Emergency department intelligent triage system based on machine learning

Abstract

The invention relates to the technical field of image communication, in particular to an emergency department intelligent triage system based on machine learning, and the system comprises the steps: collecting a remote consultation image; obtaining the gray variance and the gray variance continuity of each pixel point, and further obtaining the gray value loss degree of each pixel point; obtaining the standard deviation of the sub Gaussian distribution model corresponding to each pixel point according to the gray histogram of the remote consultation image; obtaining a gray value change threshold value of each pixel point according to the gray value loss degree of each pixel point and the standard deviation of the corresponding sub Gaussian distribution model; obtaining a difference value of each pixel point; and changing a threshold according to the difference value and the gray value of each pixel point, changing the gray value of each pixel point, and further compressing and transmitting the remote consultation image. According to the method, the gray value information of part of pixel points in the image is abandoned, and then the Huffman coding algorithm is used for compression, so that the reservation of detail information in the image is ensured, and meanwhile, the compression efficiency is also improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image communication, and in particular to an intelligent triage system for emergency departments based on machine learning. Background Art

[0002] Telemedicine refers to the transmission of medical images from one location to another through remote technology. Doctors can conduct remote consultations by transmitting patients' medical images to remote experts to achieve intelligent triage in the emergency department. This method allows patients to receive diagnosis and treatment advice from experts around the world. When transmitting medical images, the images need to be compressed to reduce the size of the image files, thereby reducing the amount of image data that needs to be transmitted and increasing the transmission speed.

[0003] Huffman coding is a commonly used lossless compression method that operates on the grayscale value of each pixel in the remote consultation image. However, since the grayscale value distribution of pixels in the remote consultation image is relatively scattered, the Huffman tree obtained according to the frequency of the pixel grayscale value may be deep, resulting in too long code length when encoding the remote consultation image, low image compression efficiency, and reduced transmission speed during image transmission. Summary of the invention

[0004] The present invention provides an intelligent triage system for emergency departments based on machine learning to solve the existing problems.

[0005] The intelligent triage system for emergency department based on machine learning of the present invention adopts the following technical solutions: An embodiment of the present invention provides an intelligent triage system for emergency department based on machine learning, which implements the following steps: Collect remote consultation images; Obtain the grayscale variance of each pixel; obtain the grayscale variance continuity of each pixel according to the grayscale variance of each pixel; obtain the grayscale value loss degree of each pixel according to the grayscale variance of each pixel and the grayscale variance continuity; Obtain a grayscale histogram of the remote consultation image; obtain a mixed Gaussian model based on the grayscale histogram of the remote consultation image; obtain the standard deviation of the sub-Gaussian distribution model corresponding to each pixel point based on the mixed Gaussian model; obtain the grayscale value change threshold of each pixel point based on the grayscale value loss degree of each pixel point and the standard deviation of the corresponding sub-Gaussian distribution model; Obtain the difference value of each pixel; change the threshold value according to the difference value of each pixel and the gray value, and change the gray value of each pixel; According to the changed grayscale value of each pixel, the remote consultation image is compressed and transmitted.

[0006] Preferably, the step of obtaining the grayscale variance of each pixel point includes the following specific steps: The default sliding window size is , traverse any pixel point in the remote consultation image, and the grayscale variance of the pixel point in the corresponding sliding window is taken as the grayscale variance of the pixel point.

[0007] Preferably, the grayscale variance continuity of each pixel point is obtained according to the grayscale variance of each pixel point, and the specific steps include the following: Traverse any pixel point in the remote consultation image, obtain the grayscale variance of other pixel points in the eight neighborhoods of the pixel point, and sort the grayscale variances in descending order to obtain a sequence, obtain the pixel points corresponding to the first two grayscale variances from the sequence for analysis, record them as the first pixel point and the second pixel point, establish a rectangular coordinate system with the pixel point as the center point, connect the first pixel point and the second pixel point with the origin of the rectangular coordinate system to form two straight lines, obtain the angle between each straight line and the vertical axis of the rectangular coordinate system, record them as the first angle and the second angle, obtain the difference between the first angle and the second angle, record them as the first difference, and obtain the first difference and The difference between them is taken as the grayscale variance continuity of the pixel.

