KR-20260065555-A - FLUID DIAGNOSIS METHOD USING ULTRASOUND B-MODE IMAGE

KR20260065555AKR 20260065555 AKR20260065555 AKR 20260065555AKR-20260065555-A

Abstract

The present invention relates to a method for diagnosing a substance based on ultrasonic images. According to the present invention, a method for diagnosing a substance based on ultrasonic images is performed by a processor of a medium characteristic inspection device that generates an ultrasonic image using a return signal for an ultrasonic signal transmitted through a pipe section in which a medium and a reflector are arranged, and analyzes the characteristics of the medium passing through the pipe section. The method comprises: a correction step of training a machine learning model that determines medium characteristics by using a statistical feature value extracted from a measurement section (ROI) within an image re-captured in the corrected state as input, after analyzing an image captured with the setting variables of the inspection device set to a default value and correcting the setting variables; and a measurement step of outputting the material characteristics of the medium as a diagnosis result using the trained machine learning model from a new ultrasonic image acquired for the measurement medium based on the corrected setting variables of the inspection device and the trained machine learning model.

Inventors

김형범

Assignees

경상국립대학교산학협력단

Dates

Publication Date: 20260508
Application Date: 20251030
Priority Date: 20241101

Claims (8)

A material diagnosis method performed by a processor of a medium characteristic inspection device that generates an ultrasonic image using a return signal for an ultrasonic signal transmitted through a pipe section in which a medium and a reflector are arranged, and analyzes the characteristics of the medium passing through the pipe section, A correction step of analyzing an image captured with the setting variables of the inspection device set to default values, correcting the setting variables, and training a machine learning model that determines medium characteristics using statistical feature values extracted from a measurement section (ROI) within an image recaptured in the corrected state as input; and An ultrasound image-based material diagnosis method comprising a measurement step of outputting the material properties of a medium as a diagnosis result using the trained machine learning model from a new ultrasound image acquired for a measurement medium, based on the corrected setting variables of the inspection device and the trained machine learning model.
In claim 1, The above correction step is, A step of correcting the setting variables by correcting the sound velocity of the medium through the current ultrasonic signal and adjusting the image gain based on histogram analysis when the position of the reflector is confirmed within the ultrasonic image obtained by setting the setting variables of the inspection device to default values; A step of setting the measurement region (ROI) within the ultrasound image recaptured in a corrected state and calculating statistical feature values for the image within the measurement region; and An ultrasonic image-based material diagnosis method comprising the step of training a machine learning model that determines the presence or absence of abnormalities or physical property values of a medium using the above feature values as input.
In claim 2, The step of correcting the above setting variables is, A method for diagnosing a substance based on an ultrasound image, wherein if the position of the reflector is confirmed within the ultrasound image, sound velocity correction of the medium is performed using a signal returned from the reflector, and if the position of the reflector is not confirmed, the position of the reflector within the pipe section is adjusted using a reflector adjustment unit until the position of the reflector is confirmed, and the process of reacquiring the ultrasound image is repeated.
In claim 2, A plurality of reflectors are arranged spaced apart from each other within the above-mentioned piping section, and The step of correcting the above setting variables is, A method for diagnosing a material based on an ultrasound image, further comprising the step of checking the alignment state of the ultrasound probe with respect to the pipe section by calculating the inclination of the ultrasound probe based on the relative positional difference between two adjacent reflectors among the plurality of reflectors before adjusting the image gain.
In claim 1, The statistical characteristic values within the above measurement range (ROI) are, A method for diagnosing a substance based on ultrasound imaging, comprising at least one of a statistical value in the depth direction, a statistical value in the width direction, a curve fitting slope of the depth direction intensity value, and an intercept value.
In claim 1, The above machine learning model is an ultrasound image-based material diagnosis method implemented as at least one of a decision tree, a support vector machine (SVM), a neural network, and a K-nearest neighbor (KNN).
In claim 1, In the above correction step, An ultrasound image-based material diagnosis method that performs principal component analysis (PCA) on statistical feature values during the training of the machine learning model to reduce feature dimensions, and repeatedly trains the machine learning model using the reduced feature values.
In claim 1, The above machine learning model is an ultrasonic image-based material diagnosis method that outputs at least one of whether the medium is abnormal, the particle size distribution of the medium, the solid content concentration, the viscosity, or the moisture content by deep learning analysis of input feature values.

