Patent Application
20250134496
The present disclosure is directed to imaging and measurement of the vascular system. In particular, it is directed to the measurement of the Inferior Vena Cava, IVC, from ultrasound imaging.
Heart failure is one of the most significant chronic conditions afflicting adult populations. In the United States, 5.7 million Americans have heart failure, with 870,000 new cases annually. As the population ages, this population is growing, as approximately 10% of the population over 80 suffers from heart failure. It is estimated that by 2030, 8 million Americans will have heart failure. The cost of caring for heart failure is over thirty billion dollars per year. Twenty billion dollars of this cost is direct medical costs. This expense is expected to more than double over the next fifteen years.
In patients with chronic heart failure, a significant portion of these costs are due to hospitalization to manage acutely decompensated heart failure (ADHF). Each re-hospitalization can last up to a week, and costs approximately $10,000. ADHF is very often the result of some combination of a downturn in the heart's performance and excessive intake of fluids and/or salt. This leads to a build-up of fluid in the vascular system. Increased blood volume in the left atrium at higher pressure means higher blood pressure in the lungs, which eventually leads to fluid filling the lungs and an inability to breathe. At this stage it is imperative to hospitalize the patient to carefully manage them while drugs are delivered to remove the excess fluids.
Managing these patients to prevent the need for re-hospitalization is extremely challenging. Many non-invasive approaches to monitoring patients have been tried, such as weighing patients daily to detect fluid weight gain, having a nurse call them daily to assess their health status, and so on.
It is important to measure the onset of ADHF early enough to give the patient and/or caregiver enough time to adjust their behaviour, medication, or other factors to prevent the patient from ending up with acute congestion and the need for hospitalization.
Congestive heart failure is so named because additional blood volume backing up into the lungs causes fluid to seep out of the pulmonary circulation into the airway passages of the lungs, causing congestion of the lungs. The patients become short of breath, and typically need to be hospitalized and carefully managed while the excess fluid is removed by a combination of fluid management and aggressive use of diuretic medications.
Congestion occurs because the left ventricle is not able to pump all the volume of blood returning to the heart from the lungs. Although measurement of left atrial pressure, typically by measuring pulmonary artery wedge pressure, is commonly considered the most direct way to measure congestion in heart failure, there are other areas where congestion can be detected. When additional blood volume is added to the circulatory system, the inferior vena cava (IVC) is one of the first places for that added volume to collect. Lee et al, “Prognostic significance of dilated inferior vena cava in advanced decompensated heart failure” International Journal of Cardiovascular Imaging (2014) 30:1289-1295 states “In patients with advanced heart failure, left ventricular systolic dysfunction causes increased left atrial pressure. The pressure is transmitted back through the pulmonary circulation to cause pulmonary artery hypertension. The pulmonary artery hypertension can worsen pre-existing right ventricular dysfunction and exacerbate tricuspid valve regurgitation, leading to systemic venous congestion. If venous congestion and elevated central venous pressure are the hallmarks of heart failure, then distention of the inferior vena cava [by echocardiography] may be a good prognostic marker in patients with decompensated heart failure.”
Heart Failure (HF) is a systemic and chronic disease which can involve many organs, including the liver, kidneys, and lungs. Continuous monitoring of the arterial and venous function, in addition to pulmonary function provides insight into the progression of the disease and its side effects.
Other volumetric related conditions and therapies such as dialysis, ultrafiltration, pulmonary hypertension, and hypertension may also benefit from such a system but for simplicity the detail below will solely focus on the HF application.
From venous perspective, diameter measurement of the IVC, proximal to the right atrium has been useful for identifying fluid status and this has been shown to be useful in the management of heart failure patients.
In a proposed novel application, the efficacy of this measurement can be further enhanced with the ability to discern the extent to which the congestion of the liver and/or kidneys contributes to the overall HF condition. Additionally, monitoring the structures and flowrates of the IVC, portal vein, renal veins and aorta provide additional detail on the function of the cardiovascular system and fluid status of a patient.
Haemodynamic congestion is a complex phenomenon and current high-end care involves the use of extensive, invasive measurements to estimate the total body water content and then estimate the proportions in each compartment (intravascular, extravascular, interstitial etc.).
From a venous perspective, diameter measurement of the IVC, proximal to the right atrium has been useful for identifying hemodynamic congestion and this has been shown to be useful in the management of heart failure patients.
In addition to correlation with right atrial pressure, the diameter of the IVC may correlate with renal function and renal sodium retention, which are also very important prognostic factors of heart failure. Therefore, increasing IVC volume and/or pressure may be a very effective early indicator of worsening heart failure condition.
Recent studies have indicated that the variation in IVC volume over the respiratory cycle is a more sensitive measurement of fluid overload and/or heart failure than simple measurement of average IVC volume, diameter, or pressure. During inspiration, intrathoracic pressure decreases, thereby increasing venous return and causing collapse of the IVC. During expiration, intrathoracic pressure increases, decreasing venous return and causing an increase in the volume of the IVC.
