MONITORING THE INJECTION OF FLUID

The invention relates to monitoring the injection of a fluid into a fluid conduit in the human or animal body. There are many situations where a non-invasive measurement of fluid flow in fluid conduits in the human or animal body is desirable. Such fluid flows may include flow velocity in blood vessels (such as arteries and veins), urinary tracts, gastro-intestinal ducts (such as the bile duct) and so on. Measuring the fluid flow velocity gives knowledge of the physiological condition of a patient, and can help with identifying potential health problems. Measuring the flow velocity can also help in determining the effect of procedures that introduce fluids into conduits of the body, such as the injection of contrast medium.

Often, contrast medium is injected or infused in procedures such as angiography, computed tomography, ultrasound and MRI. It is necessary to inject the contrast medium at high flow rates for a long time period, and to do this it is injected under pressure rather than being simply gravity fed. Other procedures that require fluids to be injected in a similar manner include, for example, injections during the course of chemotherapy treatments. To aid in performing such injections a number of injector-actuated syringes and power injectors have been developed.

When fluid is injected into a blood vessel under pressure there is a risk of extravasation occurring. Extravasation is the accidental infusion of fluid such as contrast medium into tissue surrounding a blood vessel, rather than into the blood vessel itself. Extravasation can result from various situations. A common cause is operator error in placement of the needle, or movement of the needle caused by patient movement, which results in the needle effectively missing the blood vessel. In this case the injected fluid is injected into the surrounding tissue instead of into the blood vessel. Extravasation can also occur through the inability of the blood vessel to tolerate the rate of injection of the fluid resulting in leakage through the walls of the blood vessel into the surrounding tissue. This is especially the case with older patients who can have brittle veins, or patients undergoing chemotherapy treatment. If extravasation occurs then the injection and the associated procedure often has to be aborted and repeated at a later stage.

Clearly, it is beneficial to be able to detect extravasation in order to prevent pain or injury to the patient and to allow the injection procedure to be stopped if required. It is also

beneficial to be able to monitor an injection to detect other problems that may occur during the course of the infusion of fluid, for example an equipment failure.

The applicant's earlier application, WO 2004/052431, discloses the use of an ultrasound Doppler technique to monitor an injection and provide a signal to indicate the possibility of extravasation. The method described is based on the fact that during the injection process the flow velocity of fluid in the blood vessel increases. Normally the flow velocity in a peripheral vein is very low, but during contrast injection using a power injector the flow velocity increases significantly. A direct monitoring of the increased venous flow velocity induced by the infusion can be performed. A lack of a velocity increase when injection commences indicates that extravasation could be occurring and thus gives the operator an early warning, prior to the point when extravasation would be detectable from a visual inspection or by palpation. The device disclosed can send an automated signal to the power injector to halt the infusion if no velocity increase is detected, thereby limiting the risk of tissue damage. However, although the system just described provided a major advance in the art, there are limits on how effectively it can measure the flow velocity within the blood vessel under certain conditions. A signal indicative of the start of injection or of a change in the flow rate can usually be detected, but during steady injection of fluid under pressure the Doppler signal can sometimes be too weak to provide a useful measurement. Further, with the prior system it is not possible to effectively monitor an injection under some conditions involving elevated flow rates and/or elevated concentration (viscosity) of the injected fluid.

Viewed from a first aspect, the present invention provides a method of monitoring the intended injection of a fluid into a blood vessel, the method comprising: conveying the fluid to a point of injection, introducing a disturbance into the fluid flow prior to injection, and monitoring the fluid flow in the blood vessel downstream of the point of injection using a Doppler ultrasound sensor.

By introducing a disturbance into the fluid flow the strength of the Doppler signal reflected from the fluid in the blood vessel is increased due to an increase in the number of reflection planes in the disturbed fluid flow. Thus, the disturbance introduced in the fluid flow prior to injection results in a disturbed flow in the injected fluid and/or blood in the blood vessel. This enables an injection to be more easily monitored at all stages. Further advantages are also obtained, such as the ability to monitor injection of fluids at an increased

range of flow rates and fluid viscosities. As the disturbance is introduced into the fluid to be injected before it enters the body, there is no need for additional procedures to be carried out directly on the blood vessel itself, instead the disturbance is simply added to the fluid as it is injected. The invention is particularly advantageous when used to monitor the venous injection of contrast medium, as contrast medium creates various problems for the prior art Doppler monitoring as discussed below. However, monitoring of the injection of any fluid can be carried out using the method and apparatus of the invention. The invention is also particularly useful in the monitoring of an injection to identify the onset of extravasation, but may also identify other problems with the injection process, such as a blockage or kink in the tubing conveying fluid to the patient.

In preferred embodiments, the disturbance comprises turbulence, eddies and/or cavitations in the fluid flow. The disturbance is thus created by exposing the fluid to changes in dynamic parameters, including but not limited to, pressure, velocity and direction. The disturbance may advantageously be introduced just prior to a cannula or Venflon™ that conveys fluid into the blood vessel.

