Cardiac ultrasound PACS-based image processing system using artificial intelligence

Abstract

The present invention relates to a cardiac ultrasound PACS-based image processing system using artificial intelligence, comprising: a reception unit that receives cardiac ultrasound images; an image extraction unit that extracts standard cross-sectional images of the cardiac ultrasound images by using an artificial intelligence-based cardiac ultrasound image examination model that has been trained by an image processing unit with the received images on the basis of deep learning; a measurement value extraction unit that calculates examination measurement values through artificial intelligence analysis of the extracted images; a report generation unit that integrates the images and the measurement values to generate an artificial intelligence result report; a feedback unit that inputs and feeds back final interpretation results from medical staff concerning the artificial intelligence result report; and a data processing unit that converts result values input by the feedback unit into deep learning training data and transmits the deep learning training data to the image extraction unit and the measurement value extraction unit.

Description

Artificial intelligence-based cardiac ultrasound PACS-based image processing system

[0001] The present invention relates to an artificial intelligence-based cardiac ultrasound PACS-based image processing system, and relates to an artificial intelligence-based cardiac ultrasound PACS-based image processing system that analyzes cardiac ultrasound images using artificial intelligence based on PACS and provides them to medical staff to assist in their analysis.

[0002]

[0003] Medical diagnosis using ultrasound is widely used in the medical field to obtain information about the inside of a subject because it is non-invasive and non-destructive and can provide images in real time.

[0004] A cardiac ultrasound examination is a test that acquires images of the heart in real time and simultaneously determines the presence or absence of cardiac abnormalities. It uses ultrasound to image and analyze the movement, structure, blood flow, valve abnormalities, tumors in the heart, and blood clots, etc. of the heart, and diagnoses and evaluates heart disease.

[0005] Meanwhile, diagnosing heart disease requires obtaining quantitative information, particularly regarding cardiac function. This information includes left ventricular hypertrophy, stroke volume, ejection fraction, and cardiac output. Recently, interest has been focused on technologies that calculate blood flow velocity vectors and quantify blood flow vorticity based on these data for diagnostic purposes.

[0006] Since the movement of the heart muscle is usually determined by the examiner's visual judgment in a typical cardiac ultrasound examination, the examination results may vary depending on the examiner's skill level. The examination takes 30 to 40 minutes depending on the medical staff's ultrasound examination skill level.

[0007] Accordingly, technologies are being developed to supplement the objectivity of manual measurements by cardiac echocardiographers, shorten the time of ultrasound examinations, easily analyze cardiac ultrasound, or utilize artificial intelligence. Technologies such as Korean Patent Publication No. 2023-0055004 (Title of the Invention: Method and Device for Augmenting Cardiac Echocardiography Image Data) have been developed, but skill levels depending on the examiner are still required, so there is a need for reliable technologies that shorten the cardiac ultrasound examination time.

[0008]

[0009] The present invention was invented to solve the problems of the prior art as described above, and the purpose of the present invention is to provide an artificial intelligence-based cardiac ultrasound PACS image processing system that can automatically extract images necessary for cardiac ultrasound analysis without relying on the skill of an examiner, thereby shortening the examination time.

[0010] In addition, the purpose is to provide an artificial intelligence-based cardiac ultrasound PACS image processing system that can eliminate the visual error rate of an examiner analyzing an imaged cardiac image by analyzing the cardiac ultrasound image extracted through an artificial function again using artificial intelligence.

[0011]

[0012] The present invention provides a cardiac ultrasound PACS-based image processing system utilizing artificial intelligence, comprising: a receiving unit for receiving an examination ultrasound image; an image extraction unit for extracting a standard cross-sectional image of a cardiac ultrasound image through an artificial intelligence cardiac ultrasound image examination model learned based on deep learning in an image processing unit from the received image; a measurement extraction unit for calculating an examination measurement value (Measurement) through artificial intelligence analysis of the extracted image; a report generation unit for generating an artificial intelligence result report integrating the extracted image and the examination measurement value; a feedback unit for inputting a final interpretation result of a medical staff for the artificial intelligence result report and providing feedback; and a data processing unit for converting the result value input from the feedback unit into deep learning learning data and transmitting the converted result to the image extraction unit and the measurement extraction unit.