Since the IVC typically collapses in the anterior-posterior direction, some studies have suggested that the most accurate technique for measuring IVC volume changes with ultrasound is to measure the distance from the anterior wall of the IVC to the posterior wall of the IVC.
In applying this measurement to heart failure, at least one study has suggested that a variation of less than 20% (measured as maximum anterior-posterior dimension minus minimum A-P dimension, divided by the maximum A-P dimension) is indicative of ADHF.
While such vessel dimensions may be measurable using ultrasound, magnetic resonance imaging, computerized axial tomography, or other technologies, these imaging procedures must be administered in a hospital or other specialized facility, require expensive medical equipment, and skilled operators to obtain and interpret data and hence hamper daily monitoring in large populations.
As a result, the condition of a heart failure patient can worsen into a critical state before care providers become aware of it, dramatically increasing the mortality risk and cost of treatment for the patient.
If a heart failure patient can be monitored daily at home or in a remote location, this would enable the patient and physician to take proactive steps in time to prevent acute decompensation requiring re-hospitalization.
Magnetic resonance imaging and computerized axial tomography equipment is clearly limited to hospital use. Hospital ultrasound equipment is unsuitable for remote continuous monitoring due to its size, cost, and the need for trained operators, both for operation, correct transducer placement and image interpretation.
Use of a non-invasive, portable ultrasound device for vascular monitoring and intravascular volume management, which enables remote monitoring of vessel dimensions and flowrates without the requirement for a trained operator, is desirable. It is an aim of the invention to build a 3D model of the vascular system to identify structures and overcome the challenge of monitor positioning accuracy.
The multiple measurements obtainable with such a device would enable non-invasive estimation of the volumes of blood in the different areas. For example, the IVC diameter and collapsibility could provide an indication of vascular volume, IVC geometric response to other perturbations including manoeuvres such as breath hold, stand up from seated and others may provide information relating to the splanchnic activation or vascular tone as well as vascular volume, the portal vein diameter and flow rate could provide an indication of splanchnic activation and volume while the renal flow rates could provide an indication of renal function; all of these factors being useful in the diagnosis and treatment of vascular volume.
It is a further aim to improve the speed and accuracy of marking the outline of the inferior vena cava (IVC) and other structures in images acquired using external ultrasound (EXUS), intravascular ultrasound (IVUS), MRI and CT. Manual placement of markers on the outline of the vessel in the image is not only slow but also adds additional variability or error to the measurement.
The use of a module applied to the abdomen in the sub xiphoid position is described in WO 2013163605 A1. It is claimed that the module examines IVC collapsibility through the use of a cylindrical mechanical ultrasound array positioned on the patient by a medical professional in pre-hospital emergency situations. A collapsibility percentage is returned on the display of the device. The mechanical transducer may limit the imaged field of view leading to erroneous geometrical evaluation of the vessel, and, further, the need for placement of the device by a medical professional would limit its remote use, particularly in the home. The module would not provide any imaging data but just some descriptive data and is solely limited to IVC and Aorta measurements.
In addition to heart failure patients, haemodialysis patients have a chronic need for careful volume management. Large volumes of fluid are involved in the haemodialysis process and managing patients so that they don't end up hypovolemic or overloaded with fluid requires careful management. A monitor which provided immediate feedback on these patient's volume status before, during and after haemodialysis would be very helpful.
There are other groups of patients who might benefit from such a monitor. For example, patients in septic shock or occult shock due to trauma are subject to hypoperfusion which can be identified by measuring the degree of variation of the IVC with respiration (also referred to as collapse of the IVC), and fluid volume levels are key to the treatment of patients with Left Atrial Assist Devices (LVAD's).
The present invention provides a method of determining measurements of the Inferior Vena Cava, IVC, using ultrasound imaging, comprising:
Providing a three-dimensional, 3D, ultrasound image of a portion of the body in which the IVC is located may comprise combining a plurality of 2-dimensional, 2D, ultrasound images into a three-dimensional, 3D, ultrasound stack.
The plurality of 2D ultrasound images may be obtained by a body-worn ultrasound transducer array.
The plurality of 2D ultrasound images may be obtained from a plurality of directions or locations on the body, by a body-worn ultrasound transducer array or other form of ultrasound transducer.
The plurality of 2D ultrasound images may be obtained using beam forming.
Performing image analysis on the 3D ultrasound image to identify the IVC relative to other anatomical structure may comprise:
Identifying at least one anatomical landmark in the image may comprise the use of Doppler data to identify the direction and velocity of blood flow.
Identifying at least one anatomical structure in the image may comprise the use of known geometric properties of the at least one anatomic structure.
Performing image analysis on the 3D ultrasound image to identify the IVC relative to other anatomical structures may comprise identifying anatomical structures using edge detection.
The method may further comprise verifying the identification of the IVC in the image using an additional metric selected from a group comprising Doppler velocity, direction of blood flow, pulsatility, distances, pulmonary B-line measurements and audio respiratory data.
