Ultrasound detection of thrombi in the bloodstream

Abstract

The systems and methods disclosed herein relate to the detection of irregular particles in the bloodstream based on the measured relative velocity of the particles suspended in the bloodstream and / or other characteristics of the particles suspended in the bloodstream, including the relative location of the particles within the blood vessel and the tendency of the particles to cluster with other particles suspended in the bloodstream. Based on the measured relative velocity and / or other relevant factors, characteristics of the irregular particles including size, shape, and frequency of the irregular particles in the bloodstream can also be measured. Machine learning techniques can be used to identify patterns in the behavior of irregular particles suspended in the bloodstream. These patterns can correspond to specific health risks and conditions.

Description

[Technical field]

[0001] The present invention relates generally to detection and analysis of objects in the bloodstream with ultrasound. In particular, some implementations may relate to pre-symptomatic detection and analysis of objects, including thrombi, using machine learning. [Background technology]

[0002] Anomalies in the bloodstream can pose serious health risks. Irregular objects and formations in the bloodstream, including blood clots, can be particularly dangerous. Some objects, such as blood clots, grow larger over time. As blood clots and other objects in the bloodstream grow larger, blood flow becomes more and more restricted, which increases the risk of serious health conditions such as stroke, pulmonary embolism, deep vein thrombosis, and other diseases. Currently, medical technology only allows for the detection of blood clots and other objects that are large enough to be directly observed using imaging techniques such as ultrasound or computed tomography (CT). By the time these objects become large enough to be detected, patients often already suffer from symptoms and other adverse health conditions. Thus, detection of small objects and blood clots is desirable because it allows physicians to identify and treat diseases early, before patients begin to experience adverse health outcomes. Summary of the Invention [Means for solving the problem]

[0003] Described herein are systems and methods for detection of particles in a blood stream, such as a thrombus, based on the relative velocity of the particle compared to the velocity of the blood flow. In addition to relative velocity, detection can also be achieved and / or assisted by examining the position of the particle and thrombus within a cross-section of the blood vessel, and the position of the thrombus relative to each other within the blood vessel. Changes in both particle velocity and position are detectable as solid particles move through a fluid. These velocity and position parameters are affected by the size, shape, and other properties of the particle.

[0004] The above detection can be achieved using an ultrasonic sensor. In one embodiment, an ultrasonic sensor can be used for detection in conjunction with color Doppler. Color Doppler technology uses pulsed wave Doppler with short pulses to generate a sequence of images of blood flow in a target area of ​​a blood vessel. The images can contain information about the presence and characteristics of particles, including thrombi, that are suspended in the bloodstream.

[0005] Machine learning can be used to extract relevant information from a sequence of images of blood flow. A machine learning model can be trained using collected data including a sequence of sample images of blood flow and known targets. The known targets can be particles or clots of a confirmed size or frequency in a particular sample. The known targets can also be other parameters. In one preferred embodiment, the known target can be the risk of a clot causing a particular health problem within a particular time frame. In one preferred embodiment, the machine learning model can be a neural network.

[0006] Other features and aspects of the invention will become apparent from the following detailed description taken in conjunction with the accompanying drawings, which illustrate, by way of example, features according to various embodiments. This Summary is not intended to limit the scope of the invention, which is defined solely by the claims appended hereto.

[0007] The technology disclosed herein, in accordance with one or more various embodiments, will be described in detail with reference to the following drawings. These drawings are provided for illustrative purposes only and merely represent representative or exemplary embodiments of the disclosed technology. These drawings are provided to facilitate the reader's understanding of the disclosed technology and should not be considered as limiting the breadth, scope, or applicability of the disclosed technology. It should be noted that for clarity and ease of illustration, these drawings are not necessarily drawn to scale.