[0013] In addition, the cardiac ultrasound images received from the receiving unit include a parasternal extension axis cross-section image (PLAX), a parasternal shortening axis cross-section image (PSAX), an apical four-chamber cross-section image (A4C), an apical two-chamber cross-section image (A2C), and an apical three-chamber cross-section image (A3C), and a cardiac ultrasound PACS-based image processing system utilizing artificial intelligence is provided.

[0014] In addition, the image extraction unit provides an artificial intelligence-based cardiac ultrasound PACS-based image processing system characterized in that it extracts an image after segmenting the image in one cardiac cycle.

[0015] In addition, the report generation unit provides an artificial intelligence-based cardiac ultrasound PACS-based image processing system characterized by including an image processed with DICOM properties and an artificial intelligence interpretation result.

[0016]

[0017] As described above, the artificial intelligence-based cardiac ultrasound PACS image processing system according to the present invention can automatically extract images necessary for cardiac ultrasound analysis without relying on the skill of the examiner through artificial intelligence, thereby reducing the examination time.

[0018] Additionally, it has the effect of eliminating the visual error rate that occurs when an examiner analyzes an imaged cardiac image by analyzing the extracted cardiac ultrasound image again using artificial intelligence.

[0019]

[0020] Figure 1 is a block diagram functionally illustrating a cardiac ultrasound PACS-based image processing system utilizing artificial intelligence according to the present invention.

[0021] Figure 2 is a block diagram showing the operation process of a cardiac ultrasound PACS-based image processing system utilizing artificial intelligence according to the present invention.

[0022] FIG. 3 is a drawing showing an exemplary screen for segmenting and extracting images from a cardiac ultrasound image in an image extraction unit of a cardiac ultrasound PACS-based image processing system utilizing artificial intelligence according to the present invention.

[0023] FIG. 4 is a drawing showing an exemplary screen for analyzing an image in a measurement value extraction unit of an image processing system based on artificial intelligence for cardiac ultrasound PACS according to the present invention.

[0024]

[0025] The present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, where possible, identical components or parts are indicated by identical reference numerals throughout the drawings. In describing the present invention, detailed descriptions of related known functions or configurations are omitted to avoid obscuring the gist of the invention.

[0026] The terms "about," "substantially," etc., as used herein, are used in the sense of a numerical value or a value close to a numerical value when manufacturing and material tolerances inherent to the meanings mentioned are presented, and are used to prevent unscrupulous infringers from unfairly exploiting the disclosure in which exact or absolute values ​​are mentioned to aid understanding of the present invention.

[0027]

[0028] FIG. 1 is a block diagram functionally showing an image processing system based on artificial intelligence for cardiac ultrasound PACS according to the present invention, FIG. 2 is a block diagram showing an operation process of an image processing system based on artificial intelligence for cardiac ultrasound PACS according to the present invention, FIG. 3 is a diagram showing an exemplary screen for segmenting and extracting an image from a cardiac ultrasound image in an image extraction unit of an image processing system based on artificial intelligence for cardiac ultrasound PACS according to the present invention, and FIG. 4 is a diagram showing an exemplary screen for analyzing an image in a measurement value extraction unit of an image processing system based on artificial intelligence for cardiac ultrasound PACS according to the present invention.

[0029] The present invention is an invention based on a medical image information system (Picture Archiving and Communications System, PACS), and is a system that acquires and stores image information acquired through an image diagnosis device in a digital state, and comprehensively processes the functions required to transmit and search the interpretation and medical records together.

[0030] That is, digital medical images captured using cardiac ultrasound imaging equipment can be transmitted and stored in a large-capacity storage device (PACS server) and searched and viewed at multiple stations. Through this digitalization of medical images, all medical images within a medical institution can be integrated and managed, thereby improving the accuracy and efficiency of diagnosis.

