0046] • If electrical noise should affect measurements with single electrode pixel tetrapoles, combining pixels in groups of two or more allows one to achieve a scan with reduced electrical noise.

[0047] • In the context of chronic I dynamic monitoring, for example, following a scan and the localisation of a targeted structure, algorithms may be used to identify the closest pixels to the structure that would not generate redundant measurements.

[0048] • Again, in the context of chronic I dynamic monitoring, larger “effective” electrodes (in any desired or necessary shape, groups forming round or square or stripe effective electrodes) can be formed to formulate an in-line tetrapole for monitoring the location and depth of the targeted structure.

[0049] • Again, in the context of chronic I dynamic monitoring, repeating the scanning process at times (but not continuously) to re-arrange the effective tetrapole as needed to monitor in real time and with high measurement sensitivity the targeted area.

[0050] • Once again, in the context of chronic I dynamic monitoring, multiplex measurements between effective tetrapoles can be employed to target different features, either cutaneous or subcutaneous. The effective tetrapoles can be inline, square, concentric polygons or any other geometry required for each target. All of the above may be combined with impedimetric (or other) instrumentation that takes multifrequency measurements for enabling the determination of different tissue types under the area of measurement.

[0051] Aspects of this disclosure are concerned with and / or provide: • A planar impedimetric scanning method that improves spatial resolution relative to other existing approaches like Electrical Impedance Tomography (EIT) via multidirectional sweeping of an area (or in other words: Improved resolution through multi-directional sweeping)

[0052] • Fixed tetrapole sensors (or sensors) used in impedimetric scanning provide a more predictable, and repeatable outcome, and importantly, reduce the need for complex and idealised forward models to generate an outcome (or in other words, Fixing the inter-electrode geometry removes nonlinearities in the measurement that are present in all conventional approaches).

[0053] • Varying the tetrapole sensor dimensions (i.e. the distance between the middle electrodes, electrode size etc) allows for the interrogation of different areas, depths and biological feature sizes from one side of access on the surface of a biological structure (e.g. the skin) (or in other words, reconfigurability for depth selection).

[0054] • The increase in resolution and the dynamic reconfigurability of the sensor array allows for the on-demand determination of the most appropriate electrodes to be used as pixels for the formation of larger effective electrodes that can be used to monitor real-time biological processes, without any manual selection, user participation or prior knowledge of the location of the targeted features (or in other words, an operator-independent auto-reconfigurable sensor).

[0055] • The dynamic addressability of the electrode pixels and the ability to reconfigure their combination dynamically allows for the real time monitoring of morphing, shape altering, or developing biological processes in essence creating an auto- reconfigurable sensor that follows a process over time despite changes or alterations (or in other words, it “follows” the target and reconfigures itself automatically to provide the highest possible sensitivity measurement from the most appropriate location).

[0056] Referring now to Fig. 1 , there is depicted an illustrative diagram of a system that implements the teachings of this disclosure. The system 1 comprises, a system controller (labelled CPU) 3, a switching control and analogue switching array (which together provide a probe controller 5), a measurement module 7 (labelled impedimetric instrumentation), and a probe 9 (labelled electrode pixel array). Fig. 2 is a schematic representation of a probe 9 that comprises an array of electrodes 11 , and schematically depicts four illustrative different directions (denoted by arrows) that sensors may be swept over a subject or region thereof.

[0057] As aforementioned, Fig. 1 is a block diagram of a system that can implement fixed- tetrapole scanning and also grouping electrodes to form electrodes of larger effective area. The system includes a microcontroller or a microprocessor or a CPU, which sends electrode assignments and grouping signals to the switching control block as well as a trigger signal to the impedance measuring circuitry (impedimetric instrumentation) to take impedance measurements. The switching control block interprets the signal from the CPU and sends the desired signal to the analogue switching array. The analogue switching array contains the switches which interface between the impedance analyser and the electrode array. Depending on the signals received from the switching control block, the analogue switching array connects the desired electrodes to their respective channels of the impedimetric instrumentation block. The analogue switching array is configured to be capable of connecting any, all or none of the electrodes to one or more channels of the impedimetric instrumentation block. This allows the scanning and grouping of the electrodes as well as the tetrapole sensor topology to be freely re-configured as desired.

[0058] Whilst various aspects and embodiments of the system are described in the accompanying papers, the scope of the present disclosure is not limited to the particular arrangements set out therein and instead extends to encompass all arrangements, and modifications and alterations thereto. It should also be noted that the scope of this disclosure includes any combination of features herein disclosed, including features disclosed in different papers.

[0059] It should also be noted that whilst the accompanying claims set out particular combinations of features described herein, the scope of the present invention is not limited to the particular combinations hereafter claimed, but instead extends to encompass any combination of features herein disclosed.

[0060] In addition, that whilst embodiments of the present invention have been described above in the context of software modules that are executable by a processor, it should be noted that the scope of the present invention is not limited to an implementation of the teachings of the invention in software. Rather, the skilled person will immediately appreciate that the functionality described herein may equally be implemented in hardware (for example, by means of a plurality of application specific integrated circuits (ASICS)) or, indeed, by a mix of hardware and software.

[0061] Finally, it should be noted that any element in a claim that does not explicitly state "means for" performing a specified function, or "steps for" performing a specific function, is not to be interpreted as a "means" or "step" clause as specified in 35 U.S.C. Sec. 112, par. 6. In particular, the use of "step of" in the claims appended hereto is not intended to invoke the provisions of 35 U.S.C. Sec. 112, par. 6. PAPER 1 10

[0062] A novel approach to transcutaneous localization of blood vessels using a dynamically reconfigurable electrode (DRE) array

[0063] Abstract — non-invasive transcutaneous analysis is very need to be positioned directly over the blood vessels (Fig. desirable in medical applications that require blood analysis. la), however this would still require ether: a) the exact Electrical impedance spectroscopy (EIS) is one of the many location of the targeted blood vessels to be known or b) methods that has been extensively researched and is often used visual inspection to be employed especially as the location of for such applications, with its advantages including very high blood vessels differ from person to person. Even after the sensitivity and rapid response. EIS utilizes electrodes to desired electrode positioning is identified, optimal fitting may monitor conductivity variations in blood. However in order to require several attempts. minimize the effect of high impedance tissue and skin surrounding a low impedance, localized subcutaneous vascular A possible solution to this challenge could be addressed structure of interest, one has to locate the structure of interest by the use of multi-electrode array (MEA, similar to that used and then adjust the placement and size of the electrodes to in [9]), whereby electrodes, could be dynamically addressed make the measurement as targeted as possible. It is thus rather than physically re -positioned. However, in order to essential to achieve an appropriate method for a) detecting the eliminate the need for visual inspection and alignment, a vascular structure of interest and b) localizing the solution would be desirable for detecting the targeted transcutaneous measurement so that sensitivity can be structure prior to the measurement taking place. Moreover, in improved and greatly interfering measurements from order to minimize unwanted impedance measurements from surrounding tissue can be disregarded. This study proposes surrounding tissue, electrodes would ideally need to be and assesses the potential use of a multi-electrode array for physically shaped to a size corresponding to the diameter of making dynamically reconfigurable electrodes (DRE), by the targeted vessel. Thus oversized electrodes would need to treating electrode segments as pixels that can form re-locating be “trimmed” whilst very small electrodes would entail a and re-shaping electrodes. Simulations preformed in COMSOL indicated that the technique can successfully locate a very high contact impedance, potentially limiting the output transcutaneous structure and achieve double the sensitivity dynamic range of the current injection electronics and relative to conventional electrode topologies. increasing thermal noise at the input of the measurement electronics [5],

[0064] I. INTRODUCTION

[0065] The authors are researching a dynamically reconfigurable

[0066] Transcutaneous analysis of blood has been attempted electrode (DRE) system that will allow the four electrodes of using different monitoring methods in order to identify a tetrapolar configuration to automatically re-shape and rechange in the composition of blood [1], [2], One such method locate so as to both trace and focus on a subcutaneous blood is electrical impedance spectroscopy (EIS) [3] which offers vessel (Fig. lb). It will utilize a MEA, but with each very high sensitivity and rapid response. The technique electrode acting as a “pixel” that can be re-grouped and reinvolves impedance measurements taken from the surface of combined to form larger electrodes of different shapes and the skin and can be carried out using a tetrapolar (four sizes and that can also be freely positioned at any location electrode) arrangement to minimize the influence of electrode within the array. properties to the measurements [4], [5], The impedance measured can then be used to analyze different characteristics In EIS such grouping of electrodes has not been reported, such as blood flow [6], levels of blood analytes like e.g. however it has been reported in other applications of glucose [7] or even to carry out impedance cardiography [8], electrical bio-interfacing such as electrical neurostimulation. In some or all of the above applications impedance Popovic-Bijeli et al. [9] developed a 24 segment electrode measurements are taken by using ether band electrodes, array for electrical stimulation of nerves in the arm, with which wrap loosely around a section of the body, or spot segments combining to increase the size and to change the electrodes, which are usually disk shaped electrodes applied position of the stimulating electrodes. Lawrence et al. using adhesive tape [6], [8], The problem associated with the increased the number of segments (256) to improve the use of such conventional electrodes is that the measurements stimulation’s selectivity

[0010] , of low-impedance blood variations are typically Whilst the above methods were investigated for overshadowed by the high impedance of skin and tissue neurostimulation, to the knowledge of the authors they have surrounding the region of interest. [5], [6], [9], For not been proposed for use in transcutaneous impedimetric transcutaneous analysis of blood it would be highly desirable analysis of blood. Moreover, these studies presented the for its impedance contribution to the measurement to be benefits of re-shaping electrodes but did not demonstrate enhanced [3], To achieve this, the electrodes would their use for “mapping” the exact location of a structure of ideally interest.