Determining the cross-sectional area of the IVC from the cross-sectional image of the IVC in the selected slice may comprise using edge detection techniques to identify the wall of the IVC.
Determining the cross-sectional area of the IVC from the cross-sectional image of the IVC in the selected slice may comprise the fitting of an ellipse to the IVC.
Determining the cross-sectional area of the IVC from the cross-sectional image of the IVC in the selected slice may comprise the use of thresholding techniques.
The method may further comprise the use of blob detection and analysis techniques to determine the cross-sectional area of the IVC from the cross-sectional image of the IVC in the selected slice.
The steps of providing, performing, selecting and determining may be repeated over time to obtain a series of measurements.
Throughout the series of measurements, each slice selected may comprise the same cross-sectional image of the IVC.
The selected single slice may differ between subsequent measurement time points in the series of measurements due to movement of the IVC during respiration.
The method may further comprise plotting the series of measurements against time in a graph.
The method may further comprise determining at least one of minimum area, mean area, maximum area and collapsibility of the IVC from the graph.
The method may further comprise comparing at least one of the determined minimum area, mean area, maximum area and collapsibility of the IVC to a threshold to record an event.
The series of measurements may be analysed to identify trends or fluctuations against an allowable tolerance over a period of time.
Determining the cross-sectional area of the IVC from the cross-sectional image of the IVC in the selected slice may comprise the use of a neural network to mark the IVC in the ultrasound image.
The invention further provides a computer program product comprising instructions which, when the program is executed by a computer, cause the computer to carry out the aforementioned method.
The invention further provides a computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the aforementioned method.
The ultrasound images may be provided by a portable ultrasound device comprising: an ultrasound transducer comprising a 2D array of independently controllable ultrasound transducer elements for producing an ultrasonic pulse; a power source for providing an electrical signal to the ultrasound transducers; and analysis means comprising: a beam former for optimizing transmitted ultrasound signals; a system for processing received ultrasonic signals; and a communications interface for communicating data for subsequent analysis.
Each transducer element may comprise at least one piezoelectric element. Alternatively, each transducer element may comprise at least one capacitive micromachined ultrasonic transducer, CMUT. The ultrasound transducer may be configured for at least one of B-mode ultrasonography and Doppler ultrasonography.
The data communicated may comprise a plurality of 2D scans.
The analysis means may be configured to compile a 3D representation from a plurality of 2D scans. The data communicated may then be assembled to provide the 3D representation or image stack.
The portable ultrasound device may further comprise a second ultrasound transducer for imaging of B-lines. The B-line is a kind of comet-tail artefact indicating subpleural interstitial edema. The second ultrasound transducer enables the detection of the reverberation patterns of B-lines for detecting plural edema. The second ultrasound transducer may comprise a single element 1D array transducer.
The portable ultrasound monitor may further comprise display and communication means. The portable ultrasound device may further comprise a user interface. The user interface preferably permits manual activation of the device. This permits the user to self-activate the device when positioned and may also provide guidance or feedback to the user, based on automated analysis and identification of anatomical structures in the imaged region, on location and angulation of the device to provide complete imaging of the intended region of interest. The portable ultrasound device may further comprise a communications module.
Preferably, the ultrasound transducer is generally rectangular in shape. The ultrasound transducer may comprise a length of at least 5-10 cm.
Securement means for securing the ultrasound transducer on the torso in one or more predetermined positions may be provided. The securement means may comprise adhesive securement means. Such securement means may also provide the medium for the transfer of ultrasounds between the transducer and the body. The securement means may alternatively or additionally comprise at least one strap for strapping the ultrasound transducer to the torso. The securement means may additionally or alternatively comprise a subcutaneous magnetic coupling and corresponding magnetic housing coupling to enable consistent device placement.
Repeatability of position is key in use. The body may additionally be marked by way of tattoo or similar to aid the user in correctly positioning the transducer. In one embodiment the adhesive securement means may comprise means for marking the skin, such as a die or temporary tattoo, to facilitate repeat placement. In another embodiment magnetic elements could be permanently implanted, sub-dermally, in the patient. These elements would then align and attach to corresponding, magnetic elements in the ultrasound device, thus repeatedly positioning it. The processing means may be configured to compensate for transducer placement error.
The ultrasound transducer may further comprise a multiplexor. The power source may comprise a high voltage generator. The portable ultrasound device may further comprise a signal generator to generate a signal to the user. The signal may be used to prompt action such as activation or repositioning. In the latter case, the signal generator may be activated in response to the signal processor detecting an incorrect positioning of the transducer. The signal may be at least one of an audio signal, a visual signal, or a haptic signal.
The portable ultrasound device may further comprise an auscultation device for recording lung congestion sounds. The sounds from the lungs may be recorded over a respiratory cycle to provide additional data to the measurement data.
The portable ultrasound device may further comprise means for tracking the motion and position of the patient, for example by way of an accelerometer or gyroscope.
