Method for determining a target location on a body

Abstract

The present invention relates to a method for determining a target of a body using machine learning by an electronic computing device, the method comprising: a first step of obtaining image data of a subject's body by an ultrasonic probe; a second step of analyzing the image data obtained in the first step by machine learning so as to specify the position of a target and the position of an avoidance target; and a third step of determining the position of a final target by referring to information of the target and the avoidance target specified in the second step. The third step is a step of determining the final target based on at least one of first information based on the size of the target, second information based on the depth of the target, and third information indicating whether an avoidance target exists in a straight line path between the target and the skin surface.

Description

Technical Field

[0001] The present invention relates to a method for determining the location of a target of a body. In particular, the present invention relates to a method for determining the location of a blood vessel, such as a superficial vein, into which a venipuncture needle is inserted for blood collection or intravenous injection. In particular, the present invention can use machine learning to determine the location of a target. Background Art

[0002] In hospitals or clinics, needles are often inserted into superficial veins for blood collection or intravenous injection. Needle insertion is performed by medical experts such as doctors, nurses, pathologists. The current system requires a long time to train medical personnel and incurs labor costs. In addition, if the needle is not inserted in the proper position, the medical staff makes additional attempts to insert the needle into the body, and the person into whom the needle is inserted may experience unnecessary pain due to the additional needle insertion. Even experts have difficulty inserting needles into people with poor vascular conditions (e.g., infants, the elderly, patients undergoing chemotherapy, people with dark skin, etc.).

[0003] Even for experienced medical personnel, repeated blood collection work can cause high fatigue and a lot of discomfort. In addition, when blood splashes on the medical personnel or enters the medical personnel's body during the blood collection process, the medical personnel who perform blood collection are exposed to infection risks.

[0004] The following Korean patent discloses an example of an automated process for blood collection.

[0005] Prior art references

[0006] Korean Patent No.10-1601421 (published on March 10, 2016) Summary of the invention

[0007] Technical Purpose

[0008] An object of the present invention is to provide a method for determining a target of a body by accurately detecting the position of, for example, a superficial vein (target) through machine learning (eg, convolutional network learning) to insert a needle into the center coordinates of the target.

[0009] Technical Solution

[0010] The present invention provides a computer-implemented method, which includes: a first step of obtaining image data of a body part from an ultrasonic probe; a second step of analyzing the image data using machine learning and identifying targets and avoidance objects; and a third step of determining the position of a final target with reference to information on the identified targets and avoidance objects.

[0011] The final target is determined by at least one of first information regarding the size of the target, second information regarding the depth of the target, and third information regarding the presence of an object in a straight path between the target and the skin surface.

[0012] The first information may be at least one of information about the size of the maximum inscribed circle of the target, information about the lengths of horizontal and vertical lines drawn in the target, information about the distance between the center of gravity of the target and the boundary line of the target, and information about the area of ​​the target.

[0013] The image data may be at least one of: first image data obtained between a state where the ultrasound probe is a predetermined distance away from the body and a state where the ultrasound probe is in contact with the body; and second image data obtained when the ultrasound probe presses the body.

[0014] The target may be a superficial vein.

[0015] When multiple targets are identified, the present invention may also include step 3-1-1, calculating at least one of the first information to the third information, and selecting a candidate for the final target based on the comparison result of the calculated information of the identified targets; and step 3-1-2, determining the final target by executing the third step on the candidate for the final target.

[0016] In the case where there are multiple targets, step 3-1-1 may include a step in which first information of the targets are compared with each other; the target that best meets the criteria is selected as a candidate for the final target; however, when the difference in the comparison result between the candidate and the second satisfied target is within a predetermined range, a target at a predetermined position closer to the contact surface of the ultrasonic probe is selected as a candidate for the final target.

[0017] The present invention may include a fourth step of determining whether the boundary of the target intersects with the boundary of the image data. If the boundaries intersect each other, the first step to the third step are performed after the ultrasonic probe is moved so that the boundary of the target is within the image data.