Preferably, the disturbance comprises cavitation generated by a pressure reduction in the fluid flow, in particular a pressure reduction obtained by use of the Venturi effect. The method may include passing the fluid into an area of increased cross-section to suddenly decrease the pressure and hence initiate cavitation. A constriction prior to the area of increased cross section may be used to increase the pressure prior to the decrease

A splitter or baffle may be used to divide and/or mix the flow of fluid. This is thought to promote increased turbulence. Where a change in cross-section is used, a splitter may be placed just after the increase in cross-section. The monitoring preferably comprises measuring the fluid flow velocity in the blood vessel downstream of the injection site. A successful injection of fluid with result in changes to the fluid flow velocity corresponding to expected changes, typically consisting of an increase in flow velocity when injection commences, followed by a sustained elevated velocity, and finally a drop in velocity when the injection halts. If the flow velocity does not increase when the injection commences, then this indicates a problem with the injection process, such as the needle missing the vein, or an equipment failure. Similarly, if an unexpected drop in flow velocity occurs during the course of an injection, then this indicates

that a problem has occurred, for example there may be an extravasation resulting from a fragile blood vessel wall, patient movement may have dislodged or shifted the needle, or some failure of the injector equipment may have occurred.

The monitoring process may include providing feed back to the operator and/or feed back to a power injector that supplies the fluid to be injected. This enables a quicker response to potential problems in the injection process, and also the use of an automated response from the power injector if required. The feed back may include an indication of the fluid flow velocity in the blood vessel, for example a visual or aural indication. The feed back may include providing an alarm that indicates an unexpected change in the fluid flow velocity in the blood vessel, and/or providing an alarm that indicates the absence of an expected change in the fluid flow velocity. For example, if the rate of injection is constant and the fluid flow rate drops then an alarm may be provided. Alternatively, if the fluid flow rate does not increase when the injection begins or the rate of injection increases, then an alarm may be provided. The method may include sending a signal to a power injector to automatically stop the injection when a problem is detected.

The method may also include measuring a parameter of the injected fluid prior to injection, and providing an indication of an abnormality in the parameter. For example, the injection pressure or flow rate in the power injector could be measured, and a higher or lower value than expected may be used to provide an indication of a problem with the injector. This arrangement allows the operator to assess if a lack of a velocity increase or a drop in flow velocity is due to a failure of the injector mechanism rather than an extravasation or other problem with passage of fluid into the patient. For example, an increased pressure in the power injector could indicate that a pinched or kinked tubing is preventing fluid from reaching the patient. This measurement could be correlated with the Doppler ultrasound measurement of the flow velocity in the blood vessel, which would indicate a reduced velocity compared to the expected velocity. This could trigger an event-specific alarm or notification to the operator. Further, if a sudden pressure decrease in the power injector is correlated with the signal from the ultrasound sensor, and they are in accordance, it would suggest that an extravasation is about to occur, which again could trigger an alarm. Viewed from a second aspect, the invention provides an apparatus for monitoring the intended injection of a fluid into a blood vessel, the apparatus comprising: a source of disturbance for introducing a disturbance into the fluid flow prior to injection, and a Doppler

ultrasound sensor for monitoring the fluid flow in the blood vessel downstream of the point of injection.

The source of disturbance introduces a disturbance in the fluid such that the number of ultrasound reflection planes in the fluid is increased, thereby improving the signal strength received by the ultrasound sensor from the fluid in the blood vessel. Thus, as discussed above, without direct intervention in the flow of blood, a disturbance introduced into the injected fluid is used to provide a disturbance in the fluid flow in the blood vessel. In a preferred embodiment, the source of disturbance comprises a source of turbulence, eddies and/or cavitations in the fluid flow. The source of disturbance thus preferably exposes the fluid to changes in dynamic parameters, including but not limited to, pressure, velocity and direction.

The source of the disturbance may advantageously be located just prior to a cannula or venflon that conveys the fluid into the blood vessel. In one embodiment, the source of the disturbance is located in tubing that conveys fluid to the cannula. With this arrangement the apparatus can be set up in the same way as a conventional arrangement for contrast medium injections. Alternatively, the source of the disturbance may be in an insert located between the tubing and the cannula. The use of an insert means that specially adapted tubing is not required.

Preferably, the insert includes a flow path arranged to provide a drop in pressure, to thereby initiate cavitation at the appropriate fluid flow rates. For example, a change in cross- section that uses the Venturi effect may be present. The insert may include an area of increased cross-section to suddenly decrease the pressure and hence initiate cavitation. A constriction prior to the area of increased cross section may be used to increase the pressure prior to the decrease A splitter or baffle may be provided to divide and/or mix the flow of fluid. This is thought to promote increased turbulence. Where a change in cross-section is used, a splitter may be placed just after the increase in cross-section.

In a preferred embodiment the Doppler ultrasound sensor is arranged to measure the fluid flow velocity in the blood vessel. As discussed above, the fluid flow velocity is expected to have a certain profile when a successful injection occurs, and monitoring the velocity therefore enables problems with the injection to be identified.

The monitoring apparatus preferably includes a control system for providing feed back to the operator and/or feed back to a power injector that is providing the fluid for injection. The feed back can be an indication of fluid flow velocity measurements that result from possible problems with the injection. The control system may be for handling feed back as discussed above, and is preferably arranged to carry out such feed back processes.