[0067] In this paper we carried out FEM simulations to assess the front-end of our tetrapolar DRE system, comprising a PAPER 1 11 p p a t t s p

[0068] I t t a e

[0069] A s 3 c f r e t p p h o Fig. 2). The shape and size of electrodes was chosen from an

[0070] Two electric current models (ec and ec2) were set up one existing sensor. Simulations were performed using for the injection electrodes and one for the measurement COMSOL ’s AC / DC module using a quasi-static solution of electrodes. In both models the left electrode was set as the Maxwell’s equations to solve for the electrical potential. The current injection terminal with a 1 A current being assigned as electrode / electrolyte interface double layer was not included a boundary condition and the remaining electrodes were set in the simulation. The relative permittivity and electrical to ground terminal. A parametric sweep was set up which conductivity for blood and fat were taken from

[0011] , increased the conductivity of blood from its original value by 0.005S / m until the conductivity of blood was increased to 0.05S / m over the original conductivity. The sweep represents variations of blood analytes (e.g. ions) that would be measured by a transcutaneous EIS system and its purpose was to allow for the DRE sensitivity to be assessed in comparison to that of conventional electrodes.

[0071] B. DRE pixelated MEA setup

[0072] For the DRE setup, an identical second model was created, with the large electrodes replaced by an array of 8x16 round electrodes with a diameter of 0.6mm (Fig. 3). The array was designed such that two columns of 8 electrode pixels would equal the size of the first model’s large measurement electrodes. The spacing between the elements of the array was setup to be 1mm center to center from the surrounding elements. This was done so that the array would cover the same area as the previously used large electrodes.

[0073] The next set of simulations was carried out in order to assess the ability of the DRE array to identify the location of the blood vessel. For this purpose, impedance measurements need to be taken from the entire area covered by the array. This simulation was carried out by keeping the conductivity

[0074] Figure 1: Tetrapolar configuration including current carrying (CC) and pick-up (PL) electrodes using a) fixed size and positioning (conventional of blood constant, the program then selected four electrodes setup), and b) DRE formed using a pixelated MEA. horizontally adjacent to one another on the array (Fig. 4a) and ran the simulation, then changed the selection of PAPER 1 12 electrodes to the next four electrodes vertically below the previously selected ones (Fig. 4c) and ran the simulation again. This process was repeated till measurements were taken from all the eight rows of electrodes (array height), the program then shifted the selection of electrodes to the right horizontally by one electrode (Fig. 4b) and carried out impedance measurements for all eight rows. In this manner (a) (b) the program virtually scanned the entire area covered by the Figure 5. a) CC electrodes, b) PU electrodes, of the final DRE tetrapolar array. configuration, with their position and shape “focused” on the blood vessel, that has been determined by the on the scanning process (Fig. 6) .

[0075] C. Focused measurement simulation.

[0076] The results from the scan carried out in the second model III. RESULTS AND DISCUSSION (shown in the next section - Fig. 6), identified the location A. Multi electrode scan and optimum number of pixels positioned above the blood vessel, thus allowing the formation of the tetrapolar DRE The results from the scan were plotted to visualize the configuration shown in Fig. 5. Each DRE consisted of four impedance of the area under the array (Fig. 6). From the plot pixels. The resulting “focused” tetrapolar DRE configuration it can be seen that the area where the blood vessel was was then compared with the conventional tetrapolar located exhibited lower impedance from its surroundings, configuration by using the same parametric simulation due to the high conductivity of blood which allows the procedure that was used for the first model. majority of the injected current to flow through it, hence reducing the impedance measured locally by the sweeping four electrode pixels. In that manner the location of the blood vessel could be identified allowing for the selection of electrode pixels positioned above it to be grouped to form optimum DREs aiming towards a more focused impedance measurement. in

[0077] 1 2 3 4 5 6 7 8 9 10 11 12 13 ~

[0078] Array width (pixels)

[0079] Figure 6. Impedance scan plot, a) top view, b) side view. The location of the low impedance subcutaneous blood vessel through the high impedance fat layer is clearly detectable, especially in (a). PAPER 1 13 subcutaneous blood vessel, below a layer of fat. This leads

[0080] Impedance sensitivity for conventional and to the identification of the optimal size and positioning of DRE tetrapolar configurations electrode pixel groups that lead to the formation of DREs. The resulting “focused” tetrapolar configuration exhibits

[0081] Figure 7: Comparison of both tetrapolar configurations in terms of their respective impedance sensitivity normalized to their respective baseline values. The impedance variation was a result of the same parametric conductivity sweep applied to both models.

[0082] B. Focused electrode results

[0083] The parametric impedance variation was applied to both the conventional “large” electrode tetrapolar configuration (1stmodel) and to the DRE tetrapolar configuration (2ndmodel) that was formed following the four electrode pixel scan-and-detect sweep described above. Measurements were plotted in Fig. 7 normalized to start from the respective baseline impedance measurement for each tetrapolar configuration. The results for the conventional model show that for each increment in conductivity of blood by 0.005 S / m there was a decrease in measured impedance of approximately 2.86Q ± 0.03Q. For the same conductivity step the DRE configuration exhibited a decrease in measured impedance of approximately 5.32Q ± 0.06Q. This is almost double the change in impedance measured using the conventional electrodes. Thus, the comparison clearly indicates that the DRE configuration can greatly improve the sensitivity of EIS for transcutaneous blood analysis.

[0084] IV. CONCLUSION

[0085] In applications of transcutaneous blood analysis using impedance spectroscopy, the measurements from a subcutaneous blood vessel can be significantly degraded by the much higher impedance of surrounding tissue and skin. It is therefore desirable to accurately select the localization and size of the electrodes to be used. Within the context of their research towards a dynamically reconfigurable electrode (DRE) system the authors carried out COMSOL simulations to assess the performance of the system’s frontend. This paper presented the results of these FEM simulations which were carried out to assess (a) the ability of a pixelated MEA to scan and detect a subcutaneous blood vessel and (b) the sensitivity of a DRE vs. a conventional electrode tetrapolar configuration in cases where a blood component changes its concentration.

[0086] The results indicated to a very satisfactory level that the proposed front-end is capable of scanning and detecting a PAPER 2 14

[0087] Modelling Dynamically Re-Sizeable Electrodes (DRE) for Targeted Transcutaneous Measurements in Impedance Plethysmography in v g b in h e a in e to w t o in through finite element modelling (FEM) simulations in COMSOL. A scanning sequence made use of DRE in single segment variable Fig. 1. Simplied diagram showing positioning of electrodes using: (a) Band tetra-pole (SSVT) mode proved capable to identify the transcuelectrodes taking measurements from the whole section of the arm. (b) Stripe taneous location of the blood vessel of interest and the specific electrodes taking measurements over the vessel and the surrounding tissue and electrode segments located in its vicinity. Impedance measurements (c) Segmented electrodes grouped into localised effective electrodes to take were then taken using these segments connected to form localised measurements only over the targeted vessel. electrodes only placed over the targeted vessel. The resulting localised electrodes exhibited up to 28% increased sensitivity to blood variations relative to larger electrodes. bioimpedance) measurements are taken by injecting a low am¬

[0088] Index Terms — Impedance, impedance plethysmography, plitude alternating current and measuring the resulting potential localised measurements, segmented electrodes. developed across the tissue beneath the electrodes [5],

[0089] The method has been demonstrated with band electrodes

[0090] I. INTRODUCTION wrapped around the neck and the abdomen for cardiac output MPEDANCE plethysmography (IPG) is a simple non- measurements across the thorax [6], or around a limb (Fig. 1 (a)) invasive method that can be used for monitoring changes in [2], [7], [8], e.g. for monitoring blood volume variations in the blood volume in the limbs or in the thorax

[0001] to assess, amongst brachial artery [9],

[0010] , Band electrodes have, however, a major other metrics, cardiac parameters such as cardiac output and drawback: in addition to the impedance of the tissue of interest in stroke volume [2], Unlike other photo-plethysmography (PPG), this case blood — measurements also include the impedance of IPG uses electrical currents in order to take measurements the surrounding tissue. Given the low impedance value of blood instead of light, therefore it is not influenced by ambient light or relative to all other tissue types, this leads to a significant yet skin pigmentation. IPG can also be used to get signals from undesirable measurement baseline that overshadows the small a large number of sites [3], Through the placement of four magnitude changes caused by blood volume variations

[0011] , In electrodes (tetrapolar method [4]) on the skin, impedance ( order to address this, the use of smaller electrodes (e.g. ECG electrodes) has been proposed [1], [6],

[0011] ,

[0012] , close to the vascular structure of interest.