[0018] The image data may include acoustic wave signal image data of blood flow analyzed by Doppler effect.

[0019] According to the present invention, there is also provided a computer-implemented method for calculating the coordinates of the center of the final target. The method comprises: a fifth step of calculating the maximum inscribed circle including only the pixels of the final target; and a sixth step of determining the point with the maximum distance from the coordinates (xi) of the pixels of the maximum inscribed circle as the center of the final target.

[0020] The method may also include a seventh step of generating a circle with a predetermined diameter for each pixel inside the final target; and an eighth step of classifying the pixel as a pixel of the final target if the circle drawn for the pixel including the pixel of the final target is greater than or equal to a predetermined part, otherwise excluding the pixel from the final target.

[0021] The method may further include the step of excluding the candidate from the final target in a case where a straight line between the calculated coordinates of the center and the skin crosses the blood vessel equal to or greater than a predetermined number of times.

[0022] The first image data may include image data obtained before the upper arm is pressed and image data obtained after the upper arm is pressed; and the second image data includes acoustic wave signal image data of blood flow analyzed by Doppler effect.

[0023] The machine learning used in the method of the present invention may be convolutional neural network learning. The convolutional neural network learning is performed by a method including a ninth step of learning processed data and a tenth step of obtaining characteristic information of the target and the avoidance object and storing the information, wherein the versatility of the processed data is increased by data expansion.

[0024] The characteristic information may be at least one of the intensity of echogenicity when the ultrasound probe comes into contact with the body, a distribution pattern of echogenic components, relative positions, body information of the subject, and information on size changes.

[0025] Effects of the Invention

[0026] According to the present invention, machine learning can be used to accurately determine the location of a target in the body. Therefore, since the insertion position of the needle can be accurately determined and the needle can be inserted into the position through the present invention, automation of blood collection and intravenous injection can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a diagram of a blood collection device that performs a method for determining a location of a target according to the present invention;

[0028] Figure 2 is a flow chart of a method for determining the location of a superficial vein into which a needle is inserted according to the present invention;

[0029] Figure 3 is a flow chart of a detailed process for determining an ultimate goal according to the present invention;

[0030] Figure 4 is a flow chart of a process for determining the coordinates of the center of a final target in accordance with the present invention;

[0031] Figure 5 is a diagram for explaining a process of softening a boundary of a target according to the present invention; and

[0032] Figure 6 is an exemplary diagram of an electronic computing device that implements the present invention. DETAILED DESCRIPTION

[0033] Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

[0034] In this specification, if necessary, encryption / decryption can be used to perform the transmission / reception of information (data). It should be understood that, although not specifically mentioned, the transmission / reception described in this specification can be performed using encryption / decryption. In addition, the transmission (forwarding) from A to B or the reception from A to B includes a process via an additional medium, not limited to direct transmission or reception. The order of each step should be understood in a non-restrictive manner, unless the previous step must be performed before the subsequent step logically and in time. That is, except for the exceptions as described above, although the process described as the previous step is before the process described as the subsequent step, this does not affect the essence of this disclosure, and the scope of the right should be defined regardless of the order of the steps. In addition, in this specification, "A or B" is not only defined as selectively referring to A or B, but also defined as including both A and B. In addition, in this specification, the term "comprising" has the meaning of further including other components other than the listed components.

[0035] The term "module" refers to a logical combination of general hardware and software that performs the required functions.

[0036] In this specification, the essential elements of the present invention will be described, while the non-essential elements will not be described. However, the scope of the present invention should not be limited to the invention including only the described components. In addition, it should be understood that inventions including additional elements or without non-essential elements may be within the scope of the present invention.

[0037] Certain exemplary embodiments will now be described to provide a comprehensive understanding of the principles of structure, function, manufacture and use of the apparatus and methods disclosed herein. One or more examples of these embodiments are shown in the accompanying drawings. It will be understood by those skilled in the art that the apparatus and methods specifically described herein and shown in the accompanying drawings are non-limiting exemplary embodiments, and the scope of the present invention is limited only by the claims. The features shown or described in conjunction with an exemplary embodiment may be combined with the features of other embodiments. These modifications and variations are all included within the scope of the present invention.