The present invention further provides a method for vascular monitoring and intravascular volume management.
The present invention further provides systems and methods for marking ultrasound images, including but not limited to those obtained using the portable ultrasound device described herein.
In one embodiment of the invention the marking is performed on ultrasound images obtained using an intravascular ultrasound system. The intravascular ultrasound system may be used in conjunction with an intravascular sensor such as those disclosed in WO2016131020 A1 Implantable Devices And Related Methods For Heart Failure Monitoring, WO2017024051 A1 Devices And Methods For Measurement Of Vena Cava Dimensions, Pressure, And Oxygen Saturation, WO2018102435 A1 Wireless Resonant Circuit And Variable Inductance Vascular Implants For Monitoring Patient Vasculature And Fluid Status And Systems And Methods Employing Same, WO2018220143 A1 Implantable Ultrasonic Vascular Sensor, WO2018220146a1 Implantable Sensors For Vascular Monitoring, and WO2019232213A1 Wireless Resonant Circuit And Variable Inductance Vascular Monitoring Implants And Anchoring Structures Therefore.
In accordance with some embodiments of the invention, at least one marker on an intravascular sensor is used in the marking process. Preferably the intravascular sensor provides a plurality of markers around the circumference of the vessel. Where the sensor is expandable and contacts the vessel wall, the markers may be sections of the sensor. For example, for a zig-zag shaped sensor having a plurality of struts linked by crown sections, the struts or crowns may form markers visible in the ultrasound images. Marking of ultrasound images may be semi or fully automated.
In a fully automated embodiment, a multitude of images would be required to cover all or almost all the possible scenarios for the sensor struts appearance in the produced images. The locations of the struts may then be marked manually in this training data set to train an algorithm to find those locations in images without annotations. A sub-set of the training data may be used to test the functioning of such system. It may be required to retrain the system for everyone. Then other images may be required for refining the method for the individual before it may be used in practice to automatically detect the sensor location.
The present invention provides improvements in throughput and accuracy in ultrasound based IVC measurement. Automatic processes address inaccuracies in manual marking of ultrasound images and allow the visualisation of small changes in the IVC consistently.
Embodiments of the invention will be described, by way of example only, with reference to the accompanying drawings in which:
The present invention uses ultrasound imaging in combination with advanced image processing to identify the appropriate structures, determine measurements. Clinical status may be inferred from these measurements, to enable remote diagnoses and the information may be reported to a clinical team to aid in the monitoring of HF patients.
The techniques/processes described herein provide a non-invasive, automated evaluation of the volumetric flow dynamics and geometry of the abdominal vasculature, both arterial and venous, from the level of the Aortic and IVC bifurcations cranially to the right atrium of the heart. Further, cardiac output and pulmonary function may also be monitored.
Step 6 describes the use of image analysis to identify structures such as the aorta, the IVC, renal arteries, veins including hepatic veins, the right atrium, and the diaphragm. A number of factors could potentially be used to identify the relevant structures, such as Doppler to identify the direction and velocity of blood flow with the IVC flow being cranial and the aorta flow being caudal. In another embodiment, the movement and/or thickness of the vessel walls could be used for differentiation, with the IVC having a more mobile, thinner wall, while the aorta is a thicker less mobile wall. Another differentiator may include the temporal dynamics of the flow or vessel wall signals, with the IVC containing strong respiratory and cardiac elements while the aorta has a more predominant cardiac influence. In step 7, one or more anatomical landmarks are identified in the 3D image stack. Anatomical landmarks may identified using multiple techniques including, but not limited to edge detection and/or Doppler in the identification of branches. In step 8, a single ultrasound slice containing the IVC, located a known distance from the previous identified landmark/s, is chosen. Having identified a structure, or a combination of them, the device can be used to make additional measurements. Additional measurements might include Doppler velocity, direction of blood flow, pulsatility, distances, pulmonary B-line measurements and audio respiratory data.
It will be appreciated that such additional metrics may be used in the overall/final assessment of the vessel. For example, blood velocity from Doppler could be incorporated into an alert in combination with IVC dimension data. Blood velocity profile (speed and direction) over the respiratory and cardiac cycles could be informative in relation to fluid status. For example retrograde flow in the IVC may be indicative of valvular regurgitation while flow dynamics may provide information on congestion. Velocity can be measured if Doppler ultrasound is deployed. The knowledge of IVC segment in the brightness ultrasound image allows for extracting velocity profiles for IVC in conjunction with measures of Doppler ultrasound information.
In step 9, the single slice chosen is analysed to identify the IVC vessel wall using techniques including but not limited to edge detection and gradient analysis. Step 10 determines the IVC cross sectional area. In one embodiment, this may be determined by calculating the cross-sectional area of an ellipse fitted to the image of the IVC as defined in
The following are examples of additional metrics which may be recorded:
As listed above in respect of step 6, the pulsatility of identified structures or direction of flow may be used for example to identify the Aorta and IVC to be blood vessels. A Doppler velocity evaluation would show there to be blood flow in these two large vessels, with the Aorta identified as the vessel having a higher maximum blood flow velocity. The direction of blood flow in the vessels would also serve to confirm the identity of the vessels, the Aorta having a blood flow direction from the heart in an opposite direction to the IVC. The technique may also evaluate the shape of the vessels, the Aorta appearing circular and the IVC more ellipsoidal. This detail would be recorded.