[0091] Experimental comparisons have shown that such electrodes provide a larger signal to noise (redundancy) ratio compared to band electrodes

[0011] in addition to other benefits including causing less discomfort to the patients [6] and potentially removing the effects in the measurement of surrounding vessels. Such measurements are, however, very sensitive to the placement of the electrodes [2],

[0012] relative to the targeted vasculature. PAPER 2 15 resulting localised electrodes are more sensitive to variations in

[0092] Fig. 2. Overall dimensions fortop view of simulation model for both electrode the blood vessel properties relative to the larger stripe electrodes. setups, (a) a 3D representation of the model geometry, (b) Stripe electrode setup with dimensions (top view), (c) Segmented electrode setup with some segment combinations shown (top view). A. Geometric Model Parameters

[0093] A parallelepiped geometric model with length, width and height of 45 mm, 20 mm and 10 mm respectively was designed

[0094] This issue is not easy to address practically both because that to examine the aforementioned two research question Fig. 2(a). target may be difficult to locate and also because once located The model was used as a template for both the gelatine phantom the electrodes would have to be re-positioned until aligned, used in experiments and for the FEM simulations. It was based causing discomfort and potentially destroying their adhesion. on the brachial artery of the arm, which has been used with It is therefore desirable to develop impedance measurement IPG

[0013] , with a worst-case scenario configuration whereby the electrodes whose active surface can be adjusted to target only artery is surrounded by fat, which is significantly less conductive the exact area of interest, forming localised electrodes, without than blood. It featured three main structures of homogeneously manually re -positioning them. conductive geometries: a 2 mm thick skin layer, a 8 mm thick fat

[0095] In this paper a novel approach to targeted impedance plethyslayer and a 5 mm diameter cylindrical blood vessel embedded mography measurements with dynamically re-sizeable elecin the latter. The blood vessel’s longitudinal axis was on a plane trodes (DRE) is presented. A system involving electrodes of parallel to- and 7 mm deep from the skin surface plane on adjustable effective size has been developed. Four elongated which the electrodes were attached. Assuming that, realistically, rectangular electrodes [8] (named hereafter stripe electrodes perfectly in-line placement of the electrode tetrapole relative as in Fig. 1(b)) have been partitioned into smaller electrode to the blood vessel would be very difficult to achieve, tests - segments (Fig. 1(c)) which are interconnected through cusmainly in simulations as described later - were designed to mimic tom circuitry. This allows any of the segments of each of imperfect placement by involving a number of orientations (0) the four partitioned electrodes to be individually addressed to of the vessel relative to the tetrapole’s middle axis of symmetry form single-segment variable tetrapoles (SSVT, Fig. 2(c)); or (i.e. the axis connecting the electrodes across their centres). to be grouped so as to just cover the targeted vasculature, thus forming effective localised electrodes to provide more targeted measurements. When all segments of each of the four partiB. Tissue Properties tioned electrodes are shorted they each form an effective stripe The electrical properties of biological tissue are frequency or a pseudo-stripe electrode. The developed prototype DRE dependent and have conductive and reactive elements

[0014] and system was simulated using finite element modelling (FEM) impedance plethysmography measurements have been mostly PAPER 2 16 r I p m o st th m F f an . p y set to 29010 911 54 and 5248 2 res ectivel corres ondin to im in u ch fr sp in

[0096] C at th to have a relatively large area to reduce the electrode-electrolyte interface impedance so as to avoid overloading the current Fig. 3. Simplified lumped impedance model, (a) shows the simplified lumped impedance from the side, (b) shows the simplified lumped impedance looking injection circuitry. The voltage measurement electrodes can be through the electrodes (top view). made smaller — as their interface impedance can be ignored — increasing the measurement precision and the signal-to-noise ratio of measured data

[0022] , Therefore here four stripe elecand blood layers, with respective impedance magnitudes Zs , Zp trode sites A, B, C and D were considered Fig. 2(b), with the and ZB, respectively. A side view is shown in Fig. 3(a), while current injection electrodes A and D having a 2 mm x 8 mm Fig. 3(b) is the top view (from the skin down to the fat layer) of geometry, and the voltage measuring electrodes B and C being current paths along the tetrapole, assuming the blood vessel is 1.5 mm x 8 mm. The length of the electrodes was chosen so as on the axis connecting the electrodes across their centres. to adhere to the dimension limitations of the model - so as to

[0097] The nature of electric / ionic current is to travel between two approximate the arm skin as being flat - and it closely resembled electrodes by taking the path of least resistance, and with the skin electrodes used in

[0023] , Each stripe was subdivided into eight being the most insulating layer, currents will mostly traverse it 0.6 mm long segments separated by a 0.4 mm gap Fig. 2(c). The flowing through it to the more conductive layers below

[0027] , centre to centre distances were 15 mm for electrodes A and D and Currents originating directly above the vessel will cross the skin 10 mm for B and C, resulting to a 2.5 mm distance between each and fat layers almost vertically and will enter the blood, flowing of electrodes A-B and C-D. The electrode spacing was designed across it for a length approximately equal to the distance A-D. to cover adequate length of the blood vessel so as to increase the Currents flowing between electrodes A-D outside of the boundvolume of the blood contributing to the measurements. aries - but in the vicinity of the blood vessel will deviate from their straight geometric path and pass through it. As the length of

[0098] D. Lumped-Impedance Simplified Representation the injecting electrodes extends away from the centrally located

[0099] In order to ascertain the theoretical basis of research question vessel currents adequately far from it will experience a minimum 2, prior to the experimental and the simulation assessments, resistance path by traversing the skin and traversing fat tissue an approximate analysis can be offered through a lumped of equal length to the distance travelled inside the vessel by the impedance representation of the cases involving localised vercentrally located field lines. The corresponding fat impedance sus longer electrodes. It has generally been shown that the magnitude in that case is shown as Zpp. For a specific value impedance of different tissue layers can be approximately repreof ZB, the baseline impedance value ZBoci,aseif localised sented by simplified lumped impedance models

[0024] —

[0026] with electrodes are used, and the minimum corresponding baseline impedance magnitudes at the applied frequency considered. Folimpedance value ZgTRbase when longer stripe electrodes are lowing the geometric model presented previously, and assuming used, are calculated below. The stripe electrodes used in this 9 = 0, Fig. 3 illustrates schematically a simplified lumped calculation are long enough so as to allow n current paths to impedance model for a cross section of juxtaposed skin, fat pass through fat surrounding the blood vessel. PAPER 2 17

[0100] Fig. 4. Circuit diagram, (a) shows the block diagram for the complete system, (b) shows the connections to and from the micro-controller and the switch connection for the electrode segments sites to their corresponding impedance measurement channel.

[0101] Localised electrodes baseline impedance: As none of a, fi, n and ZB are negative (being impedance magnitudes) and although realistically |x| < 20%, it is evident

[0102] ZLOCbase = 2ZS± 2Z p ± Zi; (1) that RLS is greater than unity even for an extreme - theoretical

[0103] Larger stripe electrodes baseline impedance (n paths not - range of |x| < 100% which would result to ZB varying from passing through the vessel): 0 to double its baseline value. That indicates that localised elec trodes will always exhibit higher sensitivity to blood impedance

[0104] ZsTRbase = 7 / - _ 1 _ i_ _ n _ 7 \ (2) variations relative to their larger counterparts and thus research question 2 is theoretically justified by the above model. because ZsTRbase is the parallel combination of Z / yjo, and n parallel branches of 2Zs+Zpp. E. Experimental Setup

[0105] In order to examine which electrode type is more sensitive to blood variations, ZB is assumed to change by x, with pulse 1) Circuit Design: The DRE system was designed to related blood impedance variations typically amounting up to function according to the block diagram shown in figure 4(a). ±20% of the baseline value [4],

[0028] , The resulting variations to A printed circuit board (PCB) was designed featuring the the overall impedance between A-D using localised vs larger four segmented electrode sites, as illustrated in Fig. 2. The stripe electrodes are calculated below. Localised electrodes board was fitted with a 3D printed rectangular wall attached impedance, using 2Z^ ± 2Zp = a, for clarity: around the electrodes using epoxy glue, allowing the use of liquids on the electrodes without shorting any circuitry.

[0106] ZLOC = ct ± ZB(1 ± x) (3) The electrode addressing electronics were positioned on the

[0107] Stripe electrodes impedance, using 2Z$ + Zpp = / 3 for bottom layer of the same PCB, in order to reduce the distance clarity : between the electrodes and the addressing circuitry. This served to minimise the influence of parasitics and noise in the

[0108] ZsTR — impedance measurements and to further protect the latter from any accidental spillage of liquids. The addressing circuitry was developed for interconnecting and individually addressing the

[0109] Each configuration’s sensitivity can be derived by the respec32 segments of the DRE system and for interfacing the ones tive partial derivatives with respect to x

[0029] , which respectively selected from each group A, B, C or D with the respective result to: four channels of a tetrapolar impedance analyser (Key sight dZLOc = ZB(5) E4980 A / AL Precision LCR Meter

[0030] ). The requirements on the circuit, for real diagnostic use rather than for the static dZSTBZBP gelatine phantom scenario examined here, included minimal and — - -2

[0110] (6) ax (n(l ± X)ZB ± (3 ± no)2impedance contribution when active, fast switching and stable frequency response at the chosen carrier frequency. Fig. 4(a)

[0111] The ratio RLS of localised to stripe sensitivities yields: shows the block diagram of the system which was implemented

[0112] , p r e (nZB(l ± x) ± / 3 ± na) through the circuit diagram shown in Fig. 4(b). Switching and RLb = - -5 -2