The apparatus may include and/or may be integrated with the tubing and/or connectors that convey fluid from the source of injected fluid, generally a power injector, to the patient. The apparatus may also include a sensor for measuring a parameter of the fluid to be injected, i.e. the fluid prior to injection, and a control device arranged to provide an indication of an abnormality in the parameter. The parameter may for example be a pressure or flow rate as discussed above.

The method and apparatus of the invention allow continuous monitoring of the infusion induced venous flow, rather than only monitoring of certain stages of the injection process. The qualitative use of an ultrasound Doppler technique for monitoring of injection is enabled. Further, it is possible to monitor the injection of any concentration or viscosity of injected fluid, which is not possible using conventional Doppler injection monitoring as discussed below in relation to Figures 1 to 4.

The invention also encompasses computer program products containing instructions that when executed on a data processing apparatus will configure the data processing apparatus to carry out the method discussed above. In a preferred embodiment this comprises software loadable onto or stored on a computer readable medium and consisting of computer readable program code for performing the method when the software is executed on a computer.

Monitoring the intended injection may include the storage of the measured blood vessel flow velocity data. For example, the data may be stored on an electronic storage medium such as a hard disk drive, or it may be output as a hard copy, such as a printout of velocity in a table or graph. Storing the data allows for a better record of the procedure, and the data can be accessed later for quality assurance purposes, or to improve the operation of the injection monitoring. In preferred embodiments of the above aspects, an ultrasound probe having an array of two or more ultrasound sensors is used. The array is arranged to be placed substantially transverse to the direction of flow of the blood vessel so that at least one sensor of the array

will be located over the vessel. _. With this arrangement, the accuracy with which the detector must be placed is reduced. It can be difficult to precisely locate the blood vessel downstream of the point of infusion. This arrangement makes it simple to obtain a good signal without time consuming repositioning of the probe. Provided that the ultrasound sensor is located approximately downstream of the infusion site, at least one sensor of the array will be located over the vessel, and so it is not necessary to precisely locate the blood vessel at the point of measurement prior to commencing the measurement. Instead, a change in flow velocity will be detected by whichever sensor or sensors of the array are located over the blood vessel.

With this arrangement, it is preferable that the sensor of the array that is located over the blood vessel is automatically detected using the difference in signal between sensors. The signal from an ultrasound sensors varies depending on whether the sensor is located over the blood vessel or over ordinary tissue. The highest signal will be obtained at the sensor in the array that is receiving measurements of the moving fluid in the blood vessel. This sensor can thus be selected to be used to monitor the intended injection. In a particularly preferred embodiment, the signal received by the sensor of the array that has the lowest signal strength is used as a baseline signal level. This signal level is subtracted from the signal level of the sensor that is used to monitor the flow velocity in the blood vessel. The use of a baseline signal level in this way aids in noise cancellation, and will help reduce or eliminate erroneous signals caused by movement, palpation or other sources of noise, including external sources such as electrical devices. The baseline sensor may alternatively be a separate sensor intended to be positioned away from the blood vessel for the purpose of detecting signal changes due to movement and noise.

In all the embodiments discussed above, the ultrasound probe may include a microphone, and/or the ultrasound signal may be used as the basis for a sound signal. This enables audible feed-back to be provided. Where a microphone is used, the operator can simply listed to distinguish different types of blood flow, for example arterial and venous flow. When the ultrasound signal is used, this enables audible feed-back representing the change in ultrasound signal strength to be given to the operator. Consequently, the operator can listen for the expected change in strength as an injection commences, and/or monitor an ongoing injection by ear, with a change in signal strength resulting from a potential problem being audibly indicated, perhaps in addition to a visual indication.

The invention involves monitoring an injection of fluid into a blood vessel. However, in some circumstances the fluid will never enter the blood vessel, and thus strictly speaking there is no injection of fluid into a blood vessel. For example if the needle is badly located and directly injects into tissue instead of into the blood vessel. Thus, in the aspects above, reference is made to monitoring an intended injection, in order to encompass monitoring of both an injection that is successful in introducing fluid into the blood vessel, as well as an injection that for whatever reason is not successful.

Preferred embodiments of the present invention will now be described by way of example only and with reference to the accompanying drawings in which: Figure 1 is a plot of Doppler signal strength when 100'mL ofOptiray (320 mg I/mL) fluid is injected in a known way with a flow rate of 5 mL/s at room temperature,

Figure 2 is a plot produced as for Figure 1, but with 10 mL injected at an injection flow rate of 1.0 mL/s,

Figure 3 is a plot produced as for Figure 1 , but with 30 mL injected at an injection flow rate of 3.0 mL/s,

Figure 4 is a plot produced as for Figure 1, but with 50 mL injected at an injection flow rate of 5.0 mL/s,

Figure 5 is a plot of Doppler signal strength when 50 mL ofOptiray (320 mg I/mL) fluid is injected with a source of disturbance in accordance with the invention, wherein the fluid has an injection flow rate of 5.0 mL/s,

Figures 6 to 8 show the results of repetitions of the experiment of Figure 5,

Figure 9 is a plot of Doppler signal strength when 50 mL of fluid is injected as for Figure 5, with the addition of a non-return valve in the simulated blood vessel, wherein the fluid has an injection flow rate of 5.0 mL/s, Figures 10 to 12 show the results of repetitions of the experiment of Figure 9,

Figure 13 shows schematically an arrangement for monitoring the injection of fluid,

Figure 14 shows a Bayesian network used in a system for processing and classification of the Doppler signal,

Figure 15 illustrates an insert for introducing a disturbance in flow, and Figure 16 is an exploded view of the insert of Figure 15.