[0038] The method of the present invention may be an electronic computing device. The electronic computing device may be a device such as a computer, a tablet computer, a mobile phone, a portable computing device, a fixed computing device, a server computer, etc. In addition, it should be understood that one or more of the various methods or aspects thereof may be performed by at least one processor. The processor may be implemented on a computer, a tablet computer, a mobile device, a portable computing device, etc. A memory configured to store program instructions may also be implemented in (one or more) devices, in which case the processor is specifically programmed to execute the stored program instructions to perform one or more processes, which will be further described below. In addition, it should be understood that the following information, methods, etc. may be performed by a computer, a tablet computer, a mobile device, a portable computing device, etc. including a processor in combination with one or more additional components, as described in detail below. In addition, the control logic may be implemented as a non-transitory computer-readable medium on a computer-readable medium, which contains executable program instructions executed by a processor, a controller / control unit, etc. Examples of computer-readable media include, but are not limited to, ROM, RAM, compact disc (CD)-ROM, tape, floppy disk, flash drive, smart card, and optical data storage device. The computer readable recording medium can also be distributed in network coupled computer systems so that the computer readable media is stored and executed in a distributed fashion, eg, by a telematics server or a Controller Area Network (CAN).

[0039] Figure 6 An example schematic diagram of an exemplary device architecture according to an embodiment of the present invention is shown. Figure 6 As shown, the device 609 may include multiple components, including but not limited to a processor (e.g., a central processing unit (CPU); 610), a memory 620 (also referred to as a "computer-readable storage medium"), a wired or wireless communication unit 630, one or more input units 640, and one or more output units 650. It should be noted that Figure 6The architecture depicted in is simplified and provided for demonstration purposes only. According to the claims, one of ordinary skill in the art will understand that the architecture of device 609 can be modified in any suitable manner. In addition, according to the claims, as one of ordinary skill in the art will understand, the components of device 609 themselves can be modified in any suitable manner. Therefore, Figure 6 The device architecture described in the present invention should be taken as an example only and should not be taken as limiting the scope of the present invention.

[0040] The processor 610 is capable of controlling the operation of the device 609. More specifically, the processor 610 is operable to control and interact with a plurality of components installed in the device 609, such as Figure 6 As shown. For example, the memory 620 can store program instructions and data that can be executed by the processor 610. The process described herein can be stored in the memory 620 in the form of program instructions for execution by the processor 610. The communication unit 630 can allow the device 609 to transmit data to one or more external devices via a communication network and receive data from one or more external devices. The input unit 640 can enable the device 609 to receive various types of inputs, such as audio / visual input, user input, data input, etc. To this end, the input unit 640 can be composed of a plurality of input devices for accepting various types of inputs, including, for example, one or more cameras 642, a touch panel 644, a microphone (not shown), a sensor 646, a keyboard, a mouse, one or more buttons or switches (not shown), etc. The input device included in the input unit 640 can be manipulated by the user. The output unit 650 can display information on the display screen 652 for the user to view. The display screen 652 can also be configured to accept one or more inputs by various mechanisms known in the art, such as a user tapping or pressing the screen 652. The output unit 650 can also include a light source 654. The device 609 is shown as a single component, but the device may also be composed of multiple separate components that are connected together and interact with each other during use.

[0041] Certain exemplary embodiments will now be described to provide a comprehensive understanding of the principles of structure, function, manufacture and use of the apparatus and methods disclosed herein. One or more examples of these embodiments are shown in the accompanying drawings. It will be understood by those skilled in the art that the apparatus and methods specifically described herein and shown in the accompanying drawings are non-limiting exemplary embodiments, and the scope of the present invention is limited only by the claims. The features shown or described in conjunction with an exemplary embodiment may be combined with the features of other embodiments. These modifications and variations are all included within the scope of the present invention.