[0008] Preferably, the grayscale value loss degree of each pixel point is obtained according to the grayscale variance and grayscale variance continuity of each pixel point, and the specific steps include the following: In the formula, For the Grayscale variance of pixels; For the The first angle of pixels; For the The second angle of pixels; Representative Continuity of grayscale variance of pixels; For the The degree of gray value loss of each pixel; Represents an exponential function with a natural constant as base.

[0009] Preferably, the step of obtaining the grayscale histogram of the remote consultation image includes the following specific steps: The grayscale value of the pixel in the remote consultation image is used as the horizontal coordinate, and the number of pixels corresponding to the grayscale value is used as the vertical coordinate to draw a grayscale histogram of the remote consultation image.

[0010] Preferably, the method of obtaining a mixed Gaussian model according to the grayscale histogram of the remote consultation image; and obtaining the standard deviation of the sub-Gaussian distribution model corresponding to each pixel point according to the mixed Gaussian model includes the following specific steps: Use the grayscale histogram of the remote consultation image The algorithm fits the mixed Gaussian model, obtains the mixed Gaussian model, and obtains the standard deviation of each sub-Gaussian model. The standard deviation of the sub-Gaussian model to which each pixel belongs is the standard deviation of the sub-Gaussian distribution model corresponding to each pixel.

[0011] Preferably, the gray value change threshold of each pixel point is obtained according to the gray value loss degree of each pixel point and the standard deviation of the corresponding sub-Gaussian distribution model, and the specific steps include the following: In the formula, For the The degree of gray value loss of each pixel; For the The standard deviation of the sub-Gaussian distribution model corresponding to each pixel; For the The gray value of each pixel changes the threshold.

[0012] Preferably, the step of obtaining the difference value of each pixel point includes the following specific steps: Get the difference between the gray value of each pixel and the gray mean of other pixels in the eight neighborhoods, and record it as the difference value.

[0013] Preferably, the gray value of each pixel is changed by changing the threshold value according to the difference value of each pixel and the gray value, and the specific steps include the following: Traverse any pixel in the remote consultation image. When the difference value of the pixel is greater than the grayscale value change threshold, the grayscale value of the pixel does not change. When the difference value of the pixel is less than the grayscale value change threshold and the grayscale value of the pixel is less than the grayscale mean of other pixels in the eight neighborhoods, the grayscale value of the pixel is increased according to the difference value. When the difference value of the pixel is greater than the grayscale value change threshold and the grayscale value of the pixel is greater than the grayscale mean of other pixels in the eight neighborhoods, the grayscale value of the pixel is reduced according to the difference value.

[0014] Preferably, the method of compressing and transmitting the remote consultation image according to the changed grayscale value of each pixel point includes the following specific steps: The frequencies of the changed gray values ​​are counted to construct a Huffman tree, the remote consultation images are compressed, and the compressed remote consultation images are transmitted remotely.

[0015] The beneficial effects of the technical solution of the present invention are as follows: the present invention obtains the grayscale variance and grayscale variance continuity of each pixel according to the characteristics of the remote consultation image, and then obtains the grayscale value loss degree of each pixel according to the grayscale variance and grayscale variance continuity of each pixel, obtains the mixed Gaussian model according to the grayscale histogram of the remote consultation image, and then obtains the standard deviation of the sub-Gaussian distribution model corresponding to each pixel; obtains the grayscale value change threshold of each pixel according to the grayscale value loss degree of each pixel and the standard deviation of the corresponding sub-Gaussian distribution model, and obtains the difference value of each pixel; changes the grayscale value of each pixel according to the difference value of each pixel and the grayscale value change threshold, and compresses and transmits the remote consultation image according to the changed grayscale value of each pixel. The present invention abandons the grayscale value information of the tissue area in the image to make the range of the grayscale value more concentrated, and the Huffman tree obtained according to the changed grayscale value is shallower, while ensuring the retention of important information in the remote consultation image, and improving the compression efficiency and transmission speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 The present invention is a flowchart of the steps of implementing an intelligent triage system for emergency departments based on machine learning. DETAILED DESCRIPTION

[0018] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, features and effects of an intelligent triage system for emergency department based on machine learning proposed by the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0019] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0020] The following is a detailed description of a specific scheme of an intelligent triage system for emergency department based on machine learning provided by the present invention in conjunction with the accompanying drawings.