The inventors have found that the introduction of a disturbance, such as bubbles, eddies, cavitation and so on significantly increases the ability to measure fluid velocity by the

use of Doppler ultrasound technology. This is particularly the case where the injected fluid would otherwise tend to form a homogeneous, laminar flow, which occurs, for example, with high flow rates of contrast medium.

In the Doppler ultrasound measurement of flow velocity an ultrasound Doppler probe is used, which consists of one or more transducer elements. The probe is located against the skin of a patient proximate to a vein into which the infusion is being made and downstream of the infusion site. The probe is fixed to the patient's skin using an adhesive. In accordance with standard practice, a coupling medium (ultrasound gel) should be applied to the patient's skin under the transducer elements. A signal indicative of the flow velocity in the blood vessel is produced from the Doppler probe in a conventional manner.

When monitoring contrast medium or other injections by an ultrasound Doppler method, the reflected signal strength depends on the degree to which the medium is able to reflect the ultrasound waves to the ultrasound detector. A Doppler transducer consists of one transmitter and one receiver. The transmitter emits a sound wave and a signal reflected from the measured substance is picked up by the receiver. An increase in the number of reflection planes in the flow of injected fluid can be induced by the addition of disturbance, which may be a disturbance in the form of turbulence, cavitations (micro bubbles), eddies or just plain mixing of different fluids. Thus, by introducing a disturbance into the injected fluid, the strength of the reflected Doppler signal is increased. During contrast medium injection, or injection of other fluids, the blood flow velocity of the vein increases. Normally the flow velocity in a peripheral vein is very low, but during contrast injection using a power injector set at e.g. 5 ml/sec the flow velocity increases significantly. Using ultrasound Doppler technology this increased flow can be continuously monitored at the appropriate scale setting as long as the flow contains reflection planes to provide the required reflected signal. However, a regular injection of a homogeneous contrast medium into an inert medium creates very few reflection planes suitable for Doppler ultrasound. This is because the contrast medium will strongly displace any other liquid flow in the vein, leading to a flow of homogeneous contrast medium that does not readily result in mixing of the contrast medium with the blood. Very few reflection planes will be created without the presence of mixing with other fluids than contrast medium, and thus very poor reflected Doppler signals result, except at the start and end of injection, as discussed below in relation to Figure 1.

Further, contrast medium is supplied by a number of manufacturers in a variety of concentrations. The osmolality and viscosity of the contrast medium varies with both concentration and temperature of the medium. The chemical structure of the medium will also influence the viscosity (such as non-ionic monomers or dimers in Iodine contrast medium). The surface chemistry will influence the behaviour of the medium when injected into a blood vessel. The diagnostic procedure and the patient physiology will determine the concentration and volume of contrast medium to be administered, as well as the flow rate. Certain required injection regimes cause problems for conventional Doppler monitoring as shown by Figures 1 to 4. As illustrated in Figure 1, a conventional injection of a homogeneous contrast medium into an inert medium using a high flow rate and high concentration creates very few reflections suitable for Doppler ultrasound. Figure 1 shows readings collected using a 'phantom' set up to simulate injections into a blood vessel. The phantom consisted of a water filled box to simulate body tissue, and a silicone tubing to simulate the blood vessel, the tubing passing through the water filled box. The tubing used had a 3 mm internal diameter and a 5 mm external diameter. De-aerated water was passed through the silicone tubing to simulate blood. Contrast medium was injected into the tubing upstream of the phantom, and an ultrasound sensor of conventional type, such as the sensor discussed in WO 2004/052431, was placed to detect the flow velocity in the tubing within the phantom. The injected fluid was at room temperature and was introduced into the silicone tubing using a green venflon. The sensor was placed 4 cm from the venflon tip.

For Figure 1, the injection flow rate was 5.0 mL/s, and 100 mL of Optiray (320 mg I/mL) was injected at room temperature.

The peaking signal strength at start and end of the injection is characteristic of injections of contrast medium through catheters. As a homogeneous fluid is injected through a venflon or catheter, it creates cavitations close to the tip for as long as the injection speed increases, and these cavitations provide the increased Doppler signal strength. However, as a steady state injection flow is established, very little cavitation occurs, and it is therefore not possible to effectively monitor the injection process. As the injection speed decreases at the end, the same phenomenon of cavitation occurs, leading to another peak.

During the start and end points, the flow can be characterised as turbulent. The turbulence creates reflection areas in the fluid, suited for measurement by Doppler ultrasound.