[0008] These drawings are not intended to be exhaustive or to limit the invention to the precise forms disclosed, It is to be understood that the invention can be practiced with modification and alteration, and that the disclosed technology is limited only by the claims and their equivalents. [Brief description of the drawings]

[0009] [Figure 1] FIG. 1 illustrates an example of a vessel segment through which a thrombus travels at a different velocity compared to the blood flow velocity. [Diagram 2] FIG. 2 is a diagram showing an example of a cross-section of a blood vessel, illustrating the relative position of a thrombus within the cross-section. [Diagram 3] FIG. 2 shows an example of a cross-section of a blood vessel, illustrating their relative positions with respect to each other within the cross-section. [Figure 4] 1 is a flow chart of an example of a method for detecting abnormal blood flow. [Diagram 5] 1 is a flow chart of an example of a method for detecting abnormal blood flow. [Figure 6] FIG. 1 illustrates an example of a blood flow abnormality detection system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Detailed Description Early detection of thrombi and / or other abnormalities in the bloodstream can be life-saving. The systems and methods disclosed herein are directed to the detection of thrombi and other particles in the bloodstream, including detection of their frequency, size, relative location within the vessel, and other characteristics. These systems and methods employ non-invasive methods in which thrombi and other particles can be detected based on their observed hydrodynamic behavior. Thrombi and other particles having a diameter of 90 microns can be detected with the systems and methods described herein.

[0011] The technique is based on an important principle of fluid mechanics - under certain conditions, quanta traveling through a fluid do not travel at the same speed as the fluid itself. The speed of individual particles suspended in a fluid depends on the size of the particle and other factors. The movement of these particles within the fluid therefore produces a change in the frequency of waves relative to the velocity of the fluid. These changes in wave frequency can be detected and correlated to the presence, size, and frequency of particles within the fluid. This is known as the Doppler shift. JPEG2024529795000002.jpg40134

[0012] The Navier-Stokes and Newton-Euler equations are two important sets of equations that interact and combine to describe the flow of fluids containing solid particles. The Navier-Stokes equations describe fluid flow as follows:

number

[0013] When solid objects or particles are suspended in a fluid and move through the fluid, the Navier-Stokes equations must be combined with other equations that describe the motion of the suspended particles. The Newton-Euler equations describe the motion of such particles as follows:

number

[0014] An important parameter for these equations is the Reynolds number. The Reynolds number associated with a fluid describes how the fluid behaves, including properties such as inertial and viscous forces in the flow. In laminar flow, if the Reynolds number is less than 1, the particles indeed travel at the same speed as the fluid. However, once the Reynolds number exceeds a threshold of 1, there is a velocity difference. The Reynolds number of the fluid itself is generally denoted as Re. The flow around a particular particle that is suspended in the fluid has its own Reynolds number, denoted as Re_p. Re_p creates a velocity difference between the particle and the fluid in which it is suspended. This is referred to in current technology as "slip". In addition to this velocity difference, particles moving through a fluid can also move between the center of the fluid flow and the walls, and can tend to cluster together within the fluid flow.

[0015] The Reynolds number of an average blood flow is approximately 2000. Thus, the Reynolds number of blood is three orders of magnitude greater than the threshold of 1, allowing the differential velocity of particles moving through the blood flow to be detected. This detection can be achieved by imaging the blood flow using ultrasonic Doppler techniques. Each particle or thrombus moving through the blood flow is an "event" that only stays within range of the ultrasonic sensor for a short period of time, as the blood carries it away. Color Doppler techniques can be used to image the blood flow. In addition, the magnitude and frequency of the observable Doppler shift also depends on the size of the particle or thrombus. Thus, the Doppler signal can indicate multiple pieces of information about the particles in the blood. For example, the Doppler signal can indicate not only the presence of a particle or thrombus, but also characteristics such as frequency, size, location, and other factors. These additional factors can help distinguish thrombus and other factors from noise.