[0113] (7) between segments was carried out using four ADG714 PAPER 2 18 C e r d s between 1 and 8. An example of this is shown in Fig. 2(c)

[0114] Fig. 5. Gelatine phantom comprising skin (yellow), fat (white), and blood where segment pair A± B\ and its combination with all the Cj Dj vessel (red). segment pairs is shown with dashed-red lines. A sinusoidal 10 kHz, 100 pA current was injected between Ai and Dj with Bi and Cj measuring the resulting voltage. The magnitude of the eight-analogue-switch ICs, featuring low on resistance (Ron = voltage between Bi and Cj was divided by the injected current 2.5 fl) and requiring about 28 ns each to go from OFF state to magnitude to give the value of | Zij | between the electrodes Bi ON and then OFF again. The ADG714 ICs were labelled ‘Ul ’ and Cj . Each of the 8 Ai Bi pairs was combined in turn with each to ‘U4’. To change the segment allocation, an 8 bit binary word of the 8 CjDj pairs (dotted lines in Fig. 2(c)). The outcome of needs to be sent serially to the ADG714 IC, each bit of the byte the sequence was thus an 8 x 8 matrix of 64 \Zij | values. corresponding to one switch in the IC. In order to send the byte When a scan is preformed using SSVT, the length of the active of switching data, the address pin of the desired IC needs to tetrapole varies. Even in a homogeneous volume this variation be held low and then high again to implement the switching is the main factor that introduces non uniformity between the data. The ADG714 ICs were controlled by a PIC18F26K22 impedance values measured by each tetrapole [5], with other micro-controller, with the switching data transmitted through factors including boundary conditions and the fact that in this the serial peripheral interface (SPI) pins, allowing them case the recording electrodes are not always in line with the to connect each of the electrode segments individually current injection ones. In order to remove that scaling effect, or in combinations to the respective LCR meter channels each impedance value obtained from a SSVT tetrapole was (Ii , Vi , V2, h)- The user could select the segment combinations divided by the length of that tetrapole. This approximate process through switches activating pre-programmed sequences in significantly reduced the non-uniformity from the results. Subsethe micro-controller. The pre-programmed sequence is then quently, the values in the | Z^ | matrix were normalised between 0 triggered using push-buttons. Three main sequences were and 100% and values lower than 10% of the resulting impedance pre-programmed into the micro-controller; the first sequence range of the matrix were used to identify the electrodes closer shorted all the segments to form the pseudo-stripe electrodes, to the vessel. the second sequence was the SSVT scanning sequence and the For research question (2) impedance measurements were then third sequence was the localised electrode sequence. performed to the five phantoms P2a-e, first activating only the

[0115] 2) Gelatine Phantom Design: The experimental setup feaelectrode segments from each group corresponding to lowest tured gelatine phantoms constructed in accordance to the prevalues, grouped to form the largest possible localised viously described biological model, with the aforementioned electrode covering only the area of interest. Each measurement dimensions, featuring skin and fat layers, each with a cylinder was repeated with pseudo-stripe electrodes. Between each mearepresenting a blood vessel embedded in the fat layer (fig. 5). surement sequence the electrodes were cleaned using ethanol The phantoms were created using 3D printed moulds in which to maintain good contact across the experiments and to remove gelatine solutions with conductivities corresponding to each any residue leftover from the previous experiments. respective layer were poured successively. The skin side of the Pseudo-stripe electrode (i.e. using all AiBi and CjDj elecphantoms was attached to the electrode PCB. To address research trode segments) measurements were also carried out on saline question (1), two phantoms Pl and P2 were constructed with solution, poured inside the 3D printed wall over the electrodes so their respective blood vessel cylinders placed at orientations of as to occupy the same overall volume with the gelatine phantoms. 20° and —20° relative to the axis passing across A-D through the The conductivity of the saline used was 1.55 S / m as measured array’s centre Fig. 2(b). These values represented the maximum using Model 3540 pH / conductivity meter. These measurements angles that allowed for the vessel to still be fully within the were carried out for comparison with FEM simulations also boundaries of the electrode area. To assess research question featuring a saline. (2), five phantoms P2a-e, (variants of P2) were made with blood conductivity values between 700.04 mS / m (P2a = P2) and 1400.08 mS / m, in steps of 25%. F. FEM Simulations

[0116] 3) Experimental Procedure: SSVT Sequence and Localised Manufacturing phantoms controlling the electrical properties Measurements: For research question (1) and for each of the and accurately changing the orientation of the blood vessel phantoms Pl and P2 described above, an impedance measuring was an elaborate time-consuming process. Implementing the sequence using DRE in SSVT mode was first carried out in different orientations in simulations would be quicker and would order to find out which electrode segments were closest to allow a more precise control over different parameters. Finite the blood vessel. In each measuring sequence the injecting Element Modelling (FEM) simulations using the COMSOL 5.3 electrode segment from group A was always combined with the electric currents (ec) module under the AC / DC physics were neighbouring measurement electrode segment from B, while the first performed using SSVT on a saline parallelepiped model PAPER 2 19

[0117] Fig. 6. Scaling and Normalising data, (a) Saline scan data from simulation, (b) Electrode tetra pole length, (c) Normalised and scaled data.

[0118] (conductivity of 1.55 S / m as measured previously), with identiIII. RESULTS AND DISCUSSION cal dimensions to the geometric model and the phantom. In the Both experimental and simulation results are presented and ec module, quasi-static solutions of Maxwell’s equations were discussed for each research question. used to solve for the electrical potential. No electrode-electrolyte interface was included in the model. Electrode site A was assigned as a current terminal with a 1 A current. Electrode site A. Research Question (1) D was assigned as a ground terminal of zero potential. It was 1) Factoring in the Weighting Factor for Different Size assumed that no current flows through the exterior boundaries Tetrapole Formations: Applying SSVT to the FEM saline-only of the phantom, as the exterior boundary was air which is a volume resulted in a \Zij | matrix ZssAL which was then divided poor conductor of electric current. Electrodes B and C were by the length Lt of each corresponding tetrapole, i.e. the outer assigned as boundary probes used to obtain the resulting poedge-to-edge distance between the current injecting electrodes, tential, which was then used to obtain the transfer impedance. to remove the electrode spacing influence from the measurement These simulations were performed to observe the impedance (Fig. 6). This resulted to normalised values that were within 3% sensitivity distribution in a homogeneous volume and to verify of the average scaled value of the scaled ZSSAL dataset. that division by the tetrapole lengths allowed each data point to 2) Scanning and Localisation: Fig. 7(a) and (b) illustrate be scaled to the same level Fig. 6(c). Moreover these simulations the data from the SSVT simulations for 0 = —20° and 0° allowed for a performance comparison between the simulated respectively. The scaling procedure was applied to all SSVT stripe electrodes and the simulated as well as the experimental scans for both the experimental and the simulation results. The pseudo-stripe electrodes. scaled results were then normalised to a 0% — 100% range and by

[0119] Further assessing research question ( 1 ), simulations were then applying a 10% threshold the desired segments were identified carried out for eight models directly reproducing the gelatine as shown by the examples in Fig. 7(c-f). phantom described previously, with blood vessel cylinder orienAs the normalised figures indicate, the threshold can be used tations of 0 = -20°, -10°, 0°, 10°, 20°, 30°, 60° and 90° with to identify the electrode pairs that correspond to the lowest respect to the longitudinal axis of symmetry of the electrode impedance values, i.e. to the area over the blood vessel (top array. The values of ±20° were chosen so as to verify that the view). The ij combinations that are below the threshold are model produced similar results to the experiments. For each shown in Fig. 8 for 0 = —20°, together with the ij pairs ge0 simulations of SSVT sequences were carried out that led ometrically located right above (top view) the vessel labelled to estimates of the values of the \Zij\ matrix for each model, as above vessel segments. The outcome indicates a very close thus allowing for the selection of the electrode segments that match of 85% with 12 / 14 segments matching. For all cases of 0 were closer located to the blood vessel in each case. The \Zij | considered, Table I summarises the comparison between all of matrix was then divided by the lengths of the related tetrapoles the ABi-CDj pairs found to exhibit the lowest | Zij | values after to remove the electrode spacing influence from the measurescaling and normalising the results versus those located over the ment. By applying a 10% threshold from the overall change in corresponding blood vessel. In most cases the results are very the impedance value, the desired segments satisfying research closely matched with the ij pairs that are geometrically located question (2) could be identified. The 10% threshold was selected right over the vessel. empirically and it worked for all cases. In non-controlled measurements, plotting the angle 0 of each

[0120] To assess research question (2) simulations were carried out SSVT orientation <f> vs the scaled normalised data provides an for 0 = —20° with the blood conductivity being varied from its approximate value for the angle 0. This is shown in Fig. 9 for the original value (700.04 mS / m) that corresponded to the exper0 = —20°, 0°, 20° cases where the lowest \Ztj \ values exhibit iments up to twice that value (1400.08 mS / m), in increments minima at respective f = —16.5°, —3.4°, 0°, 3.4°, 16.5°. These of 10%. Moreover, uniquely in the simulations, an additional values are within 83% and thus approximately near the 0 cases calculation was carried out for each conductivity variation step considered. of the blood vessel using stripe and pseudo-stripe electrodes, in The scanning and localisation function of the system have order to establish whether they perform similarly. been successfully demonstrated using the idealised model used PAPER 2 20 esults