[0042] In this specification, embodiments where the target is a superficial vein are mainly described. However, it should be understood that the target may be any body organ, such as an artery, a deep vein, etc. The scope of the present invention is not limited to the method for determining the location of a superficial vein.

[0043] Figure 1 1 is a block diagram of a blood collection device 1 of the present invention, which performs a method for determining the location of a superficial vein and inserting a needle into the determined location of the vein. For ease of explanation, inserting a needle for blood collection is described in this specification. However, it should be understood that the present invention can also be used for other needle insertion purposes, such as intravenous injection, etc.

[0044] The blood collection device 1 of the present invention includes a central processing module 10, a control module 20, an image data processing module 30, a communication module 40, a display module 50, an operation module 60, a storage module 70, a first driving unit 80, an ultrasonic probe 90, a second driving unit 100, a needle unit 110 and an upper arm pressing unit 120.

[0045] Those skilled in the art can make appropriate modifications to the device 1 based on the appended claims according to its application. Each component of the device 1 can also be appropriately modified by those skilled in the art based on the appended claims. Figure 1 The structures shown in the drawings are merely exemplary and should not be construed as limiting the scope of the present invention.

[0046] The central processing module 10 performs arithmetic operations of the present invention and controls the overall operation of the device. The control module 20 controls the first driving unit 80 that moves the ultrasonic probe 90 and the second driving unit 100 that moves the needle unit 110. The central processing module 10 and the control module 20 may be combined into one component.

[0047] The image data processing module 30 performs conversion of image data obtained by the ultrasonic probe 90 into data having a format and size that can be analyzed by the present invention.

[0048] The communication module 40 enables the device 1 to communicate with the machine learning server 140 so that the image data processed by the image data processing module 30 can be analyzed by the server.

[0049] The display module 50 displays information notifying the status of the device to the operator visually, audibly, etc. The display module 50 may be a monitor, a speaker, etc.

[0050] The operation module 60 may be a button, a keyboard, a microphone for receiving oral commands, etc.

[0051] The storage module 70 can store program codes for operating and controlling the blood collection device 1 and data related to the blood collection device. The storage module 70 can be a hard disk, an SSD, a flash memory, etc. The storage module 70 can also be a cloud storage medium that can be remotely accessed through the communication network 130.

[0052] The first driving unit 80 drives the ultrasonic probe 90 according to a command of the control module 20, thereby moving the ultrasonic probe 90 in two dimensions or three dimensions. The first driving unit 80 may be provided in a single number or in a plurality of numbers.

[0053] The ultrasonic probe 90 approaches the lower arm of the subject; comes into contact with the lower arm; presses the lower arm; and obtains image data of the inside of the body.

[0054] The second driving unit 100 moves the needle unit 110 toward the center of the superficial vein identified by the method described below upon receiving a command from the control module 20. The second driving unit 100 may be provided in a single or plural number.

[0055] The needle is inserted into the blood vessel, and then blood is collected and / or intravenous injection is performed. If the purpose is not blood collection or intravenous injection, the needle unit 110 is not required.

[0056] The upper arm pressing unit 120 may include an annular member (not shown) that presses the upper arm to easily identify the superficial vein. After a person places his / her upper arm into the annular member, the diameter of the annular member is reduced, and the upper arm is pressed with a predetermined pressure in an automatic or manual manner. The upper arm pressing unit 120 may include a compression cuff of a conventional blood pressure monitor.

[0057] The machine learning server 140 performs machine learning for identifying targets (e.g., superficial veins) and avoidance objects. Machine learning can be performed by a PC, a smart phone, a tablet PC, a machine learning server, etc. The machine learning server 140 can analyze data obtained by a plurality of automatic blood collection devices 1 using learning data trained by a convolutional neural network, etc. According to an alternative aspect of the present invention, the automatic blood collection device 1 can have a lightweight machine learning that can be installed therein, and perform machine learning for identifying targets and avoidance objects.

[0058] Figure 2 is a flow chart of a method for determining the location of a superficial vein according to the present invention.