[0016] Thrombi and other particles in the bloodstream can be detected because they behave differently than the normal cells that compose the bloodstream. Thrombi and other particles exhibit distinguishable hydrodynamic behaviors; that is, while moving through the bloodstream, they travel at a different velocity than the surrounding bloodstream. Thrombi and other particles have a tendency to occupy specific locations within a vessel and specific locations relative to each other. These three behavioral patterns, namely (i) differential velocity, (ii) relative location within a vessel, and (iii) relative location to other particles in the bloodstream, provide for the detection of thrombi and their properties.

[0017] Individual thrombi and other particles moving through the bloodstream are visible as "events" in the Doppler spectrum as they pass through the range of the ultrasound probe. Measuring the Doppler shift relative to the frequency of the blood flow in this portion of the imaged vessel provides a strong reference point. In other words, the Doppler shift is measured relative to the center frequency. Thus, thrombi or other particles moving through the bloodstream can be detected with a high level of reliability even when there are changes in the blood flow. For example, changes in blood flow may exist depending on which portion of the patient's vessel is measured, whether the patient has recently eaten, and other factors. This thrombi detection is reliable in different patients who may have different baselines of blood flow.

[0018] Machine learning techniques can be used to identify and interpret patterns consistent with these events. Machine learning techniques can be used to distinguish events from spectra, thereby confirming the presence of thrombi or other particles in the bloodstream. Machine learning techniques can also be used to interpret patterns of events to characterize the size and frequency of different particles or thrombi in the bloodstream. Although principles of fluid mechanics support the idea that there is a detectable event when thrombi or other particles are present in the bloodstream, there is no closed-form solution for this event. In other words, although there are patterns for the different velocities, preferred locations, and tendencies to cluster of thrombi and other objects, these patterns have not been and cannot be rigorously identified beyond highly simplified example scenarios by observation and traditional mathematical techniques alone.

[0019] Although the embodiments described below relate to thrombus detection, the techniques described herein can be used to detect any particle or object in the bloodstream that poses a health risk. For example, these techniques can enable the detection of foreign or abnormal objects in the bloodstream, such as cancer cells, and events indicative of other health conditions. Thus, machine learning models are important for detecting and identifying characteristics of thrombus or other particles. Machine learning models can be trained to effectively detect and characterize thrombus or other particles.

[0020] 1 is a diagram illustrating an example of a blood vessel segment through which a thrombus travels at a different velocity compared to the blood flow velocity. The blood vessel segment 100 includes normal blood cells 102, 104 and a thrombus 106. Within the segment 100, the blood cells 102, 104 and the thrombus 106 together form a blood flow 108. The blood flow 108 can have a blood flow velocity 112. The thrombus 106 can be moving within the blood flow 108. The thrombus 106 can be moving at a velocity 110 that is different than the blood flow velocity 112.

[0021] 2 is a diagram illustrating an example of a blood vessel cross-section and the relative location of a thrombus within the cross-section. The blood vessel cross-section 206 includes normal blood cells 102, 104 as well as a thrombus 106. The blood vessel cross-section has a radius 200. The thrombus 106 may occupy a particular location relative to a center 208 and an outer wall 210 of the blood vessel cross-section 206. The location of the thrombus 106 may be described as the distance 202 along the radius 200 of the thrombus 106 relative to the outer wall 210 of the blood vessel cross-section 206. The location of the thrombus 106 may also be described as the distance 202 along the radius 200 of the thrombus 106 relative to a center 218 of the blood vessel cross-section 206.

[0022] 3 is a diagram illustrating an example of a blood vessel cross-section, showing the relative positions of blood cells relative to one another within the cross-section. The blood vessel cross-section 206 includes normal blood cells 102, 104 as well as thrombi 106, 300. The relative positions of the thrombi relative to one another can be expressed as the distance 302 between the two thrombi 106, 300. The blood vessel cross-section 206 can also include multiple thrombi, each of which can occupy a position relative to one another.