[0121] (e) normalised experiment -20° (Q normalised experiment 4“20sscan scan

[0122] F s normalised simulation and experimental scan data for (—20°, —10°, 0°, 10°, 20°, 60°, and 90°) orientation data, with a threshold of 10% of the maximum Fig. 9. SSVT orientation <j> vs the scaled normalised data. impedance. measured with an LCR meter and it was found to be 122 fi. The here. It is worth mentioning that it does not take into consideradifference of about 7 Q with respect to the simulated value was tion the anisotropic properties of muscle layer and the insulating attributed to the resistance introduced by the switching circuit. properties of bone, or the influence of neighbouring vascular

[0123] It is expected that an increase in blood conductivity will structures. result in a decrease of the total skin-fat-blood vessel impedance, and vice versa. Fig. 10(a) illustrates simulation results of the

[0124] B. Research Question (2) aforementioned blood conductivity variations worked out for

[0125] 1) Pseudo-Stripe vs Full Size Stripe: The simulation results stripe versus pseudo-stripe electrodes. The results are similar on saline solution with stripe and pseudo-stripe electrodes within 0.01% for all concentration steps, highlighting further demonstrated that the impedance for both types of electrodes the similarity in performance between full solid stripe electrodes was very similar: the simulated impedance for the stripe and and the grouping of segments covering a similar outline area. for the pseudo-stripe electrodes for saline was 114.88 and 2) Monitoring Blood Variations With Localised vs. Pseudo115.41 Q, respectively. The small difference of less than 1% Stripe Electrodes: Fig 10(b) illustrates the performance of lobetween the two values leads to the conclusion that the pseudocalised electrodes comprising the selected ij pairs for 0 = —20° stripe electrodes emulate satisfactorily their equivalent stripe in the previous section, compared to pseudo-stripe electrodes electrodes. Experimentally, the saline solution impedance was in detecting such impedance changes, both with simulations PAPER 2 21 ventional methods and in the future such a system could be

[0126] Fig. 10. Impedance vs conductivity plot for simulation and experimental made portable, taking advantage of the latest state-of-art bioresults, using different types of electrodes to measure impedance while changing impedance ASICs

[0031] —

[0033] . Still there are some areas in deconductivity of blood from 100% conductivity to 200% conductivity. signing and implementing such a system for diagnostics than require attention. In terms of electronics, the interface circuits will be influenced by varying the effective electrode area which and experimentally. The impedance measurement using the will change the contact impedance. Thus the injecting electronDRE localised electrodes right above the blood vessel is not ics output impedance should be designed accordingly. Other the same as that monitored by the pseudo-stripe electrodes. potential issues include influence by neighbouring vasculature The localised impedance measurements exhibit a larger decrease and an increase in complexity in the circuitry relative to other as a function of blood conductivity than when pseudo-stripe impedimetric systems. In terms of application areas, the DRE electrodes are used. A similar behaviour, albeit with a smaller will mainly be used for plethysmography to extract flow and relative difference is observed in the simulation results. The volume information from arteries that are relatively solitary and overall conclusion from these measurements is that, localised in the extremities. electrodes exhibit significantly higher sensitivity to blood variation. In fact absolute impedance change for each concentration step change is 25% higher for localised verses pseudo-stripe electrodes in simulations, while the corresponding experiment results exhibit a 22% — 28% improvement. While this is a smaller variation than the one demonstrated in the theoretical lumped impedance model, the electrode length assumed in that model would have been significantly larger for the field lines to pass straight between the injecting electrodes. Therefore it is commendable that, based on the achieved outcome, the localised electrodes exhibit higher sensitivity even when compared to relatively smaller stripe electrodes such as the ones examined here. This indicates that the performance improvement over significantly larger electrodes (eg: band electrodes) will be even more pronounced. PAPER 3 22

[0127] A Quadruple-Sweep Bioimpedance Sensing Method for Arterial Stenosis Detection

[0128] I ABSTRACT Current carotid atherosclerosis diagnostic protocols do not feature techniques that would allow for early or frequent medical examinations, leaving a significant number of asymptomatic carotid stenosis cases undetected and often leading to strokes. In this work we used idealised biological models to demonstrate a new rapid, potentially inexpensive and operator-independent diagnostic method, aimed at detecting whether a stenosis exists, rather than seeking to accurately quantifying or localising it. An array of electrodes was used to obtain sequential bioimpedance values over the skin, through a novel scanning technique, covering an area over the artery of interest. FEM simulations, verified through in-vitro experiments on gelatine phantoms, were used to validate the method. The final results, obtained through image processing algorithms, were in the form of planar bio-impedance maps and were successful both in identifying arterial features and detecting the presence of stenoses of different sizes. The results could also be used to indicate the artery’s relative orientation to the sensor, eliminating the need for manual alignment by a specialist operator. Therefore, this method shows promise for routine medical examination, either in primary care, or even at home, to indicate whether a patient would require further, more detailed examinations at a specialist clinic.

[0129] I INDEX TERMS Bioimpedance sensing, Impedance scanning, Cardiovascular Diagnostics, Planar bioimpedance mapping, FEM modelling, Image processing, Carotid artery screening, Stenosis detection

[0130] I. INTRODUCTION rotated by a specialised operator, visualising the structure of

[0131] AROTID artery disease can have devastating conseinterest from different angles [5] (Fig. 1(a)). However, the quences, as it is one of the main causes of stroke. Stroke quality and reliability of the outcome strongly depends on affects 15 million people around the world annually, with the expertise of the person handling the probe, making the more than 60% of all cases caused by carotid atherosclerosis, process highly operator-dependent. This results in up to 18% 80% of which are asymptomatic [1], [2]. The resulting arteof the diagnoses being erroneous, either as false positives or rial stenosis or occlusion can become symptomatic in people false negatives [1], [6]. This operator dependency coupled over 40 years of age, while asymptomatic stenosis may occur with the cost of a full ultrasound imaging system makes at an earlier age [3]. the technique suitable for use predominantly in specialist

[0132] To date, ultrasound imaging [4] is the gold standard for clinics [7], being carried out only when clear symptoms non-invasive detection of carotid artery stenosis once sympoccur [1], which is often too late. More importantly, these toms appear, referring positive outcomes for further tests issues preclude ultrasound carotid scanning from being part based on CT or MRI scans. The method is almost always of routine medical checks. As a result, for the asymptomatic used by means of a hand-held probe, which is moved an majority of patients [1], such a scan will not take place at all PAPER 3 23 before a potentially life changing or fatal stroke occurs. tetrapole and the EIT approaches have been described in

[0133] In order to carry out regular carotid screening tests to

[0011] ,

[0023] . For the desired method, topological information every person meeting certain criteria (e.g., age, diet), it is is needed to detect the outline of the artery, in terms of desirable to have a low cost, operator-independent, non- position and orientation, to avoid the operator dependence invasive method to provide an indication of arterial stenosis of manually re-positioning the sensor. While the presence of or occlusion without waiting for symptoms to occur. Such a features, such as arterial bifurcations and possible stenoses, method would not aim at replacing ultrasound, but rather it need to be detected in order to generate an indicative diwould allow for suspected cases to be referred for ultrasound agnostic outcome, there is no need at a routine screening examination and subsequent further tests, if needed. It would test to extract a detailed image involving elaborate hardware, ideally be available in primary care, as well as low resourced algorithms and forward models. healthcare settings and even for home use. Enabling routine The work presented here proposes and assesses a multitests on all patients that visit a doctor could potentially allow electrode sensing method using bioimpedance measurefor preventive action for a great number of stroke sufferers ments to provide what we introduce as localised planar annually [1]. bioimpedance mapping (PBiM) of transcutaneous vascular

[0134] A sensing modality that would allow for a point-of-care features. A possible neck-worn realisation is shown in Fig. (PoC) diagnostic device to satisfy the above criteria is bio1(b), along with the methodological steps followed in this impedance (or electrical impedance). Bioimpedance has been research. While - in a similar fashion to EIT - we use used in a variety of medical applications, ranging from carimaging techniques to extract the aforementioned data, there diovascular plethysmographic measurements [8]—

[0011] , lung is no requirement to use a forward model or computationally monitoring

[0012] ,

[0013] , detection of breast cancer

[0014] , mapdemanding reconstruction algorithms. ping wound areas in chronic ulcers

[0015] and many others More specifically, in this work an electrode array is used

[0016] ,

[0017] . Bioimpedance systems are inexpensive, easy to to carry out a novel bioimpedance scanning sequence, both use without specialist training, providing rapid outcomes by using finite element modelling (FEM) simulations and exwithout the use of harmful ionising radiation [8]. While perimentally on gelatine phantoms, to detect non-invasively measurements can be prone to errors due to the surface the position and orientation of a blood vessel underneath electrochemistry of the electrodes used, this can be overcome the skin, thus eliminating any need for sensor alignment through the use of tetrapolar (four-electrode) configurations that would require a specialised operator. The associated

[0018] . In a tetrapolar setup, measurements are carried out by bioimpedance images are processed for blood vessel features, applying a small amplitude alternating current using a pair of including bifurcations and, in particular, the presence of electrodes placed on the skin and then measuring the resultstenosis. ing potential using a second pair of electrodes

[0017] . Recent The novelty of this approach lies in the aforementioned studies have shown that it is possible to detect atherosclerotic properties, which allow for portability and minimal operator plaques through tetrapolar bioimpedance measurements