[0059] In step 200, the subject places the upper arm into the upper arm pressing unit 120, and then the unit 120 presses the upper arm. The unit 120 may include an annular belt with a width of 1 to 5 cm. The belt can press the upper arm 3 to 12 cm above the elbow, preferably 6 to 8 cm above the elbow. The pressure applied to the upper arm can be 10 to 60 mmHg, preferably 20 to 30 mmHg. The following process can be performed even without pressing the upper arm.

[0060] Despite Figure 2 , the upper arm is shown to be pressed before the ultrasound probe 90 moves, but the upper arm may be pressed after the ultrasound probe 90 moves and starts to acquire image data. If the upper arm is pressed after the ultrasound probe starts to acquire the first image data, the first image data may include image data acquired before the upper arm is pressed and image data acquired after the upper arm is pressed.

[0061] In step 205, according to the command of the control module 20, after the unit 120 presses the upper arm, the ultrasound probe 90 moves to a position separated from the body part by a predetermined distance, for example, 1 mm. The ultrasound probe 90 can be moved from the antecubital fossa toward the hand to the median cubital vein, which is 0 to 8 cm, preferably 1 to 3 cm.

[0062] In step 210, first image data is obtained while the ultrasound probe 90 is moved toward the body part. The first image data is obtained before the ultrasound probe 90 comes into contact with the body part. The first image data may include image data obtained before the upper arm is pressed and image data obtained after the upper arm is pressed.

[0063] The second image data is obtained when the ultrasound probe 90 is in contact with the body part and compresses the body. According to an alternative aspect of the present invention, the body may be compressed by an alternative means.

[0064] The first image data and the second image data may include image data analyzed by Doppler effect among the acoustic wave signals of the blood flow. According to another embodiment of the present invention, only one of the first image data and the second image data may be used.

[0065] In step 220, the image data processing module 30 converts the first image data and the second image data into data having a format and size that can be processed by the program code of the present invention. Alternatively, the image data may be processed without conversion.

[0066] In step 225, the machine learning server 140 analyzes the transformed image data using, for example, a convolutional neural network; and then identifies pixels of target and avoidance objects.

[0067] In this specification, a target may be a blood vessel such as a superficial vein or an artery; an avoidance object may be an artery, a nerve, a bone tissue, etc. The target and the avoidance object are defined differently according to the purpose of finding the target. An artery, a nerve, or a bone tissue considered as an avoidance object when the purpose is to find a superficial vein may be classified as a target for another purpose.

[0068] The target and avoidance objects may be displayed by pixels or bounding boxes. The bounding box may be generated using a center value, a border, and a distance. Information that may define the shape of the target or avoidance object may also be stored. The information may be a center value, information about the shape of the target or avoidance object, such as a circle / ellipse, rectangle, etc. The information may also include a radius, a distance between at least two points, a length of a side, etc.

[0069] The present invention using convolutional neural network learning (hereinafter referred to as "CNN learning") is described below.

[0070] In order to perform CNN learning, learning through data should be performed in advance. The image data used in CNN learning can be pre-processed using data augmentation techniques such as random cropping, size conversion, horizontal flipping of the first image data and the second image data to increase its versatility.

[0071] The machine learning server 140 obtains feature information of the target and avoidance object through CNN learning and stores the feature information.

[0072] Characteristic information may include the intensity of the echogenicity of components identified in the ultrasound image data; the distribution pattern of echogenic components; information about the relative positions of the distribution of components; height, weight, gender and age, morbidity, and information about past and / or present treatments of the person; information obtained by ultrasound probes and other devices during compressions; and real-time blood flow information analyzed by the Doppler effect.

[0073] When the skin surface is pressed with an ultrasonic probe or other device, the size of the vein is directly reduced, or the proximal portion is expanded due to the pressing, thereby causing congestion, which is different from the artery. The information obtained during the pressing process performed by the ultrasonic probe and other devices can include information on the state change of the vein, etc.

[0074] Since arteries beat periodically and their blood flows quickly, the Doppler signal from arteries is strong. Therefore, real-time blood flow information is useful for detecting targets such as superficial veins or arteries.