[0031] Additionally, the invention is based on PACS and includes medical equipment, servers, storage, PCs, monitors, and networks, and the software that operates them as one is combined and used.

[0032]

[0033] The present invention can be functionally divided into an image acquisition unit that acquires images from a cardiac ultrasound diagnostic device and converts the images into DICOM standards, an image processing unit that includes a server, storage, and a PC that stores and analyzes images, an image output unit that includes a monitor, printer, and the like that searches for and outputs images, and a network unit that connects all equipment of the PACS wirelessly or by wire so that they are linked.

[0034]

[0035] As described above, the cardiac ultrasound PACS-based image processing system utilizing artificial intelligence according to the present invention based on PACS may include a receiving unit, an image extraction unit, a data processing unit, a measurement value extraction unit, a report generation unit, a feedback unit, and a data processing unit, as shown in FIG. 2, according to the operating process.

[0036] The above-mentioned receiver receives the ultrasound image and is directly connected to the ultrasound equipment. The image output in real time is input and cropped, and preprocessing is performed to adjust the resolution, contrast, brightness, and balance image properties.

[0037] The ultrasound images received from the above receiver include parasternal extension axis cross-section image (PLAX), parasternal shortening axis cross-section image (PSAX), apical four-chamber cross-section image (A4C), apical two-chamber cross-section image (A2C), and apical three-chamber cross-section image (A3C), and the images must be received in units of cardiac cycle.

[0038] The above cross-sectional views can be obtained depending on the position of the body where the cardiac ultrasound equipment probe is in contact, and a parasternal long axis cross-sectional image (PLAX) can be obtained by positioning the probe in a direction parallel to the long axis of the heart on the front of the subject's chest, a parasternal short axis cross-sectional image (PSAX) can be obtained by positioning the probe in a direction perpendicular to the long axis of the heart on the front of the subject's chest, and an apical four-chamber image (A4C) can be obtained by positioning the probe in the direction of the apex, which is the end of the heart, on the lower left side of the subject's chest, so that all four chambers of the heart (left atrium, left ventricle, right atrium, right ventricle) can be viewed, an apical two-chamber image (A2C) can be obtained by rotating so that two chambers of the heart (left atrium, left ventricle) can be viewed, and an apical three-chamber image (A3C) can be obtained by rotating so that three chambers of the heart (left atrium, left ventricle, aorta) can be viewed.

[0039] The above image extraction unit extracts a standard cross-sectional image of a cardiac ultrasound image through an artificial intelligence cardiac ultrasound image examination model learned based on deep learning in the image processing unit.

[0040] The image extraction unit, as shown in FIG. 3, divides an image into multiple images in one cardiac cycle, and then obtains anatomical location and structure information of the heart in the image through a Semantic Segmentation (precise object detection) model based on a Convolution Layer or Vision Transformer, compares the location information of each part of the heart, such as RV (Right Ventricle), LV (Left Ventricle), LVOT (Left Ventricle Outflow Tract), AO (Aorta), AV (Atrioventricular), MV (Mitral Valve), LA (Left Atrium), and LVPW (Left Ventricular Posterior Wall), in the standard image and the input image, calculates a score based on the similarity, derives the average of the scores for each cardiac cycle, and extracts the image with the highest average score.

[0041] The above measurement value extraction unit calculates the inspection measurement value (Measurement) through artificial intelligence analysis of the extracted image.

[0042] The above measurement value extraction unit uses a standard model pre-learned through a large amount of accumulated raw data to compare the image extracted from the image extraction unit with the standard model to produce an inspection measurement value (Measurement) including a prediction probability for a lesion.

[0043] That is, the image extracted from the image extraction unit is segmented and the characteristics of the image are judged based on the set criteria, and the type and existence probability of the lesion are measured by applying a learning model.

[0044] In addition, as shown in Fig. 4, by learning a mask (an image that distinguishes and displays a specific area in the image) that indicates the left atrial diameter (LAD) and the interventricular septum (IVS), left ventricular diameter (LVD), left ventricular free wall (LVFW), right ventricular free wall (RVFW), and right ventricular internal diameter (RVD) areas in M-mode echocardiography, the number of pixels in each area can be calculated to derive a test measurement value.