[0019] , dependency. To the best of the authors’ knowledge, there is making this sensing modality a possible candidate for early no state-of-the-art solution for planar mapping of the carotid detection of carotid stenosis. artery. PBiM utilises tetrapolar bioimpedance interfacing. It

[0135] However, single tetrapole arrangements ideally need to features further innovative aspects in the methodology, as it be aligned to the targeted artery, thus requiring the latter’s interrogates tissue through tetrapoles of fixed inter-electrode location and orientation to be known in advance. Failing distance, as opposed to EIT, where the tetrapole dimensions to place and size the electrodes accordingly, degrades meavary. In this manner, the measurements are directly comsurement sensitivity due the surrounding tissue’s impedance parable without the need for a forward model. Moreover,

[0011] . Moreover, the approach of analysing plethysmographic given the fact that a tetrapole provides measurements over waveform features for indications of stenosis presence, while the underlying tissue volume rather than point measurements valuable, does not reveal any topological information. Such at the tetrapole centre

[0017] , we introduced a novel scanning information can typically be offered by imaging, with elecmethod, whereby an area is sweeped four times in each of trical impedance tomography (EIT)

[0018] meeting several of four different directions (i.e., Vertical, Horizontal, Diagonalthe above criteria. However, the method is too cumbersome, right and Diagonal-left). The four resulting images, or biocomplicated and expensive for routine medical examinations impedance "planar maps", are combined so as to achieve as, in its most portable embodiment, it requires an electronic higher spatial resolution and less dependency on the underlysetup similar to a "tower" desktop computer on wheels, making blood vessel’s orientation. A COMSOL model was used ing it unsuitable for the majority of primary care practices, to assess the method, while experimental verification was remote areas, or home use

[0020] ,

[0021] . Additionally, for EIT- performed on a smaller physical model (gelatine phantom). based carotid artery scanning, electrodes would need to be In section II, the procedure for setting up the simulation applied around the neck of the patient and, given its poor and performing the experimental studies with four variants of spatial resolution

[0016] ,

[0017] ,

[0022] , it would not offer useful a biological model and a multi-electrode array is discussed. information about a localised structure, such as the carotid In particular, the parameters, i.e., vessel orientation, bifurcaartery. Solutions that include elements from both the single tion angles and percentage stenosis, used to assess the per- PAPER 3 24

[0136] FIGURE 1: Carotid artery scanning using: (a) A standard sweeps or "scans" of impedance measurements by moving ultrasound probe, (b) The proposed - operator independent fixed-size electrode tetrapoles in four different directions, - planar bioimpedance mapping method. In this work, the using an electrode array placed over the blood vessel. biology is represented by a biological FEM model and a The simulations were carried out using the COMSOL 5.3 gelatine phantom. electric currents (ec) module, where the quasi-static solutions of Maxwell’s equations were used to solve for the electrical potential. The electrodes placed on the top surface were used formance of the method, are described. Section III presents as current injection or recording terminals. As biocompatiand discusses the results obtained from the application of bility and electrode corrosion are issues in simulation, the the method. Finally, section IV presents the conclusions and simulated injection current was kept at 1A for simplicity. avenues for further work. Adaptive physics controlled meshing at the finer setting was used, where the mesh density was kept very high at locations

[0137] II. METHOD of interest, i.e. the vascular features.

[0138] A three-dimensional biological model was designed to calA cuboid shape model was considered with two layers, a culate electrical impedance sensed by electrodes placed on 2 mm thick skin layer and an 8 mm thick fat layer, where the top surface, representing the skin. Aiming to explore the the carotid artery was considered to be in the fat layer (Fig. principles and merits of planar mapping and similar to other 2(a)) and the anatomical properties of the model were kept studies (e.g.

[0011] ) this model was simplified in terms of its consistent with

[0024] ,

[0025] . The typical depth of the carotid composition and its geometrical complexity, as it was limited artery in the neck from the outer skin layer is (6 ± 3) mm to three homogeneous tissue layers (see section II- A) without

[0026] , and in the model the depth was taken as 7 mm. For any other vasculature in the vicinity. Two versions were the sake of model simplicity, it was assumed that the carotid realised, as described later: an FEM model and a gelatine artery is embedded in a fat layer underneath the skin and phantom for experimental verification. An array of electrodes no muscle tissue is present between the skin and the arterial was used with the active tetrapole being dynamically rewall. It was also assumed that all media are homogeneous configurable in terms of position and orientation, in resemand that the top surface of the skin is flat. The conductivity blance to

[0011] , modelled to assess local impedance variations and relative permittivity values assigned to the skin, fat and corresponding to vascular structures underneath the skin. blood layers were obtained from

[0027] ,

[0028] . The first o bjective w as t o c heck w hether i t i s p ossible to Specifically, the conductivity values for skin, fat and blood detect the presence of a blood vessel and subsequently, to were set to 2.937 mS / m, 42.954 mS / m and 700.04 mS / m, detect its orientation, towards the goal of making the method respectively, while the relative permittivity values were set operator-independent by eliminating the need of manually to 29010, 911.54 and 5248.2 respectively, at a frequency of aligning the sensor to the vessel. The next objective was to lOKHz, calculated using Eq.(l)

[0028] ,

[0029] : assess the possibility of detecting the presence of a stenosis within the blood vessel from the electrical impedance scan. (1) Blood vessels with and without stenoses were studied, both of straight-line and bifurcated shapes. The latter was included In the above equation E (u>) is the complex relative perto represent the most usual region of atherosclerotic plaque mittivity as a function of frequency, is the permittivity PAPER 3 25

[0139] FIGURE 2: Overview of the biological model with the 20 x 30 electrode array used for the FEM simulations on top: (a) 3D COMSOL model, (b) Top view of the —37° straight artery model with stenosis shown as a yellow region in the straight vessel. The smaller group of dark-coloured electrodes in the middle of the main array represent the 8 x 16 sub-array used for the experimental verification, (c) Top view of the bifurcated artery model with stenosis shown as a yellow region in one branch of the bifurcation. at frequencies, where LOT > > 1, and T represents a time variations in this study can be categorised into the following constant. The magnitude of the dispersion is given as Asm, four test cases: ai is the static ionic conductivity and EQ is the permittivity of

[0140] • Case I: Straight vessel without stenosis. free space. The distribution parameter, a, is a measure of the

[0141] Parameter variations: Orientation = {0°, —15°, —37°, broadening of the dispersion. —45°, —90° }. For each of the orientations, the simula¬

[0142] Atherosclerotic plaques within the blood vessel comprise tion was carried out for five slightly different variants in fat deposits and calcium hydroxyapatite (CHA, commonly the vessel’s position, in order to test whether the scanreferred to as pathological calcification), which can make up ning method and the image processing algorithm could to a maximum of 30% of the total stenosis

[0030] . Therefore, still detect the vessel with some degree of randomness the conductivity and relative permittivity of fat

[0028] and those added to the blood vessel parameters. The five (x,y) of CHA at room temperature

[0031] at 10kHz were taken and position coordinates of the vessel’s centre in relation combined in the ratio of 70 : 30 to obtain the electrical to the initial central position were (-2 mm,0), (0,0), properties of the stenosis to be used in our simulations. (2 mm,0), (0,-2 mm) and (0,2 mm), with the vessel’s Considering this, the conductivity and relative permittivity of diameter fixed at 5 mm. Subsequently, with the vessel the stenosis were set to 1.4 (mS / m) and 280, respectively. at its original position, four different diameters were

[0143] The dimensions of the COMSOL model were 100 mm x considered with values of {5.5 mm, 6 mm, 6.5 mm, 7 60 mm x 10 mm with a 2 mm thick skin layer at the top mm] resulting in a total of 45 simulations. and an 8 mm fat layer underneath the skin. The blood vessel • Case II: Straight vessel with stenosis. was placed in the fat layer with its centre 7 mm below the The percentages of stenosis were {0%, 20%, 40% 60%, top surface of the skin. By changing various parameters, it 80%}. For each of the stenoses, the same nine total was possible to change the location, orientation and diameter variations in the vessel’s location and diameter as in case of the blood vessel. Both straight (Fig. 2(b)) and bifurcated I were considered, while keeping the vessel orientation (Fig. 2(c)) blood vessels were considered. It was also possible at —37°. A total of 45 simulations were considered. to vary the size and the location of an atherosclerotic plaque • Case III: Bifurcated vessel without stenosis. stenosis within a blood vessel. The stenosis was modelled as Again, as in case I, five variations of location and an ellipsoid with radii of riengthX rwidthx rheight. Using four variations of diameter were considered, keeping the top view of the model (Fig. 2(b)), its centre was placed the orientation of the main vessel at 0° and the angle on the vessel-fat boundary at the middle of the blood vessel’s between the bifurcated vessels at 73°, bringing the total length on the side facing the bottom left of the electrode array. number of simulations to 9. The depth of its centre was 7 mm and its length riength was • Case IV: Bifurcated vessel with stenosis 6 mm, parallel to the vessel’s longitudinal axis. Its height Five variations of stenosis, i.e., {0%, 20%, 40%, 60%, rheight = 3 rnm was vertical to the electrode plane. Its radius 80%} were simulated. For each of the stenoses, five along its width was = Ddvx Xocc, where Ddvis the variations of the blood vessel’s location, and four varidiameter of the vessel in mm and Xoccis the stenosis size as ations of its diameter were considered, while keeping a percentage of the vessel diameter. the orientation of the main vessel at 0° and the angle