However, between the start and end points of the injection shown in Figure 1 the flow of infused contrast medium does not produce a Doppler signal. In this phase of the infusion the flow can be characterised as laminar. A homogeneous fluid in a laminar flow generates very little reflections. Thus, with this high concentration and high flow rate the conventional method can only monitor the injection process effectively at the start and end of the injection. Therefore, any problems occurring during the steady state phase cannot be detected.

Figures 2 to 4 show the results of a progressive increase in concentration and flow rate from a measurable low rate toward the high concentration and flow rate of Figure 1.

Most concentrations at low injection flow rates result in a good mixing of blood and contrast medium, which induces a good Doppler signal. Figure 2 shows the signal produced when an injection with a low flow rate of 1.0 mL/s and a low volume of 10 mL. As can be seen, an elevated signal strength is produced throughout the injection process showing that the conventional monitoring technique is capable of producing useful results with this injection regime. However, higher flow rates, especially at higher concentrations (i.e. high viscosity), result in poor mixing with the blood. The flow tends to be more homogeneous, and it seems to displace the blood from the vein. There is also a great difference with respect to contrast medium temperature. A highly viscous contrast medium at room temperature tends to create a homogeneous/laminar flow with very few reflection planes compared to the same contrast medium at body temperature. Higher concentrations of viscous contrast medium will, because of the flow properties, form a 'plug' which is much more difficult to push into the blood vessels. Some time after the injection starts, the vein will reach a maximum pressure limit for the injected flow rate, and the contrast medium will then start to flow backwards in the vein (from the venflon) as well as forwards. At this point, as there is no blood present in the vein, and hence no mixing to cause reflection planes, it is very difficult to obtain a good Doppler signal.

Figures 3 and 4 illustrate the deterioration in measured signal when the flow rate and volume are increased to 3.0 mL/s, 30 mL (Figure 3) and then 5.0 mL/s, 50 mL (Figure 4). It will be appreciated that in Figure 4 the readings between the start and end phases of the injection are too low and erratic to provide a useful indicator of problems with the injection process.

As discussed above, to achieve an improved Doppler signal strength and allow monitoring of an injection throughout the course of the injection and also to allow monitoring of injections at increased flow rates and concentrations, a disturbance is introduced into the fluid flow prior to injection. Thus, a conventional arrangement for injecting fluid into a patient is used, with the exception that the cannula or venflon is provided with a source of a disturbance, such as a source of turbulence. To simulate this arrangement, experiments were performed as for Figures 1 to 4, but with the introduction of a source of disturbance in the venflon.

The effect of this modification to improve the Doppler signal strength can be seen from Figures 5 to 12. In the experiments shown the injection conditions of Figure 4 were repeated, first with a non-return valve in the liquid flow prior to injection (Figures 5 to 8) and then with a further non-return valve in the liquid flow along the simulated blood vessel (Figures 9 to 12). The purpose of the second non-return valve is simply to ensure that the beneficial effects of the invention exist for varying conditions in the simulated blood vessel. Figures 5 to 8 show that when a disturbance to the flow of the injected fluid is introduced, the signal strength is greatly improved compared to Figure 4, which has an identical flow rate, volume and concentration. The non-return valve introduces turbulence as the fluid passes through it.

Figures 9 to 12 show that with a simulation of different blood flow conditions, using a non-return valve in the silicon tubing, the Doppler signal of sufficient strength is once again produced.

It will of course be appreciated that alternative devices to the non-return valve could be used to produce turbulence in the fluid prior to injection. For example an angle cock or a static mixer could be used. A preferred device is discussed in more detail below with reference to Figures 15 and 16. In addition, as discussed above, an alternative to a source of turbulence is a source of micro bubbles, which has the same effect of increasing the incidence of reflection planes to thus increase the Doppler signal strength.

Thus, as will be appreciated through the use of the phantom studies it has been shown that the introduction of a disturbance into the fluid flow prior to injection results in an improved Doppler signal, allowing monitoring of an injection throughout the entire course of the injection and also allowing injections and higher concentrations and higher flow rates to be effectively monitored. In the discussion above an obstruction placed in the liquid flow in

front of the venflon, was used, and it was confirmed that this created enough turbulence to obtain a good Doppler signal. It shows a significant difference in the Doppler signal for high flow rate injections of contrast medium compared to the injection with no disturbance as shown in Figure 4. The same phenomenon will occur when alternative turbulence sources are used or when micro bubbles are injected into the flow.

An embodiment of an injection monitoring system will now be described with reference to Figure 13. A contrast medium is infused into the patient's arm 1 from a power injector 6. The power injector 6 is controlled by an electronic controller 5, which varies the injection speed as required and starts and stops the injector. The contrast medium flows via flexible tube 7 to a cannula or venflon arrangement 8, which comprises a connector for connection to the flexible tube, and a fine bore tube 9 which has been inserted into a vein in a known manner. A source of disturbance is in the tube 7 or the cannula arrangement 8 to introduce turbulence or the like. The source of disturbance could alternatively be in a separate interconnection unit, placed between and providing a fluid connection between the tube 7 and venflon 8. As in the above examples, the source of disturbance could be a non-return valve. In preferred embodiments the source of disturbance is an insert as described below with reference to Figures 15 and 16.