[0023] FIG. 4 is a flow diagram of an example of a method for detecting blood flow abnormalities. An ultrasound probe can be used to generate a sequence of images of blood flow 400 in a target region 402 of a blood vessel. The target region 402 can be an area in the body where blood flow is desired to be measured. For example, the target region 402 can be in a blood vessel in an arm or leg of a patient. The images can be generated using different types of medical imaging techniques. For example, a color Doppler method can be used to generate the sequence of images. From the images, a center frequency 404 of the blood flow 400 in the target region 402 of the blood vessel can be measured. The center frequency 404 is a Doppler frequency shift corresponding to the entire blood flow. A difference frequency 406 of one or more particles 408 suspended in the blood flow 400 can also be measured. Particles moving in the blood flow can have a different frequency than the surrounding blood flow. This frequency difference is detectable from the sequence of images of the blood flow.

[0024] The difference frequency can be analyzed to determine characteristics of the blood flow and any particles suspended within the blood flow. For example, analysis of the measured difference frequency 406 can indicate the presence of irregular particles in the blood flow 404. The irregular particles may be thrombi.

[0025] FIG. 5 is a flow diagram of an example of a method for detecting blood flow anomalies. An ultrasound probe can be used to generate a sequence of images of blood flow 400 in a target region 402 of a blood vessel 412. The target region 402 can be an area in the body where it is desired to measure blood flow 400. For example, the target region 402 can be a blood vessel 412 in an arm or leg of a patient. The images can be generated using different types of medical imaging techniques. For example, a color Doppler method can be used to generate the sequence of images. From the images, a center frequency 404 of blood flow 400 in the target region 402 of the blood vessel 412 can be measured. The center frequency 404 is a Doppler shift that corresponds to the total blood flow.

[0026] Several characteristics of the blood flow 400 can be measured. These characteristics can reveal important information about the blood flow 400, which can indicate whether a medical risk exists. The difference frequency 406 of one or more particles 408 suspended within the blood flow 400 can also be measured. Particles moving within the blood flow may have a different frequency than the surrounding blood flow. This frequency difference is detectable from a sequence of images of the blood flow.

[0027] The relative position 402 of the particle 408 within the cross section 410 of the blood vessel 412 can be measured. Specifically, the position of the particle 408 can be measured along a radius 414 of the cross section 410 of the blood vessel 412. For example, the distance between the particle 408 and the wall of the blood vessel 412 can be measured. Alternatively, or in addition, the distance between the particle 408 and the center 208 of the blood vessel 412 can be measured. Irregular particles, such as thrombi, in the blood flow 400 can have a tendency to become suspended in certain regions of the blood flow 400. Thus, the detected position of the particle 408 can provide information about the blood flow 400 that can correspond to certain types of irregularities, which can correspond to certain types of health risks.

[0028] A clustering factor 416 may also be measured. The clustering factor may quantify the relative position of particles 408 in the blood flow 400 relative to other particles 418. Some types of particles 408, 418 suspended in the blood flow 400 may have a tendency to be located close to each other in the blood flow 400. In one embodiment, the types of particles 408 may be thrombi. Thrombi tend to cluster together in blood vessels. The measured clustering factor may provide information about the blood flow 400, which may correspond to certain types of irregularities, which may correspond to certain types of health risks. For example, a measurement of thrombi grouped very close to each other may indicate a significant risk of stroke.

[0029] The difference frequency 406, relative position 420, and clustering coefficient 416 can be analyzed to determine characteristics of the blood flow 400 and any particles 408, 418 suspended within the blood flow 400. For example, the measured difference frequency 406, relative position 420, and clustering coefficient 416 can indicate the presence of irregular particles in the blood flow 400. The irregular particles can be thrombi.

[0030] The difference frequency 416, relative position 402, and clustering coefficient 416 may be further analyzed to determine characteristics of particles detected in the bloodstream. For example, these coefficients may indicate the size, shape, and / or frequency of particles suspended in the bloodstream. These coefficients may also correspond to a particular medical condition and / or risk of a particular adverse medical event. For example, analysis of the relative position of a blood clot within a cross-section of a blood vessel may correspond to a risk of an adverse medical event, such as a stroke, or may indicate that a blood clot has reached a particular size. Further analysis of the clustering coefficient may similarly correspond to a risk of an adverse medical event, such as a stroke.