Having identified the IVC and Aorta in the initial transverse image, the same identification process is carried out on further 2D B-mode images acquired in the same plane and on images acquired above and below the initial 2D plane, to serve as confirmation of the IVC and Aorta locations. Once the IVC and Aorta are identified, the 3D transducer array begins to map the IVC and Aorta and surrounding area using b-mode ultrasound. Mapping is conducted for different times in the respiratory cycle. The anatomical map could be used as a basis for estimating the volume of the vessels.
Multiple 3-d volume and Doppler maps are acquired over multiple respiratory and cardiac cycles.
Pulsed Wave or Colour Doppler ultrasound could subsequently be used to estimate the blood flow velocity at the centre of the identified vessels (identified as the maximum flow rate) and at the edges of the vessel walls (as the minimum flow rate), from which a mean flow rate is recorded. Combining the volume and the blood flow velocity, the volumetric flowrate at each point along the IVC and Aorta could be estimated.
The system could prompt the patient to inhale and exhale with the respiratory cycles automatically identified from the ultrasound images, through examination of the IVC length between the hepatic veins and the right atrium. With inspiration the increased intrathoracic cavity pressure causes a flattening of the IVC, therefore it becomes possible to use the ultrasound device to determine collapsibility of the IVC.
Once the IVC and Aorta have been identified from the 3D image it is possible to identify veins and arteries leading into Aorta and IVC. For example, the ultrasound response of a kidney being at one end of a vein can identify the vein as a renal vein. The flow direction and expected distances from other vessels could also be used to confirm identifications. In another example, identification of the right atrium and then the diaphragm can be used to confirm that the next lumens below that entering the IVC are therefore the hepatic veins.
Ideally the length of the IVC imaged includes branching of the hepatic and renal arteries from the liver and kidneys, respectively. Therefore, evaluating the blood flow velocity profile and vessel volume profile, combined, provides a measure of the volumetric flow rate along the IVC. This helps to characterise the level of fluid responsiveness and to indicate the congestion contribution from each of the liver and kidneys.
Algorithms are then used to determine diagnoses. For example, the processor could compare results to previous results and make therapeutic decisions/recommendations based on pre-programmed criteria and limits. Automated algorithms are used to determine specific diagnoses. These may be adapted from existing guidelines (ref IVC collapsibility in echo guidelines) or may solely be enabled based on the present invention.
The algorithms undergo an initial machine learning/training phase in order for reproducibility to be ensured. In this training phase, the placement of the transducer is carried out on multiple occasions, establishing a range of depths/shapes/flow rates of the landmark structures.
Upon the completion of the diagnosis portion the data is then communicated by the communications interface from the module to the cloud via GSM, wireless, Bluetooth® or similar. The raw data and diagnosis data are both transferred. In another embodiment a subset of the data is transferred in order to minimize the communication burden.
In one embodiment, the data is stored in the cloud and is accessible by nursing or clinical staff via a web portal. Alerts are pre-programmed based on specific thresholds being exceeded and physicians are alerted of a change in patient status, thus facilitating a modification in therapy.
Other volumetric related conditions and therapies such as dialysis, ultrafiltration, pulmonary hypertension, and hypertension may also benefit from the portable ultrasound device of the present invention. In the dialysis and ultrafiltration applications, the ultrasound module could be linked to the dialysis filtration machine thus providing a feedback loop to monitor the physiological variables and modulate the parameters on the machine to optimize the dialysis process.
In the hypertension examples the device could be used to predict elevated pressures in the patient via metrics such as the IVC collapsibility and this information communicated to the clinical team to inform the treatment of the patient.
Another potential use for the present invention is to monitor the progression of aortic aneurysms. These weakened, enlarged sections of the aorta can be potentially deadly if they rupture. This device would facilitate the daily monitoring of any changes/progression/growth/dilation of the aneurysm remotely.
In accordance with the present invention, it is also possible to discern the extent to which the congestion of the liver and/or kidneys contributes to the overall HF condition. Additionally, monitoring the structures and flowrates of the IVC, portal vein, renal veins and aorta provide additional detail on the function of the cardiovascular system and fluid status of a patient. Further, information on the respiratory rate, its variability and pattern can prove useful in identifying progression and specific type of HF. Continuous monitoring of the arterial and venous function, in addition to pulmonary function, provides insight into the progression of the disease and its side effects.