[0075] Based on the feature information, training through CNN learning can be performed to display pixels at the locations of target and avoidance objects.

[0076] Learning (training) can be supervised learning, unsupervised learning, or semi-supervised learning.

[0077] In supervised learning, training can be performed by using images with the correct answer. If the target and avoidance objects recognized in the image by CNN learning are different from the correct answer, learning can be performed so that the loss of the loss function decreases. The loss function can be BCE loss (Binary Cross Entropy loss), cross entropy, etc. The optimizer can be Adam optimizer, RMSprop, stochastic gradient descent, etc. Evaluation can be performed by Dice coefficient loss, etc.

[0078] After identifying the target and / or avoidance object by machine learning, such as CNN learning, step 230 is performed to determine the final target based on the identified object. Steps 230 and 235 can be performed by the central processing module 10. However, the steps can also be performed by a remote server that receives pre-processed data of the first image data and the second image data via the communication network 130.

[0079] Figure 3 A flowchart of the detailed process included in step 230 is shown.

[0080] In step 300, it is determined whether the target to be found in the previous step is identified. If the target exists, the number of targets is counted in step 305.

[0081] If the number of the identified targets is one, it is determined in step 335 whether the boundary of the target intersects with the boundary of the image. If they do not intersect each other, step 310 is performed. Otherwise, in step 355, the ultrasonic probe 90 is moved to another position, for example, to the center of the image, so that the coordinate value of the boundary of the target is inside the image. Subsequently, the process returns to step 210, and then the steps after step 210 are performed. The body of the subject may be moved instead of moving the ultrasonic probe 90.

[0082] In step 310, the diameter of the largest inscribed circle is calculated as one of the size-based information of the object. The largest inscribed circle means the largest circle including only the pixels of the object. The size-based information of the object may be the maximum / minimum length of horizontal and vertical lines that can be drawn in the object, the maximum / minimum distance between the center of gravity of the object and the boundary of the object, the area of ​​the object, etc., in addition to the diameter of the largest inscribed circle.

[0083] In step 315, the depth of the object as one of the depth-based information is calculated. In step 320, it is determined whether the object satisfies the first determination criterion and the second determination criterion. The depth-based information includes various information related to the depth.

[0084] In the case where the target is a superficial vein, the first determination criterion is whether the diameter of the largest inscribed circle is greater than or equal to a predetermined value (first value). For example, if the diameter of the largest inscribed circle of the identified target is less than 2 mm, the target can be classified as an avoidance object in step 375. If the diameter of the largest inscribed circle is greater than or equal to the first value, it is determined that the first determination criterion is met. The first value can be determined with reference to the device operation error range.

[0085] If the size-based information relates to horizontal and vertical lines that can be drawn in the object, the first determination criterion may be whether the maximum length and / or minimum length of the horizontal / vertical lines is greater than or equal to a predetermined value. For example, if the minimum length of the horizontal and vertical lines of the identified target is less than 2 mm, the target is classified as an avoidance target. If the target that meets the first determination criterion is not a single one, in step 370, the object having the maximum value in the sum, product, average, etc. of the length may be determined as a candidate for the final target.

[0086] If the size-based information is the maximum distance and the minimum distance between the center of gravity and the boundary line, the first determination criterion is whether the distance is greater than or equal to a predetermined value. For example, if the minimum distance between the center of gravity of the target and the boundary line of the target is less than 2 mm, the target can be classified as an avoidance object. If the target that meets the first determination criterion is not a single one, in step 370, the target having the maximum value in the sum, product, average, etc. of the distances can be determined as a candidate for the final target.

[0087] If the size-based information is the area of ​​the object, the first determination criterion is whether the area is greater than or equal to a predetermined value. For example, if the area of ​​the object is less than 10 mm 2 , the target can be classified as an avoidance object. If the target satisfying the first determination criterion is not single, in step 370, the object with the largest area can be determined as a candidate for the final target.