An ultrasound Doppler probe 3 is placed above the same vein at a convenient distance downstream so as to be clear of the infusion site. The Doppler probe 3 consists of a number transducer elements 2 which in use are placed at an angle to the vein to create and detect a Doppler shift resulting from the flow of fluid in the vein.

The transducer elements 2 comprise ultrasound sensors arranged in an array, which is placed substantially transversely over the vein. The signal levels of the sensor elements in the array is measured, and this is used to select appropriate sensors to measure the flow velocity and to provide a baseline signal for noise cancellation. The sensor in the array with the highest signal is selected to be used to as a first sensor, which measures the flow velocity. The sensor in the array that has the lowest signal strength is used as a second sensor, for noise cancellation. As will be appreciated, the sensor with the lowest signal level will be a sensor that is not affected by the movement of fluid in the blood vessel. The signal level of the second sensor will be the result of background reflections from body tissue and noise from external sources and from body movement. This baseline signal is deducted from the signal

level of the first sensor in order to give a more accurate signal indicating the flow velocity in the blood vessel and changes in this flow velocity.

The probe 2 is connected via a flexible lead 4 to a controller 5 which includes a processor unit and display. This converts the output from the probe 3 into a form that may be displayed as an image on display unit in the conventional manner. ^• In addition it provides a digital signal proportional to the flow velocity detected by the probe 3. This value is then also displayed on the display.

The controller 5 receives data concerning flow velocity and injection parameters, and provides a signal to the power injector 6 and a scanner as required in order to control the injection and an associated scanning procedure. It will be appreciated that various arrangements could be used to implement the required control device that monitors and controls the injection process. There can also be data passed back to the control device from the power injector, such as data from a pressure sensor, which can be used to indicate if there is a potential problem with the power injector or other injection. The operator inputs the desired injection flow rate and the duration of the injection into the processor unit of the controller 5. The operator also inputs into the processor unit of the controller 5 the details of the proportion of the injection that triggers a change in the content of a second signal when a problem is detected during the course of the injection. An associated scanning process, such as angiography or MRI etc., is set up as required. When the injection is to commence the operator inputs a start signal into the controller 5. This in turn transmits a start signal to the controller 5 which energises the power injector 6 and causes it to run at the desired speed. The same start signal can also be used to co-ordinate initiation of the scanning process.

The processor unit then checks the flow velocity signal produced by the Doppler probe 3 as described above. If it is not satisfactory within a pre-determined short period of time, i.e. if the expected increase in velocity does not occur, then the infusion and associated scanning process will be stopped. This is achieved by sending a first signal to the controller 5 and also to the associated scanning equipment.

If the injection commences satisfactorily then the controller 5 continues to monitor the flow velocity signal. If an unexpected drop in flow velocity occurs then the controller 5 issues the second signal, the content of which depends on the proportion of the injection that has been completed and/or on the size of the drop in flow velocity. Thus, the processor unit

also monitors the extent to which the injection has been completed, in this case the elapsed injection time as compared to the total duration of the injection.

In general, as discussed above, the second signal is a signal to halt the injection process as well as the scanning process if a flow velocity drop larger than a set limit occurs before the bulk of the injection has been completed, but if most of the injection has been completed then the second signal is a signal to produce a notification to the operator without stopping the injection process. As noted above, the set proportion of the injection at which the content of the second signal changes can be input by the operator before the injection is commenced, in order to allow the control of the injection to be tailored to the particular injection and scanning process. However, as an alternative or in addition to the possibility of customising this proportion, the processor unit may have a default proportion, such as 50%. Thus, in this situation, if a drop in flow velocity occurs prior to 50% of the injection being completed then the injection and scanning process is stopped, whereas if the drop in flow velocity occurs after 50% of the injection is complete, then the processor unit 4 issues a notification or alarm.

If the drop in flow velocity is smaller than the set limit for stopping the procedure, and the scanning procedure is one which utilises a waiting time prior to beginning, then the second signal can be used to delay the scanning procedure dependent upon the size of the drop in flow velocity as discussed above. It will be appreciated that in general a drop in the flow velocity corresponds to a drop in the dB level of the signal provided by the Doppler sensor. Thus, in the foregoing discussion, references to the size of a drop in flow velocity correspond to measurement of the size in the drop in dB of the signal from the sensor.

The processor unit includes signal processing means to classify the Doppler measurements and identify anomalies. This can be done using known techniques, such as neural networks as discussed by Guler I. and Ubeyli E. in "A recurrent neural network classifier for Doppler ultrasound blood flow signals", Pattern Recognition Letters, Volume 27, Issue 13, 1 October 2006, Pages 15601571 or Support Vector Machines (SVMs) as discussed by Ubeyli E in "Doppler ultrasound signals analysis using multiclass support vector machines with error correcting output codes", Expert Syst Appl. 33(3): 725733 (2007). Bayesian classifiers have also been used to classify medical Doppler signals, as discussed in "Bayesian Classifier for Medical Data from Doppler Unit" by Malek J., Acta Polytecnica, Vol. 46, no.