Description

Fluid Diagnosis Method Using Ultrasound Imaging {FLUID DIAGNOSIS METHOD USING ULTRASOUND B-MODE IMAGE} The present invention relates to a method for diagnosing a material based on ultrasonic images, and more specifically, to a method for diagnosing the characteristics of a medium by applying machine learning to an ultrasonic image acquired from a medium characteristic inspection device. In general, diagnosing the presence of large particles that are not ground or are bound together during the process of mixing various components in a mixer after grinding them to a size below a certain threshold is important in various technical fields such as chemistry, biology, pharmaceuticals, and environmental monitoring. In particular, carbon, graphite, and other metal components are mixed in the negative or positive electrode materials of secondary batteries, and if there are large particles that are not well mixed, a low-quality battery may be manufactured. Therefore, it is important to diagnose the presence of particles larger than a certain size, but it is difficult to diagnose the presence of particles in an opaque fluid. As a method for measuring particles in such fluids, ultrasound is used to determine the presence and concentration of particles. However, while the ultrasonic method calculates the fluid flow rate by estimating the maximum fluid velocity through the measurement of the maximum frequency of the Doppler power spectrum generated by the movement of particles within the fluid during measurement, there is a problem in that other related parameters, such as the concentration of fluid components, cannot be accurately estimated due to limitations in accurately estimating the maximum frequency of the Doppler power spectrum. The technology forming the background of the present invention is disclosed in Korean Published Patent No. 10-2021-0104733 (August 25, 2021). FIG. 1 is a diagram illustrating the configuration of a medium characteristic inspection device to which an embodiment of the present invention is applied. Figure 2 is an enlarged view of section 'A' shown in Figure 1. Figure 3 is a configuration diagram showing a plurality of reflectors and a reflector adjustment unit installed with respect to Figure 1. FIG. 4 is a diagram showing the procedure of an ultrasound image-based material diagnosis method according to an embodiment of the present invention. Figure 5 is a diagram specifically explaining each step of Figure 4. Then, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification have been given similar reference numerals. Throughout the specification, when a part is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. The present invention proposes a technology capable of diagnosing the physical properties or the presence of abnormalities of a measurement medium by analyzing ultrasonic images acquired from a medium characteristic inspection device based on image statistics and machine learning. Prior to the detailed description of the technology of the present invention, an example of the configuration of a medium characteristic inspection device to which an embodiment of the present invention is applied is described based on FIGS. 1 to 3. Referring to FIGS. 1 to 3, a medium characteristic inspection device (hereinafter, inspection device) to which an embodiment of the present invention is applied may include a piping part (100), a reflector (200), an ultrasonic transceiver (300), and a processor (400), and may inspect the physical characteristics (particle size, dispersion, viscosity, solid content concentration, moisture content, etc.) of particles in the medium based on ultrasonic images. The piping section (100) can guide the medium to flow in one direction while containing the medium to be measured inside. Of course, the medium may also be placed in a stationary state. The medium present in the piping section (100) may be a fluid mixed with particles. The above piping section (100) may be equipped with a mixing tank (110) for storing a medium in which particles are mixed, a circulation line (120) and a pump (130) for guiding the medium stored in the mixing tank (110) to circulate between the piping section (100) and the mixing tank (110). A stirrer may be provided inside the mixing tank (110) to agitate the medium and particles to prevent sedimentation of particles within the medium. The pipe section (100) may be provided with an incident surface (10