[0144] In the simulation, the top surface was considered as the between the bifurcated vessels at 73°, resulting to 45 xy plane and the depth was along the z axis. All parameter simulations. PAPER 3 26

[0145] B. SIMULATION SCANNING SEQUENCE C. EXPERIMENTAL VERIFICATION

[0146] All simulations featured "shifting" tetrapoles formed by inThe experimental part of the work was carried out to validate line neighbouring electrodes, with the outer pair always the simulation outcomes. One of the straight vessel orienbeing the current-carrying (CC) electrodes, and the inner tations of Case I, namely —37°, was replicated in gelatine. pair assigned the voltage pick-up (PU) function. One of the Generating a more sophisticated system beyond expanding CC electrodes was modelled as a current terminal with a the slightly the one presented in

[0011] was beyond the scope boundary condition of 1 A, while the other was modelled as of this work and thus due to circuit complexity constraints a ground terminal of zero potential. It was assumed that no it was deemed adequate to confirm the simulation findings current flows through the exterior boundaries of the phantom, using a smaller array and a smaller phantom. Therefore a modelled as air. The two PU electrodes were modelled as downsized version of the model was simulated and a matchboundary probes, where the resulting potentials were meaing experimental model was generated. They both used a sured. The transfer impedance was then obtained by dividing smaller 8 x 16 electrode array (termed elsewhere in this text the difference of the voltage values at the two boundary as the "sub-array") which was just enough to achieve a vessel probes by the applied current. A 20 x 30 electrode array was orientation identification and thus to allow for a comparison used in the FEM simulations (Fig. 2). of simulations and experiments. Six identical phantoms were constructed with dimensions 30 mm x 50 mm x 10 mm

[0147] FIGURE 3: Four directional scanning. The vertical, horizontal, diagonal left and diagonal right scanning sequences are indicated in green, red, blue and magenta outlines and arrows, respectively. FIGURE 4: Gelatine phantom used in experiments with different layers designated, placed face-down on the electrode

[0148] Fig 3 illustrates a 4 x 4 segment of the full array to viarray PCB. sualise the four scanning sequence orientations, i.e., vertical, horizontal, diagonal left, and diagonal right scans. Tissue conductivities were calculated at lOKHz using Eq. (1) and values taken from

[0028] . The conductivities of the

[0149] The vertical scan (red outline in Fig. 3) started with the corresponding phantom layers were matched to those of the first f our h orizontal e lectrodes o f t he a rray. After obtaining respective biological tissues at lOKHz, by calculating the the transfer impedance, the tetrapole was shifted by one row required NaCl concentrations through Eq. (2) and according at a time until the last row, before shifting by one column and to the work in

[0033] , similar to the work in

[0032] —

[0034] : repeating the sequence until the bottom right four electrodes in the last row were reached. In a similar fashion, rotated by a = 200 X — MNaCl

[0150] (2) 90°, the horizontal scan (green outline in Fig. 3) started with » Solution the left-top vertical tetrapole, shifting to the right, until evenwhere a (S / m) is the target conductivity of the layer in S / m, tually reaching the bottom four electrodes of the rightmost Mj^aCi is the mass of required NaCl in grams and Vsoiution column of the array. Each of the diagonal left and diagonal is the volume of the solution in ml. right scans (blue and magenta outlines respectively in Fig. Table 1 lists the resulting NaCl concentrations used for 3) started with first p ossible f orward s lash a nd backslash making the gelatine phantom layers. A PCB with a planar diagonal tetrapoles at the upper left and upper right corners respectively. As in the previous scans, the transfer impedance TABLE 1: NaCl Concentration of the phantom layers was obtained for all similar tetrapoles in the array. Layer Layer NaCl

[0151] The four scanning sequences were used in all simulation Conductivity (S / m) Concentration (g / 100 ml) steps, using the full 20 x 30 electrode array. For each step, the ~ Skm - 0003 - 00014 - resulting impedance values of the overlapping areas covered Fat 0.043 0.02

[0152] Blood 0.70004 0.325 in the four respective directions were used to calculate the overall impedance of the underlying volume of that area, as described in section II-D. electrode array with 8 x 16 electrodes was fabricated to PAPER 3 T1 experimentally verify the results obtained from a simulation of the electrodes belonging to the skeleton of the vessel model with identical dimensions. The PCB electrodes were was calculated and used to classify them as follows: disk-shaped with the same diameter and spacing as in the electrode is an endpoint simulations. Using instrumentation as in

[0011] , a Keysight electrode is part of the E4980A impedance analyzer was used to obtain the transfer skeleton impedance from each tetrapole, following the same scanning electrode indicates the sequence as in the simulation studies, described in section ise, location of a bifurcation II-B . The analyzer was configured to give real and imaginary (R & I) values at 10kHz. The impedance analyser features This produced a list of candidate bifurcation points four terminals, which are labeled Lcur, Hcur,Lpot and Hpot. along the skeleton of the vessel object, which could The Lcur and Hcur terminals are used to inject current and then be used to segment it into main vessel and side were connected to the CC electrodes. The Lpot and Hpot branches. terminals are used to measure potential and were connected (d) Identification of the parts of the object, which belong to the PU electrodes. to the main vessel and side branches was performed by finding the intersection points of the normal vectors

[0153] D. IMAGE PROCESSING FOR VESSEL FEATURE on either side of the bifurcation and the end points ESTIMATION belonging to the skeleton of the vessel, using the vessel boundaries detected in step (b).

[0154] The objectives of the image processing were: a) to combine, (e) The orientation of each segment of the vessel object for each of the cases, the data from the four individual scans was estimated by averaging the orientations of the left in one 2D impedance map so as to detect the vessel, if this is and right vessel boundaries. Boundary orientation was located under the electrode array; b) to calculate vessel oridetermined by calculating the least squares estimates entation relative to the array; c) to perform the identification of the best fitting straight line to the boundary pixels. of features, including vessel boundaries and bifurcation using (f) The bifurcation angle was estimated by subtracting the image processing methods; and d) to detect the occurrence of orientations of the two vessel branches, obtained in vessel stenosis using appropriate information. step (e).

[0155] The data obtained with the scanning procedure set out in (g) To detect the presence of stenosis in a vessel, the k- section II-B correspond to horizontal (0°), vertical (90°) and means++ method

[0038] was used. We ran the experdiagonal (45° and 135°) scanning directions. For all models iments using four clusters, corresponding to image using m x n electrodes (including the smaller sub-array regions, i.e., background, blood-filled vessel region, used in experimental verification), these resulted in arrays external vessel wall and internal vessel wall. The latter with respective sizes of m x (n — 3) for the vertical scan, region includes pixels belonging to the stenosis. The (m — 3) x n for the horizontal scan and (m — 3) x (n — 3) output of the algorithms is a number of connected for each for the two diagonal scans. In order to produce a components, each with a unique label. The occurrence composite image, which combines the information from each of a stenosis is detected by identifying more than one of the scans, the impedance measurement of each tetrapole blood-filled regions within the vessel, or equivalently, was allocated to the (x, y) coordinates of the corresponding the disconnection of the blood-filled region of the two PUf electrodes on the top surface of the volume. In order vessel due to the presence of a stenosis. to reduce the effect of measurement noise, distance-based

[0156] This method was applied on all simulation and experimenweighted averaging was then performed, taking into account tal models. the scanning orientation. The Euclidean distance was used in the experiments, with the corresponding weights being III. RESULTS AND DISCUSSION proportional to 1 for the horizontal / vertical matrices, and x / 2 A. BIO-ELECTRIC FIELD PERTURBATION BY VESSEL for the diagonal ones, respectively. The result of this process,

[0157] The perturbation by the subcutaneous vessel of the biowhich combined together the measurement data from the four electric field generated by each electrode tetrapole is evident scanning directions, is an array of size (m — 2) x (n — 2). in the simulation results shown in parts (a) to (c) of Fig.