[0088] The second determination criterion is whether the depth of the object is less than or equal to a predetermined value (second value), for example, 1 cm. If the depth of the target is greater than the second value, the target can be classified as an avoidance object in step 375. This is to exclude deep veins, because hemostasis of deep veins is more difficult than hemostasis of superficial veins, and to exclude arteries as avoidance objects. In addition, if the depth of the target is greater than the second value, the needle may cause a deep wound when penetrating the body, and there may be important anatomical structures or nerves in the penetration path. This is why the second determination criterion is necessary. If the depth of the object is less than or equal to the second value, it is determined that the second determination criterion is met.

[0089] If the first determination criterion and the second determination criterion are satisfied, it is determined whether another object exists in the straight path between the candidate and the skin surface in step 325. This is referred to as the third determination criterion. The other object existing in the path may be a nerve as an avoidance object.

[0090] If the result of the determination in step 325 is “No”, the candidate is determined as the final target in step 330 .

[0091] Alternatively, a target satisfying at least one of the first determination criterion, the second determination criterion, and the third determination criterion may be determined as the final target.

[0092] An embodiment in which a plurality of targets are identified in step 225 will be described below.

[0093] If multiple objects are identified, the size of the maximum inscribed circle of each object is calculated in step 360. In step 365, the sizes of the maximum inscribed circles of each object are compared with each other, and then in step 370, the candidate of the final object is determined according to the third determination criterion.

[0094] The third determination criterion is as follows. In principle, the target having the largest size of its largest inscribed circle is selected as a candidate for the final target. However, if the difference between the largest size and the second largest size is within a predetermined value (third value), such as 10%, a target closer to the contact surface of the ultrasonic probe (e.g., a target closest to the center of the contact surface) may be selected as a candidate for the final target.

[0095] Even in an embodiment where the first information is other size-based information, if the difference in the compared sizes is within a predetermined value, a target closer to the contact surface of the ultrasound probe (e.g., a target closest to the center of the contact surface) can be selected as a candidate for the final target.

[0096] Step 315 and steps subsequent to step 315 are performed after selecting a candidate for the final target, and then it is determined whether the candidate is the final target or an avoidance object.

[0097] If it is determined in step 300 that the target does not exist, it is determined in step 303 whether the search for the target has been performed greater than or equal to a predetermined number of times. If it is determined that the search has been performed greater than or equal to a predetermined number of times, for example, 2 or 3 times, a search for the target is performed on the opposite arm in step 380. If it is determined that the search has been performed less than the predetermined number of times, in step 210, the ultrasonic probe 90 is moved to a position different from the last search, and then the first image data is obtained.

[0098] After determining the final target according to the above steps, the coordinate value of the center of the final target is calculated in step 235. In step 240, the second driving unit 100 receives a command from the control module 20 and moves the needle unit 110 to the center of the final target, where blood collection is performed.

[0099] The center of the final object is defined by the point x that has the maximum distance from any point xi inside the largest inscribed circle that includes only the pixels of the final object.

[0100] If a straight line between the center and the skin surface crosses the surface of the blood vessel greater than or equal to a predetermined number of times, for example, twice, the candidate for the final target may be excluded from the final target.

[0101] According to an optional embodiment of the present invention, a circle with a predetermined diameter (e.g., 1 mm) is drawn for all pixels inside the final target; then, if a predetermined proportion (e.g., 30% to 95%, preferably 50% to 70%) of the pixels within the circle are inside the final target, the pixel is determined to be included in the final target. Otherwise, the pixel is excluded from the final target.

[0102] like Figure 5 As shown, since all pixels in circle 520 are inside the final target, pixel 510 is determined as a pixel of the final target. Since a predetermined portion of pixels of circle 511 are not inside the final target, corresponding pixel 530 is excluded from the final target.

[0103] These processes are used to soften the boundaries of the final object and eliminate the possibility of misclassified pixels in the final object. However, these processes are optional and are not required in the present invention.