[0031] FIG. 6 illustrates an example of a blood flow anomaly detection system. The blood flow anomaly detection system can include an ultrasonic sensor 600. The ultrasonic sensor 600 can generate an image sequence 602 of blood flow 604 in a target region 606 of a blood vessel 608. The blood flow anomaly detection system can also include a set of anomaly detection parameters 610. A few examples of relevant anomaly detection parameters can be the relative velocity 632 of a particle in the blood flow compared to the velocity of the blood flow itself, the relative position 634 of a particle in the blood flow, and the relative position of a particle in the blood flow to other particles in the blood flow. The values ​​of these parameters can indicate important information about the blood flow, including the presence of abnormal or irregular particles in the blood flow, and the characteristics of any irregular particles, including the shape, size, and frequency of the irregular particles.

[0032] All of these anomaly detection parameters 610 can be measured from a high quality image sequence of the blood flow through the vessel. In one embodiment, the high quality image sequence can be generated using color Doppler technology. The ultrasonic sensor 600 can use pulsed wave Doppler with short pulses to generate an image sequence 602 of the blood flow 604 that includes the anomaly detection parameters 610.

[0033] The blood flow anomaly detection system may also include one or more sample data sets including training data 614. The sample data sets may be sample image sequences 616, 620, 624 of sample blood flows that are associated with known parameters of interest 618, 622, 626. The known parameters 618, 622, 626 may be known values ​​of anomaly detection parameters. For example, the known parameters 618, 622, 626 may correspond to known values ​​of size, shape, and / or frequency of irregular particles detected in the sample image sequences. The detected irregular particles may be thrombi. The known parameters 618, 622, 626 may also correspond to risk thresholds established for a particular health condition. For example, the known parameters 618, 622, 626 may correspond to samples having thrombi where there is a greater than 50% chance of stroke.

[0034] The blood flow anomaly detection system can include a machine learning model 612. The machine learning model 612 can be trained using training data 614. The trained machine learning model can then apply the trained parameters to analyze an image sequence 602 of blood flow 604 within a target region 606 of a blood vessel 608. The machine learning model can determine information about the blood flow 604, including the presence of any irregular particles in the blood flow and a characteristic of any detected irregular particles. The machine learning model can generate output data 642. The output data 642 can include values ​​of the characteristic of the irregular particles detected in the blood flow. For example, the output data 642 can include a measured size 628 of an anomaly 630 present in the blood flow 604. The output data 642 can include a measured shape 638 of an anomaly 630 present in the blood flow 604. The output data 642 can include a measured frequency 640 of anomalies 630 present in the blood flow 604.

[0035] In one embodiment, the blood flow anomaly detection system of FIG. 6 can be specifically configured to detect a thrombus in the blood flow. The thrombus detection and classification system can include an ultrasonic sensor. The ultrasonic sensor can generate an image sequence of the blood flow in the target area. The thrombus detection and classification system can also include a set of thrombus detection parameters, which include: (i) velocity; (ii) intravascular location; and (iii) thrombus location, where velocity is the relative velocity of the thrombus in the blood flow to the velocity of the blood flow itself, intravascular location is the relative location of the thrombus in the blood flow along the radius of the blood vessel, and thrombus location is the relative location of the thrombus in the blood flow to other particles in the blood flow. The ultrasonic sensor in the thrombus detection and classification system can use a pulsed wave Doppler technique with short pulses to generate an image sequence of the blood flow including the thrombus detection parameters.

[0036] The thrombus detection and classification system can also include one or more example databases, the example database including a sequence of images of sample blood flows and corresponding known values ​​of the thrombus detection parameters. The thrombus detection and classification system can also include a machine learning model, the machine learning model trained with the one or more example datasets to detect the presence of thrombus in the blood stream based on the thrombus detection parameters, measure the frequency of thrombus in the blood stream, measure the diameter of thrombus in the blood stream, and measure the shape of thrombus in the blood stream.