[0158] The image processing steps were as follows: 5. To demonstrate the effect, a side view of the model is

[0159] (a) Thresholding was applied to separate the vessel from used with a single tetrapole placed along the top, in the background tissues using Otsu’s method

[0035] middle over the skin layer. In line with the tetrapolar im-

[0160] (b) The vessel boundaries were detected

[0036] , followed by pedimetric theory defining the sensitivity distribution

[0017] , the application of thinning

[0037] , using mathematical i.e. the contribution to the impedance measurement by each morphology operators. This resulted in single-pixel point inside the measurand volume, both the injection and width vessel boundaries. the measuring electrode pairs can be shown as generating

[0161] (c) In order to detect the presence of bifurcations, skelebio-electric field lines. In Fig. 5(a)-(c) the respective current tonization

[0037] was applied to the binary vessel region, density fields are shown in red and blue and a vessel vertical obtained in step (a). The number of neighbours, Nskeito the plane of view is moved from outside the frame to the PAPER 3 28 m v r p s l s e a F T s a m h ) r h e c e ( t t

[0162] B. IMAGE PROCESSING

[0163] (c) Diagonal left scan (d) Diagonal right scan

[0164] Fig. 7 illustrates the image processing results from simulation data for each of the four scans of case I, and orientation FIGURE 7 : Image processing results from simulation data of of —37°: (a) vertical scan, (b) horizontal scan, (c) and (d) each of the four scans for case I, and orientation of —37°: (a) diagonal left and right scans, respectively. It can be seen vertical scan (electrodes located at the left and right edges that the results differ according to the sequence directions. In did not contribute), (b) horizontal scan (electrodes located particular, sequences where the tetrapoles were perpendicular at the top and bottom edges did not contribute), (c) & (d) to the blood vessel axis resulted in larger vessel diameter diagonal scans (electrodes located at the image borders did than those featuring tetrapoles parallel to the blood vessel. not contribute). This suggests that it is not adequate to exclusively rely on the results of a single scanning direction for the study of a straight blood vessel. For a bifurcated blood vessel, informaC. ESTIMATION OF VESSEL ORIENTATION AND tion obtained from a single scan would not be sufficient for EXPERIMENTAL VERIFICATION the entire region. The procedure of combining measurements To evaluate the proposed method’s capability in accurately from the four scanning sequence helps eliminate bias due to identifying the orientation of blood vessels with respect to the geometric relationship between the scanning directions the electrode geometry, blood vessel orientations of 0°, — and vessel orientations, as detailed in section II-D. The 15°, — 37°, —45° and —90° were simulated. Using the PAPER 3 29 l r ite image for gelatine phantom vessel, (d) gelatine phantom

[0165] FIGURE 8: Image processing results from simulation data vessel boundary. for case I, and orientation of —37°: (a) Composite image using the measurements of the four scanning directions. ( b) TABLE 2: Vessel Orientation: Estimated values from the Thresholded image, (c) Vessel boundary. image data vs reference values pr th co m the orientation of —37 shown in Fig. 8(a). Thresholding was then applied, indicating the location of the blood vessel relative to the electrode array, as shown in Fig. 8(b). The shows the composite image for case III without the presence extracted orientations for simulated vessels at different oriof stenosis. Applying thresholding to this image results to entations are given in Table 2. It is also noteworthy that when Fig. 10(b), while the vessel edges are shown in Fig. 10(c). the vessel orientations are 0° , — 45° or —90°, the angles As it can be seen, increasing the size of the electrode array can be estimated with minimal errors. For other angles, the results in better identification of the bifurcation. This shows extracted angles are within ±3° of the actual angles. that, with a sufficient number of electrodes, straight vessels, as well as vessels with bifurcations, can be identified using

[0166] 1) Experimental Verification the method described in this work.

[0167] The detection of a blood vessel with an orientation of —37° was experimentally verified through measurements on six E. DETECTION OF STENOSIS gelatine phantoms fabricated through the same procedure. The detection of stenosis was studied for two cases, i.e., case The composite image and the resulting detected boundaries II (stenosis in a straight vessel) and case IV (stenosis in a of the vessel at the orientation of —37° in the simulation vessel with bufurcation). studies using the smaller model and the electrode sub-array are shown in Fig. 9(a,b). Fig. 9(c,d) show the composite 1) Straight Vessel with Stenosis image resulting from the experimental measurements and The results for the straight line vessel with the presence the resulting extracted boundaries of the vessel at the same of stenosis, obtained from the simulations were processed orientation. We can observe that the results of the image using the procedure described in section II-D to evaluate the processing algorithms on both simulated and experimental ability of the proposed method in detecting the occurrence data are similar. Using the image processing algorithm the of stenosis. A number of stenosis scenarios were considered, extracted angles from the gelatine phantoms resulted to an specifically, 0 % (no stenosis), 20 %, 40 %, 60 % and 80 average value of —37.37° ± 3.09°. Thus, it was conclusively %. Fig. 11(a) shows the image obtained after processing, demonstrated that the relative orientation of the blood vessel with the slightly lighter gray area within the vessel boundary with respect to the electrode array can be identified using the indicating the stenosis region, for the reference value of 40 method described in this study. %. The estimated stenosis area (shown in Fig. 11(b)) was obtained by following the method described in section II-D,

[0168] D. VESSEL FEATURE DETECTION step ’g’. Table 3 provides stenosis detection accuracies for

[0169] The procedure described in II-D for case III resulted in the different levels of stenoses in the case of straight vessels, images shown in Fig. 10 for the bifurcated vessel. Fig. 10(a) and shows that stenoses can be detected with accuracy of PAPER 3 30 d e e d 4 provides stenosis detection accuracies for different levels of

[0170] FIGURE 10: Image processing results from the simulation stenoses in the case of bifurcated vessels, and shows that data of case III and orientation of 0° for the straight part of the stenoses can be detected with an average accuracy of 91.11 % . vessel structure (no stenosis present): (a) Composite image As in the previous case, the results demonstrate that stenosis using the simulation results of the four scanning directions, detection accuracy drops to 66.67% and 88.89% in the case (b) Thresholded image, (c) Vessel boundary. of 20% and 40% stenoses, respectively, i.e., low degree stenosis.

[0171] 9 cu (d re w th a s e , e s enos s area s c ear y v s e w e same co oura on as

[0172] FIGURE 11: Image processing results for case II at an the inner vessel boundary. orientation of —37° and reference stenosis percentage of 40% (simulation data): (a) composite image (stenosis barely visible as slightly lighter grey inside the vessel boundaries), IV. CONCLUSIONS (b) segmented image showing background, inner / outer vessel In this work, a new planar bioimpedance mapping method boundaries and blood filled region. The stenosis area is was presented, which is based on a fourfold scanning proceclearly visible with the same colouration as the inner vessel dure using a planar electrode array. The proposed approach boundary. was studied through FEM simulations using a simplified biological model consisting of skin and fat layers and blood vessels with and without bifurcations and stenoses. Experi¬

[0173] TABLE 3: Accuracy of stenosis detection in a straight line mental proof-of-concept was carried out for one of the cases vessel (out of 9 scans) examined through the use of a multi-layered gelatine vascular h i i i h d h l bd d y n . ., n the quality of the outcome. Beyond its potential low cost, PAPER 3 31 the proposed system does not require specialist training to be operated, e.g., to ensure exact and well-aligned sensor placement over the targeted vasculature as in other methods. Most importantly, information was extracted regarding vascular features, such as bifurcations and stenoses, demonstrating the ability not only to detect features but to accurately detect the presence of stenosis even when it is of a low grade. Given the large amount of undetected carotid atherosclerosis cases that lead to debilitating or even fatal strokes, the findings of this study are significant a s they c an lead to a low cost diagnostic method for early screening medical examinations. Such screening examinations could be practised in primary care or at remote clinics - possibly even at home - providing a warning even without the presence of any symptoms, thus allowing for the referral of patients for further in-depth examinations when needed.

Claims

CLAIMS1. A system for investigating biological structures, the system comprising: a probe for application to a subject under investigation, said probe comprising a plurality of electrodes arranged in an array, the electrodes of said array being operable to apply electrical signals to said subject and to receive signals that are indicative of electrical properties of said subject; a probe controller responsive to control signals from a controller to select electrodes of said plurality to form a sensor, said sensor being operable to apply electrical signals to a region of said subject and to output received signals that are indicative of electrical properties of said region of said subject; a measurement module coupled to the probe controller for the receipt of said received signals, said measurement module being operable to determine from said received signals the electrical properties of said region of the subject; and a system controller operable to output control signals to said probe controller, the control signals specifying the electrodes of the array that are to be selected to form said sensor.

2. A system according to Claim 1 , wherein said sensor comprises first and second electrodes for the application of an electrical signal to said region of the subject.

3. A system according to Claim 2, wherein said sensor comprises third and fourth electrodes for the receipt of signals that are indicative of an electrical property of the region of the subject.

4. A system according to Claim 2 and 3, wherein said sensor comprises a tetrapole.

5. A system according to Claim 4, wherein electrodes of said tetrapole are linearly arranged.

6. A system according to Claim 5, wherein the linearly arranged electrodes of said tetrapole comprise a line of four electrodes that are immediately adjacent to one another.

7. A system according to Claim 5, wherein the linearly arranged electrodes of said tetrapole comprise a line of four electrodes that are each spaced from one another by one or more electrodes.

8. A system according to any of Claims 1 to 7, wherein said system controller is operable to output a plurality of control signals each specifying a different selection of sensors for investigating different regions of said subject, the arrangement being such that the control signals define a sensor that tracks across the subject to investigate multiple regions of the subject.

9. A system according to Claim 8, wherein the control signals each specify a different selection of sensors for investigating different linearly arranged regions of said subject.

10. A system according to Claim 9, wherein the control signals are configured to track the sensor over the subject in different directions to investigate multiple linearly arranged regions of said subject.

11. A system according to Claim 4, wherein the electrodes of said tetrapole are arranged in a polygon, such as a square or rectangle.

12. A system according to Claim 11 , wherein said system controller is operable to output a plurality of control signals each specifying a different polygonally arranged selection of sensors for investigating different regions of said subject, the arrangement being such that the control signals define a sensor that tracks across the subject in concentric polygons to investigate multiple regions of the subject.

13. A system according to any preceding claim, wherein said received signals are indicative of the electrical impedance of the region of the subject investigated by the sensor.

14. A system according to any preceding claim, wherein said sensor is operable to apply electrical signals at multiple frequencies to said region of said subject.

15. A system according to Claim 14, wherein said measurement module is operable to determine tissue type from received signals.

Images