The present invention further provides systems and methods for marking ultrasound images to detect the IVC and other structures of the vascular system. Manual placement of markers on the outline of the vessel in the image is not only slow but is also adding additional variability to the measurement. The present invention utilizes an automatic algorithm to mark the vessel while supervision is possible in order to intervene if the fit result is insufficient. In the semi-automated embodiment, the first image of every sequence is marked manually, and corrective action entails manually marking the vessel at rare occasions. In the fully automated version user input is not required.
For each function (lumen IVUS, strut IVUS, and lumen EXUS) three data sets were chosen as test datasets for which manually marked trace data of the cross-sectional vessel area and required marking time was available.
The validation result showed that the new marker technique marks 3-6 images per second while manually marking requires at least 10 seconds per image. The difference of the means between manual and semi-automatically marked images was 14 mm2, the standard deviation from the mean was 6 mm2 for the semi-automatic marker while it was 115 mm2 for the manual marking approach.
Images of the cross-section of the inferior vena cava (IVC) acquired using external ultrasound (EXUS) and intravascular ultrasound (IVUS) are a useful resource to know the native behaviour of the IVC and check that the area measurements taken by the FIRE1 sensor are accurate enough. However, marking the outline of the inferior vena cava (IVC) in these images could be a long and tedious task if done manually. Moreover, the noise level in this type of image is higher than in others, which makes that totally automatic marking potentially introduces a lot of errors.
A Graphical User Interface (GUI) was developed to enable a semi-automatic marking which allows performing automatic marking with real-time correction possibility in case of errors in the marking. Three algorithms are included in this GUI which enables the marking of the struts of the sensor in IVUS (Struts Marker), the marking of the IVC lumen in the IVUS (Lumen Marker) and the marking of the IVC lumen in external US images (External US marker). The advantages of this tool can include accuracy, speed, and more human independence, among others; which facilitate the work of the operator.
We hypothesize that the semi-automatic marking has also lower probability of introducing errors than the other type of markers (Table 1). It will be demonstrated/verified herein that this tool is not only better in terms of usability for the user, but also in terms of accuracy.
Three datasets of US/IVUS images for each algorithm of the GUI were selected as examples of small, medium, and large IVC area (9 datasets in total). IVUS images were extracted from pre-clinical experiments in sheep, whereas external ultrasound was sourced from anonymized human data. All participants provided written informed consent to use their data for further analysis.
The marking concepts described herein can be applied to a number of different ultrasound imaging techniques, including IVUS and EXUS. IVUS imaging can produce cross-sectional views of the vessel as per
EXUS imaging can also produce similar cross-sectional views, as well as longitudinal sections of the vessel. The techniques described below can be applied to IVUS and cross sectional EXUS images.
Manual marking and Semi-Automatic marking with the GUI of the cross-sectional vessel area were performed in all these datasets as shown in
Manual Marking was done in 1 every 5 images for all the algorithms, which means that some temporal resolution is lost. The operator identifies the IVC at a glance and marks it with points. The IVC area extracted from the images is the area of the oval defined by the marked points.
Semi-Automatic marking was performed for all the images by means of the developed GUI. The operator marks the first image, and the algorithm automatically identifies the subsequent images. In case of mismarking, the operator can always pause the algorithm, rectify the marking, and continue automatically.
The process starts by operator marking of the outline of the IVC in a first image of a set of images. The operator would be presented with the first image in a time series and be prompted to mark the outline of the IVC in the image. The semi-automated system would then determine the threshold to find edge matching marked edge using a rotating line and gradient approach. The system uses the location of eight marked locations that must be placed on the edge of the IVC in the image. The centre point of those locations will be determined by fitting an ellipse to the coordinates. Then a line is drawn through the center point and each marked position. The gradient of the pixel values along the line is computed and the edge is determined as the location at which the gradient is the largest. The threshold for the gradient is then determined as a percentage of the found value in order to locate the edge in the subsequent images. The system would then proceed to perform sequential rotating line and gradient analysis of the image to mark the outline of the remaining images. This involves the system taking a line from outside the marked IVC wall to the centre of the IVC. The greyscale values of the image along this line are evaluated and the point at which the largest change in these values is identified. This colour change identifies the vessel wall, and the system records the location of this point. The line is then rotated a number of degrees and the process repeated to identify the next point of maximum colour gradient and this point is recorded as another point on the vessel wall. This process is repeated in angular increments up to 360 degrees and the points identified are joined to define the IVC wall.
The outline of the IVC is then marked automatically on the consecutive and remaining images. An operator can observe these results and intervene in the case of any mismarking. The outline may be remarked to improve accuracy.
When the data set is complete an ellipse is fitted to the outline of the IVC. The shape is transformed into a set of Cartesian coordinates with x- and y-position. An ellipse is fitted using a minimum norm approach.
Finally, vessel dimensions such as major and minor axis, and area are extracted from the resulting images. The area and the diameters of the minor and major axis of the IVC, plus the area of the outlined region may be extracted. As the frame rate of the ultrasound is defined and constant, this can be used to plot the area and diameters over time. These results can then be presented in a graphical representation over time such as shown in
The semi-automated nature of this system relates to the partial involvement of the user in the marking process. This user involvement can be in the marking of the initial IVC image or in other embodiments could be in the marking of points inside and outside the vessel in initial images, it could also involve the reviewing of images as they are marked by the system as a quality control step.