[0037] Although various embodiments of the present invention have been described above, it should be understood that these embodiments have been presented by way of example only and not by way of limitation. Similarly, various figures may depict examples of architectural configurations and other configurations of the present invention, and are presented to aid in understanding the features and functionality that the present invention may include. The present invention is not limited to the illustrated architectural or configuration examples, and a wide variety of alternative architectures and configurations may be used to achieve the desired features. Indeed, it will be apparent to one skilled in the art how alternative functional, logical, or physical partitioning and configurations may be realized to achieve the desired features of the present invention. Also, many different component module names other than those depicted herein may be applied to the various partitions. In addition, with respect to flow diagrams, operational descriptions, and method claims, the order in which steps are presented herein does not mandate that the various embodiments be implemented to perform the recited functions in the same order, unless the context otherwise dictates.

[0038] While the present invention has been described above with reference to various preferred embodiments and implementations, it should be understood that various features, aspects, and functions described in one or more of the individual embodiments are not limited in application to the particular embodiment in which such features, aspects, and functions are described, but instead may be applied, either alone or in various combinations, to one or more of the other embodiments of the present invention, whether or not such an embodiment is described, and whether or not such features are presented as part of the described embodiment. Thus, the breadth and scope of the present invention should not be limited by any of the preferred embodiments described above.

[0039] Terms and phrases used in this document, and various variations thereof, unless expressly stated otherwise, should be construed as open ended as opposed to limiting. As examples of the above: "including" should be read to mean "including without limitation" and the like; "examples" are used to provide preferred examples, not an exclusive or exclusive list, of the items being described; "a" and "an" should be read to mean "at least one," "one or more," and the like; adjectives such as "conventional," "traditional," "usual," "standard," "known," and words of similar import should not be construed to limit the items described to those items available at a given period or time, but instead should be read to encompass conventional, traditional, ordinary, or standard technology that may be available now or at any time in the future. Similarly, when this document refers to technology that would be apparent or known to one of ordinary skill in the art, it encompasses what would be apparent or known to one of ordinary skill in the art now or at any time in the future.

[0040] In some instances, broadening phrases such as "one or more," "at least one," "but not limited to," or other similar phrases should not be read to imply that a narrower case is intended or required in some instances where such broadening phrases are not present. Use of the term "module" does not imply that all of the components or functionality described or claimed as part of the module are configured within a common package. Indeed, any or all of the various components of a module, whether control logic or other components, may be combined within a single package or maintained separately, and may be distributed in multiple groupings or packages or across multiple locations.

[0041] In addition, various embodiments described herein are set forth in the form of exemplary block diagrams, flow charts, and other illustrations. As will be apparent to one of ordinary skill in the art after reading this document, the illustrated embodiments and various alternatives thereof can be implemented without limitation to the illustrated examples. For example, the block diagrams and the accompanying description should not be construed as mandating a particular architecture or configuration.

Claims

1. capturing a sequence of images of blood flow within a target region of a blood vessel using an ultrasound imaging sensor that uses pulsed wave Doppler with short pulses; Based on the image sequence, a center frequency corresponding to a first Doppler frequency shift of the entire blood flow within the target region of the blood vessel as seen by the ultrasound sensor; a second Doppler frequency shift seen by the ultrasonic sensor corresponding to one or more objects suspended within the blood flow; and measuring determining a differential velocity between the one or more objects suspended within the blood flow and the blood flow as a whole based on the center frequency and the second Doppler frequency shift; Based on the image sequence, determining the position of the one or more objects within the cross-section of the blood vessel; and measuring the distance between the one or more objects and one or more other objects present in the blood flow within the cross-section; measuring the relative positions of the one or more objects within the cross section of the blood vessel by performing an operation consisting of: detecting, based on the differential velocity, the location, and the distance, that the one or more objects correspond to one or more abnormal objects present in the blood flow; A method comprising:

2. The method of claim 1 , further comprising determining a size of the one or more abnormal objects present in the blood flow based on the differential velocity and the measured relative position.