4/2006. It is presently preferred however to use a pattern classification method combining a Bayesian network and a sparse kernel classifier, as described below.

In the system used with this method, the transducer elements 2 comprise three continuous wave Doppler 4 MHz transducers. These are attached to the patient's arm, over the vein and proximate to the contrast cannula. The transducer array is attached to the patient perpendicularly to the vein direction, circumferentially around the arm. A saline injection is administered first and then after a pause, the contrast agent is administered. Finally another saline injection is administered. There is a synchronization signal available that specifies whether the injection of fluid has started or ended and whether the injection is saline or a contrast agent.

First, pre-processing of the output signal from the transducer elements is carried out as follows. The transducer signal is quadrature demodulated, then sampled at 25 kHz and high pass filtered. The resulting power Doppler signal is short time Fourier transformed with a Harming window of 0.2 seconds length and a two window length overlap. This results in a pre-processed 0.2 sec/2.5 Hz resolution time frequency signal. Finally, the 50Hz band and its harmonics are removed.

After the pre-processing the inputs to the pattern classification system are:

1. three x 2D pre-processed time frequency samples, and

2. a synchronization signal tuple {(start,stop), (contrast, saline)} The classification system output is a state variable capturing if the injected fluid is running normally through a vein, and if nothing seems to flow or an extravasation may be occurring. The variable "flow-state" is hence defined as: flow-state = (intravenous, noflow, extravasated)

The patient examination goes through several distinct phases. For the classifier, important phases are a possible pre-check phase where a saline injection is administered, a possible saline injection before the contrast injection, the contrast injection itself and finally a possible saline injection after the contrast injection. A variable "examination-phase" is hence defined as: examination-phase = (saline-precheck, saline-preinjection, contrast-injection, saline-postinjection)

In a general dependency model it is assumed that the flow-state characteristically influences the received Doppler signal. A Temporal Feature (TF) is defined to be some

representation of the Doppler signal at time t. The function TF(t) then statistically depends on the flow-state at time t.

It is also hypothesised that the chance for a particular flow-state is dependent on the flow-state at earlier times. For example, if it is not clear whether the flow is intravenous or extravasated during the saline-preinjection phase (e.g. the probabilities are roughly equal) then that might heighten the chance that the flow-state is seen as extravasated during the contrast- injection phase. Thus, the probability of a flow-state at time t _n depends on the flow-state at time t _n- i.

Finally, since the task is pattern classification, an independent flow-state probability estimate is defined, which stems from a classification model based on the Doppler signal. This variable, flow-state _mo d, at time t depends on the flow state at time t. Figure 14 shows these dependencies in terms of a Bayesian network. The circles are statistical variables. The arrows denote a statistical dependency. Shaded circles are observed variables and white circles are latent variables. The "..." and the arrow with dashed line on the left illustrate a chain of dependencies from time 1 to time nl.

Values of flow-state as it varies with time are defined as a vector flow-state(t), where flow-state(t) = {flow-state (1),.. , flow-state (n). Similarly, vectors flow-state _mo d(t) and TF(t) are defined. The joint probability for the sequence of ^' flow-states, model estimated flow-states and temporal features is then: p(flow-state, flow-state _mod , TF) =

IJ p(TF (/7)| flow-state (n)) f[p(flow-state _mbd (n)\flow-state(n)) (i) n=\ λ/

Ylp(flow-state(n)\flow~statθ(n- \ )) n^Y

The densities in equation (1) are assumed to be Gaussian, and so an expectation maximization algorithm can estimate the particular distributions in (1) given a set of observations, and a sum-product algorithm may be used to compute desired marginal distributions, as set out in "Pattern Recognition and Machine Learning" by Bishop C, Springer 2006, ISBN 0387310738.

A normal B ayes risk estimator can be used to make classification decisions, based on the marginal distribution $(flow-stαte(n)) and an appropriate loss function, which can be

deteπnined based on the particular procedure being undertaken. The Bayesian network thus captures the dependencies between latent variables, observed measurements, and previous system states.

The postprocessed Doppler signal is rich in frequency content, and it is thus reasonable to attempt to classify the flow-state based on a time-frequency representation of the signals. The Doppler baseband range is roughly 4 kHz. With the aforementioned 2.5 Hz resolution, there are some 1600 frequency coefficients per sample. The classification approach used should hence efficiently deal with high dimensional data in real time. '

Furthermore, it would be desirable if the classifier provides a confidence estimate, preferably as a probability that can directly serve as the p(flow-statem _od ) estimate.

A sparse kernel machine such as a Support Vector Machine (SVM) is thus a possible classifier since such a machine capable of dealing efficiently with high dimensional data and is well suited for time series analysis. However, a preferred sparse kernel machine is a machine named the Relevance Vector Machine (RVM). This sparse kernel machine is based on a Bayesian framework and is described by Tipping, M. E., in "Sparse Bayesian learning and the relevance vector machine", Journal of Machine Learning Research 1, 211-244. 2001. The RVM is used to classify the Doppler signals based on the Bayesian considerations set out above.