An automated embodiment is also described where the marking process is fully automated. In this instance the centre of the image is automatically detected as it is the location of the IVUS transducer (@half width of image and half height of image, using an assumed diameter of 3 mm of the catheter plus artefact in conjunction with the information of the dataset on pixel size) and the marking process is completed without the intervention of the user. It is envisaged that additional quality check process steps in the system could be added to increase the accuracy of the automated version.
The IVUS marking systems described above could also be used in case where an implant or sensor is placed within the vessel (
The algorithm then automatically finds the outline of the struts and presents to the operator an image with all strut outlines plotted. This is performed by using a rotating gradient of a line approach. A threshold is then determined for each strut to be used in consecutive images. The algorithm finds the outline of struts and respective centre points automatically in subsequent images of the set of images. By using the centre points of the struts, the lines are spread out equally, therefore any individual marking error is reduced. In other embodiments the marked locations can be unevenly spread. An operator can intervene in the process if any of the strut locations are mislocalised, in which case the operator can manually update the centre strut locations.
When all of the strut locations have been marked, again an oval could be fit to the shape and dimensions extracted and a trace over time outputted. From the strut locations the algorithm fits an ellipse and returns area, minor and major axis diameters as a function of respective frame times. In one embodiment, to estimate the true outer area of the sensor and considering that the counterpoints of the struts are localises, 0.5 mm is added to the minor and major diameters at both ends to compute an area that is most reflective of the outer sensor volume as this represents the distance between the centre of the struts and the vessel wall.
The struts' location is known from the first image. Only a sub-section close to the marked location is extracted (˜0.5 cm by 0.5 cm). To find the pixel value threshold to be used for each location the threshold is varied from 0.4 to 0.9 in 0.1 steps in images with pixel values scaled between 0 and 1. The images are binarized using the various possible thresholds. An edge detection filter is used to find the number of pixels in the vicinity of the marked strut location (within 1 mm). The threshold is found by finding the region with ˜1 mm diameter in 1 mm distance to the marked strut location. This threshold is variable for each strut. Once again, this system could be made automatic in a similar way to that described above.
As shown in
With reference to
This process is repeated for each subsection. Resulting in binary maps of the strut location (black and white maps) as shown in
As shown in
The algorithm is explained in the flow diagram of
In other embodiments a similar gradient-based approach could be used to mark longitudinal EXUS images and extract time series data on the measured diameter of a vessel.
The marking process of the present invention may result in the generation of an IVC trace depicting the motion of the IVC overtime (as shown in
Where an intravascular ultrasound system is used, a pullback reconstruction of the IVC is possible as the intravascular ultrasound system is withdrawn (pulled-back) at a constant speed through the IVC (as shown in
In accordance with some embodiments, neural networks are used in the marking process to identify the IVC.
With reference to
As shown in
To validate the tool, the difference between the area extracted from manual and the markers was computed for all the data points. Mean and standard deviation were extracted and compared to the ideal results, which should be as close to 0 as possible. These measurements are used to ensure that the markings and calculations performed by the algorithms in the GUI are reliable and accurate.
An area trace was extracted from each of the three datasets of small, medium, and large area from each of the manual and GUI marked data sets and compared (
Assessing the error between the expected area from ground truth marking to the area returned by the following methods was found to be 8.13±3.15 mm2 for the Struts Marker, 10.59±4.62 mm2 for the Lumen Marker and 20.36±17.95 mm2 for the External US Marker (Table 4).
The semi-automatic markers GUI have been a great development in terms of speed with respect to the manual marking and the users' feedback has been positive in all cases, improving and increasing the productivity time.
The validation results show that the GUI is reliable, and the measurements stays within ±10% of the area in all the algorithms, however the external US has more variability and wider range of error than the one extracted from IVUS.
The present invention further provides longitudinal temporal monitoring of the IVC using external ultrasound to repeatedly/conveniently capture the same cross-sectional slice while person undergoes manoeuvres.
The images may be obtained from a body worn or internal ultrasound imaging system. A portable ultrasound device which may be used to obtain the ultrasound images according to one aspect of the present invention is shown in
Ultrasound transducer 2 comprises a 2D array of independently controllable ultrasound transducer elements 6 for producing an ultrasonic pulse. Ultrasound transducer 2 is capable of 2D and 3D B-mode scanning in addition to functional Doppler modes such as Pulsed Wave, Colour and Power Doppler. The external 2D array of ultrasound transducer elements enables 3D imaging of the entire abdominal cavity, and assessment of both the anatomical structure and blood volumetric flowrates of the arterial and venous vessels, which would ideally be measured in the home by the patient, for long term monitoring by a clinical team.