[0104] Although the present invention has been described with reference to the accompanying drawings, the scope of the present invention is determined by the claims and should not be interpreted as being limited by the above-mentioned embodiments and / or the accompanying drawings. It should be clearly understood that improvements, changes and modifications of the present invention disclosed in the claims and obvious to those skilled in the art also fall within the scope of the present invention. Therefore, this description is intended only as an example and not to limit the scope of the embodiments herein in other ways.

Claims

1. A computer-implemented method for determining a target of a body using machine learning, the method comprising: In a first step, image data of a body part is obtained from an ultrasound probe; In a second step, the image data is analyzed using machine learning to identify targets including blood vessels and avoidance objects; as well as The third step is to determine the position of the final target including the blood vessel based on at least one of the first information about the size of the identified target, the second information about the depth of the identified target, and the third information that the avoidance object exists in a straight path between the identified target and the skin surface. Wherein, the third step comprises: If the identified target is a plurality of targets, determining the position of the final target among the identified plurality of targets based on the position of the contact surface of the ultrasonic probe, The method further comprises a fourth step of determining whether a boundary of the target intersects a boundary of the image data, and Wherein, in a case where the boundaries intersect each other, the first step to the third step are performed after the ultrasonic probe is moved so that the boundary of the target is within the image data.

2. The method according to claim 1, wherein: The first information is at least one of information about the size of the maximum inscribed circle of the target, information about the lengths of horizontal and vertical lines drawn in the target, information about the distance between the center of gravity of the target and the boundary line of the target, and information about the area of ​​the target.

3. The method according to claim 1, wherein: The image data is at least one of the following: first image data, which is obtained between a state where the ultrasound probe is away from the body by a predetermined distance and a state where the ultrasound probe is in contact with the body; and second image data, which is obtained when the ultrasound probe presses the body.

4. The method according to claim 1, wherein: The target is the superficial vein.

5. The method according to any one of claims 1 to 4, wherein: In the case where multiple targets are identified, selecting a candidate for the final target based on the comparison result of the first information of the identified targets; And the final target is determined by performing the third step on the candidates of the final target.

6. The method according to claim 5, wherein: The first information of the targets are compared with each other; the target that best meets the criteria is selected as a candidate for the final target; however, when the difference in the comparison result between the candidate and the second satisfied target is within a predetermined range, a target at a predetermined position closer to the contact surface of the ultrasonic probe is selected as a candidate for the final target.

7. The method according to any one of claims 1 to 4, wherein: The image data includes acoustic wave signal image data of blood flow analyzed by Doppler effect.

8. The method according to claim 2, further comprising the steps of calculating the coordinates of the center of the final target: The fifth step is to calculate the maximum inscribed circle including only the pixels of the final target; and In the sixth step, the point having the largest distance from the coordinates (xi) of the pixel of the maximum inscribed circle is determined as the center of the final target.

9. The method according to claim 8, further comprising: The seventh step is to generate a circle with a predetermined diameter for each pixel inside the final target; And an eighth step, if the circle drawn for a pixel includes the pixel of the final target in a portion greater than or equal to a predetermined portion, classifying the pixel as a pixel of the final target, otherwise excluding the pixel from the final target.

10. The method according to claim 3, wherein: The first image data includes image data obtained before an upper arm is pressed and image data obtained after the upper arm is pressed; and the second image data includes acoustic wave signal image data of blood flow analyzed by Doppler effect.

11. The method according to any one of claims 1 to 4, wherein: The machine learning is convolutional neural network learning; the convolutional neural network learning is performed by a method including a ninth step of learning processed data and a tenth step of obtaining characteristic information of the target and avoidance objects and storing the information, and the versatility of the processed data is increased by data expansion; the characteristic information is at least one of the intensity of echo reflectivity when the ultrasonic probe comes into contact with the body, the distribution pattern of echo components, the relative position, the body information of the subject, and the information on size changes.

12. The method according to claim 8, further comprising an eleventh step of excluding the final target when a straight line drawn between the center of the final target and the skin surface passes through the surface of a blood vessel greater than or equal to a predetermined number of times.

Images