3. The method of claim 2 , further comprising the step of verifying that at least one of the one or more abnormal objects has a diameter no greater than 90 microns.

4. The method of claim 1 , further comprising determining a shape of the one or more abnormal objects present in the blood flow based on the differential velocity and the measured relative position.

5. The method of claim 1 , wherein the ultrasound imaging sensor constructs the image sequence using color Doppler.

6. 10. The method of claim 1, further comprising training a machine learning model to detect the presence of the abnormal object in the blood flow, wherein the machine learning model is trained using one or more collected datasets, the one or more collected datasets including a sequence of sample images of the blood flow in the target region and corresponding values of a parameter of interest.

7. The method of claim 6 , wherein one of the parameters of interest is the size of the anomalous object in a controlled experiment.

8. The method of claim 6 , wherein one of the parameters of interest is the frequency of the anomalous object in a controlled experiment.

9. The method of claim 6 , wherein the parameter of interest corresponds to a risk threshold associated with a particular health problem.

10. The method of claim 9 , wherein one of the parameters of interest is the size of the abnormal object, which corresponds to the size of a thrombus associated with a stroke.

11. The method of claim 6 , wherein the machine learning model is a cognitive neural network.

12. The method of claim 1 , further comprising treating a medical condition corresponding to the one or more abnormal objects detected in the bloodstream.

13. detecting that the one or more objects correspond to one or more abnormal objects present in the blood stream includes detecting that the one or more objects correspond to the one or more abnormal objects present in the blood stream based on the differential velocity, the position, and the distance using a trained machine learning model; 2. The method of claim 1, wherein the machine learning model is trained using a dataset including a plurality of image sequences of a sample blood flow and corresponding known values, the known values including a differential velocity between an object suspended in the sample blood flow and the sample blood flow, a position of the object suspended in the sample blood flow relative to the blood vessel, and a distance between the object suspended in the blood flow and one or more other objects suspended in the blood flow.

14. 14. The method of claim 13, further comprising using the trained machine learning model to determine a size of the one or more abnormal objects present in the blood flow based on the differential velocity and the measured relative position.

15. 15. The method of claim 14, further comprising: using the trained machine learning model to confirm that at least one of the one or more anomalous objects has a diameter no greater than 90 microns.

16. 14. The method of claim 13, further comprising using the trained machine learning model to determine a shape of the one or more abnormal objects present in the blood flow based on the differential velocity and the measured relative position.

17. 14. The method of claim 13, further comprising using the trained machine learning model to determine a frequency of the one or more abnormal objects present in the blood flow based on the differential velocity and the measured relative position.

18. The method of claim 13 , wherein the machine learning model is a neural network.

19. 2. The method of claim 1, wherein detecting that the one or more objects correspond to one or more abnormal objects present in the blood stream comprises detecting that the one or more objects correspond to the one or more abnormal objects present in the blood stream based on the differential velocity, the location, the distance, and one or more risk thresholds associated with the differential velocity threshold, the location threshold, and the distance threshold.

20. the one or more abnormal objects are one or more thrombi; detecting that the one or more objects correspond to one or more abnormal objects present in the blood stream includes detecting the presence of the one or more thrombi in the blood stream based on the differential velocity, the location, and the distance using a trained machine learning model; the machine learning model is trained using a dataset including a plurality of image sequences of a sample blood flow and corresponding known values, the known values including a differential velocity between the sample blood flow and a thrombus suspended in the sample blood flow, a position of the thrombus suspended in the sample blood flow relative to the blood vessel, and a distance between the thrombus suspended in the blood flow and one or more other thrombi suspended in the blood flow; The method of claim 1.

Images