[0045] In addition, by examining all pixels of the image segmented in the image extraction unit, the texture characteristics for the identification area are analyzed, and by quantifying the characteristics of each identification area (Pixel) through the texture characteristic analysis, an inspection measurement value (Measurement) can be produced compared to normal reference data.

[0046] The above image extraction unit and measurement value extraction unit can build a deep learning-based learning model by utilizing the 'device for converting a captured image into deep learning learning data in a vision maker' of Korean Patent No. 2529688.

[0047]

[0048] The above report generation unit generates an artificial intelligence result report that integrates the image extracted from the image extraction unit and the inspection measurement value (Measurement) calculated from the measurement value extraction unit.

[0049] The above report generation unit is produced to assist the medical staff in their final interpretation, and the artificial intelligence result report may include the original cardiac ultrasound image to assist the medical staff in their final interpretation, and the image extraction unit and measurement value extraction unit may include various images and result values ​​utilized.

[0050] Additionally, the report generation unit may include images processed with DICOM properties and artificial intelligence interpretation results for lesions analyzed with artificial intelligence in the measurement value extraction unit.

[0051]

[0052] The above feedback section inputs the medical staff's final interpretation results for the above artificial intelligence result report and uses them as learning data for the learning model to improve the accuracy of the artificial intelligence.

[0053] A data processing unit is included that converts the result value input from the above feedback unit into deep learning learning data and transmits it to the image extraction unit and the measurement value extraction unit.

[0054] The above data processing unit collects the result values ​​input by medical staff in the feedback unit and transmits them to the image extraction unit and the measurement value extraction unit. It would be desirable for the feedback unit to be divided into various sections so that various feedbacks from medical staff can be input.

[0055]

[0056] As described above, the artificial intelligence-based cardiac ultrasound PACS image processing system according to the present invention can shorten the examination time by automatically extracting images necessary for cardiac ultrasound analysis without depending on the skill of the examiner through artificial intelligence, and can eliminate the visual error rate of the examiner analyzing the visualized cardiac image by analyzing the cardiac ultrasound image extracted through the artificial function again with artificial intelligence, thereby providing it to medical staff to assist the medical staff in their analysis.

[0057]

[0058] The present invention described above is not limited to the above-described embodiments and the attached drawings, and it will be apparent to a person skilled in the art to which the present invention pertains that various substitutions, modifications, and changes are possible within a scope that does not depart from the technical spirit of the present invention.

Claims

1. A receiving unit that receives ultrasound images; An image extraction unit that extracts a standard cross-sectional image of a cardiac ultrasound image using an artificial intelligence cardiac ultrasound image examination model learned based on deep learning in the image processing unit; A measurement value extraction unit that calculates inspection measurement values ​​(Measurement) through artificial intelligence analysis of the extracted image; A report generation unit that generates an artificial intelligence result report that integrates the extracted images and inspection measurement values; A feedback section that inputs the final interpretation results of medical staff for the above artificial intelligence result report and provides feedback. An artificial intelligence-based cardiac ultrasound PACS-based image processing system, characterized by including a data processing unit that converts the result value input from the feedback unit into deep learning learning data and transmits it to the image extraction unit and the measurement value extraction unit.

2. In paragraph 1, An image processing system based on artificial intelligence for cardiac ultrasound PACS, characterized in that the ultrasound images received from the above-mentioned receiving unit include a parasternal extension axis cross-section image (PLAX), a parasternal shortening axis cross-section image (PSAX), an apical four-chamber cross-section image (A4C), an apical two-chamber cross-section image (A2C), and an apical three-chamber cross-section image (A3C).

3. In paragraph 1, The above image extraction unit is an artificial intelligence-based cardiac ultrasound PACS-based image processing system characterized in that it extracts images after segmenting the images in one cardiac cycle.

4. In paragraph 1, An artificial intelligence-based cardiac ultrasound PACS-based image processing system characterized in that the above report generation unit includes images processed with DICOM properties and artificial intelligence interpretation results.

Images