Briefly, the RVM estimates a conditional distribution: piflow- state\x β ^{' )} ) (2) where x is the input vector, w is a parameter vector, N is the normal distribution and β is the noise precision. y(x) is defined in terms of a set of basis functions that with the RVM formulation is constrained to the Support Vector Machine (SVM)-like form:

y[x)=∑ w _n k{x,x _λ )+b (3)

Here, in contrast to SVMs, the kernel k does not need to be positive-definite.

As set out in the paper by Tipping, on benchmark data the RVM has about equal accuracy as the SVM, training times are roughly an order higher, but there is no need to do exclicit parameter optimalization for RVM kernels. Classification of new data is significantly faster for RVMs than for SVMs.

The RVM is thus used to provide an effective probalistic estimate of flow-state using the Bayesian network set out above. This can then be used to set an output value of flow- state, and the control device can hence provide appropriate signals to the power injector and operator when a potential problem has occurred. As noted above, Figure 15 illustrates an example of an insert 10 for introducing a disturbance into the fluid flow prior to injection. The insert 10 is for placement prior to the catheter 8 shown in Figure 13. The insert 10 has an inlet part 11 with a Luer thread 12 for fitment to the incoming fluid line 7, and an outlet part 13, fitted with a nut 14 so that it can be joined to an appropriate fitting to the catheter 8. The central part 15 of the insert 10, and the internal connections thereto, forms the source of disturbance for the fluid flow.

In Figure 16, the insert 10 of Figure 15 is shown in exploded view, and the internal arrangement of the central part 15 can be seen in more detail. A chamber 16 is formed between two halves of the body of the insert 10, and has a relatively large cross-sectional area. The inlet part 11 has a flow path with a relatively small cross-sectional area, which opens suddenly into the chamber 16. This hence forms an expansion chamber, with a pressure reduction resulting from the Venturi effect as fluid flows from the inlet part 11 into the suddenly increased cross-section of the flow path through the chamber 16, and this causes cavitation to occur when fluid is flowing at appropriate flow rates.- To further disturb the flow, the chamber 16 is fitted with a splitter 17, which is a baffle with an X-shaped cross- section, that divides the flow into four parts, before it is recombined in the outlet part 13, and the fluid continues to the catheter 8.

By means of the insert 10 of Figures 15 and 16 a disturbance is introduced into the fluid flow in the form of cavitations and turbulence. As discussed above, the result of this disturbance is that when the injected fluid passes into the blood vessel, an increased number of reflection planes is present and the strength of the reflected Doppler signal is advantageously increased.

In some embodiments described above, an injection of contrast medium alone is described. However, as is known in the art, the contrast medium injection could occur in conjunction with an injection of saline, as described in connection with the preferred signal processing method. Such injections are achieved using 'double-barrelled' injectors. Saline can be used alone prior to or after the injection of contrast medium. If injected before the contrast medium the saline acts to open the vein, and prepare it for the contrast medium

injection. When injected after the contrast medium the saline can be used to push the contrast medium further along the blood vessel by applying pressure behind the contrast medium. This is beneficial as it allows contrast medium at a higher concentration to be driven to a desired location for imaging, which might be some distance along the circulation system from the point of injection.

Further, as noted above, the invention, at least in some embodiments, is not limited to monitoring the injection of contrast medium, but also provides benefits when utilised during the intended injection of other fluids such as cytotoxins, anaesthetics, chemotherapy drugs and so on. As will be appreciated, the flow detection apparatus of the invention is inherently capable of producing a measurement of a blood flow velocity, or other fluid flow velocities in fluid conduits in the human or animal body. Therefore, the processing unit 4 may usefully be provided with the ability to provide a straightforward velocity measurement, in addition to having the capability to detect extravasation and the like as discussed above. The system shown in Figure 13 is illustrated as using a wired connection 3 between the probe 2 and controller 5. In an alternative embodiment, data transmission between the probe 2 and controller 5 is by a wireless connection. The use of wireless data transmission is useful as it avoids a potential entanglement risk, and gives more freedom of movement by the patient and around the patient. Further, the various processing units and control devices of the embodiments described and claimed can be a single unit, such as a CPU, or could for convenience be separated. For example, local processing of ultrasound sensor signals could occur at the probe, enabling the first and second sensors to be selected and the baseline signal used to adjust for noise without the need to transmit data to and from the processing unit or control device that controls the power injector. Another modification to the system of Figure 13 involves the use of sound. As is well known, venous and arterial flows have distinguishable sounds. A microphone can be provided with or as part of the ultrasound probe in order to enable the sound of blood vessels at the location of the probe to be heard. These sounds can be provided to the operator by head phones or by a speaker system. This allows the operator to more easily position the probe over the desired blood vessel.

In addition, the variation in signal strength from the ultrasound probe during the course of a procedure can be relayed as sound to the operator, i.e. the ultrasound signal can be

used as the basis for a sound signal. This allows any the change in signal strength resulting from correct operation or from potential problems to be presented as audible feed-back to the operator. The audible feed-back can be used in addition to a visual signal and/or a separate alarm.