In the embodiment shown in
The ultrasound transducer 2 in this embodiment extends to approximately 10 cm in length, and approximately 5 cm in width with 128×64 piezo crystals/CMUT cells. Depending on patient size the dimensions of the ultrasound transducer and the number of cells could vary (10 cm×5 cm, 12 cm×6 cm, 14 cm×7 cm . . . etc.). The transmission frequency range of the ultrasound transducer varies between 2-8 MHz, and this could be tuned in the initial training phase, dependent on the size and shape of the patient. Prior art ultrasound transducers/probes typically have an N×N square aperture. In the present embodiment, a rectangular transducer aperture (longer than the conventional smaller footprint transducers) allows improved extended-angle imaging capability of structures/organs of interest related to cardiac output and congestion.
It is intended that the patient would periodically affix the ultrasound transducer 2 to the body throughout a monitoring period of about one week. The ultrasound transducer 2 in this embodiment is secured to the torso by a pad 8 formed from an adhesive/gel/sponge like substance. The pad 8 provides an ultrasound transfer medium between the transducer and the skin.
An alternative embodiment of pad in shown in
Adhesion of the transducer 2 to the skin is not essential to the invention, therefore other forms of securement of the transducer to the torso are also envisaged.
As shown in
As shown in
The processing means comprises a TX beam former for controlling the timing of each of the ultrasonic elements and to assist in the beam steering, in order that plane waves at angles to the transducer may be swept across the abdomen and anatomical structures of interest, as shown in
The image processor may further be configured to compensate for changes in transducer position relative to the target abdominal structures, such that readings remain accurate regardless of transducer position.
The system of the present invention may further comprise a single element 1-D array transducer for imaging the lungs, examining for fluid (B-Lines) due to congestion. This may be used with, or form part of, the device of the present invention. The single element 1-D array transducer may be used to image B-lines. It may be connectable via a cable, to control module 4 or may be configured to communicate wirelessly with control module 4, for example using Bluetooth®. Its use would be particularly advantageous if there was a risk of pulmonary congestion. The patient would manually apply the 1-D array ultrasound transducer to a number of sites (as per
The system may further comprise an auscultation device, in the form of a simple microphone receiver, and/or Forced Expiration Volume (FEV), embedded in a simple face mask, for connection to control module 4. This device would record sounds from the lungs over a respiratory cycle and also for forced expulsions/Valsalva in order to establish the extent of congestion (the extent of crackle/wheezing of the lungs would be analysed). In addition, The FeV1/FEV ratios could be measured. These could be compared to a baseline set of results combining both the B-line analysis and the audio congestions/FEV analysis.
An ultrasound transducer such as that described above may be positioned and secured on the patient. To enable monitoring, the ultrasound transducer is placed on the same site of the patient daily or for a period of days. This could be assisted with the use of tattooing of the skin for repeatable placement of the transducer. Specific placement of the transducer could be decided upon on a patient-by-patient basis, depending on their anatomical profile and determined during an initial training/educational phase (for example in the hospital in consultation with healthcare professional). For remote monitoring, measurements are taken at the same time each day, with the patient in the same position, to ensure repeatability.
A device may be provided that auto locates the same IVC or SVC segment over time.
A mechanical or similar set-up that ensures that the device is over same approx. area (for example, this might be a belt with a dedicated holder & straps to ensure repeatable positioning).
Alternatively, or additionally, inked (permanent or the likes of semi-permanent) markers on the patient's body could be used for positioning. The device could recognise it's in the correct location using optical recognition of say 3 tattooed dots on the patient's body. This may be assisted by the device comprising a handheld probe shaped like a computer mouse.
A plurality of tattooed dots (or other shapes—such as asymmetric triangles such that both location and orientation can be precisely tracked)—either permanently or temporary could be used. These might be stencilled on in a doctor's office or be applied via ink on an adhesive positioned by a trained operator. Orientation is key to maximise possibly of achieving a consistent biplanar slice through the IVC.
In use, the holder for that day's reading may be lined up with those dots and then affixed using some type of adhesive. Any holder set-up straps etc. could be released and the individual let breathe freely during the reading.
The above may be coupled with feature extraction (e.g., liver, kidney, diaphragm, vessel branch or atrial structure recognition or other anatomical recognition) to identify the precise CSA required. This could be used with a 3D section or a 2D section.
Such a device would be able to locate this same IVC CSA/3D volume consistently over extended durations—months to years—and capture during relative motion of respiration cycles (combination of device itself moving with respiration and image analysis).
Such a device could trigger a series of prompts for patient to engage in appropriate manoeuvres (either a set sequence and/or selected using some decision-making criteria). One example is as follows:
An extended duration patch may be provided for continuous monitoring. It may have a rechargeable battery and/or may fit into a holder.
The words “comprises/comprising” and the words “having/including” when used herein with reference to the present invention are used to specify the presence of stated features, integers, steps, or components but do not preclude the presence or addition of one or more other features, integers, steps, components, or groups thereof.
It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2022/075458 | 9/13/2022 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63243370 | Sep 2021 | US |