[0021] See also Figure 1, which shows a flowchart of the steps of implementing an intelligent triage system for emergency department based on machine learning provided by one embodiment of the present invention, the system implements the following steps: S001. Collect remote consultation images.

[0022] It should be noted that when a patient in the hospital needs a remote consultation, the hospital will collect various data of the patient to be consulted, such as routine clinical examinations, X-rays, CT scans, B-ultrasound films, etc. For patients who need a remote consultation, the medical image data of the patient to be consulted is collected and input into the computer system. The medical images of the patient to be consulted collected in the computer system are grayed, and the processed images are recorded as remote consultation images.

[0023] S002. Obtain the gray value loss degree of each pixel in the remote consultation image.

[0024] It should be noted that Huffman coding is a lossless coding algorithm for data compression. It uses shorter codes for data with higher frequency and longer codes for data with lower frequency. This can reduce the overall length of the code, thereby achieving data compression. However, since the grayscale values ​​of pixels in the acquired neurosurgery grayscale image are numerous and scattered, when Huffman coding is used, the depth of the constructed Huffman tree is deeper due to the presence of more low-frequency grayscale values. Therefore, the code length obtained after encoding each pixel of the remote consultation image is longer, and the image compression effect is not good, resulting in a lower speed of the image transmission process. It is known that in remote consultation images, the grayscale values ​​of pixels are not as high as those of pixels. Different tissues or structures usually show different grayscale values. This feature enables doctors and radiologists to distinguish and identify different tissue types according to the grayscale level of the image. Therefore, the grayscale value information of the edge points between tissues needs to be retained, and the grayscale values ​​of the pixels inside each tissue can be changed to make their grayscale values ​​consistent, so as to make the distribution range of the grayscale values ​​in the image concentrated. Therefore, an embodiment of the present invention proposes a lossy coding algorithm, which concentrates the grayscale value range by discarding the grayscale value information of certain pixels in the remote consultation image, and then compresses the data in the remote consultation image through the Huffman coding algorithm to achieve a good compression effect.

[0025] It should be further explained that, since the grayscale value information of the pixels inside each tissue area in the remote consultation image can be discarded, and the grayscale value information of the pixels at the edge between tissues is kept unchanged as much as possible, it is necessary to obtain the degree of grayscale value loss of each pixel in the remote consultation image. For the pixels inside the tissue area, the degree of grayscale value loss is large, and for the edge points between tissues, the degree of grayscale value loss is small. Because there are large differences in grayscale values ​​between each tissue area, the pixels at the edge between tissue areas have a large grayscale variance in their local areas, and the grayscale variance continuity of the pixels at the edge is large. Therefore, the degree of grayscale value loss of each pixel in the remote consultation image is obtained based on the grayscale variance of each pixel in its local area and the continuity of the grayscale variance.

[0026] In the embodiment of the present invention, it is set The sliding window size of traverses all pixels in the remote consultation image to obtain the grayscale variance of each pixel in its sliding window. In this embodiment of the present invention, set In other embodiments, the implementer can set size.

[0027] It should be noted that for the pixels at the edges between tissue regions, their grayscale variances are continuous, so it is necessary to obtain the continuity of the grayscale variances of each pixel. Therefore, in an embodiment of the present invention, any pixel in the remote consultation image is traversed and recorded as the current pixel, the grayscale variances of other pixels in the eight neighborhoods of the current pixel are obtained, and the grayscale variances are sorted in order from large to small to obtain a sequence, and the pixels corresponding to the first two grayscale variances in the sequence are obtained for analysis, recorded as the first pixel and the second pixel, and a rectangular coordinate system is established with the current pixel as the center point, and the first pixel and the second pixel are connected to the origin of the rectangular coordinate system to form two straight lines, and the angle between each straight line and the vertical axis of the rectangular coordinate system is obtained, recorded as the first angle and the second angle. It should be noted that when the difference between the first angle and the second angle is close to 180, it means that the continuity of the grayscale variance of the current pixel is large.

[0028] Get the gray value loss degree of each pixel in the remote consultation image: In the formula, For the Grayscale variance of pixels; For the The first angle of pixels; For the The second angle of pixels; Representative The grayscale variance continuity of pixels is The closer the difference between the first angle and the second angle of a pixel is to 180, the The more pixels there are, the more likely they are edge points between tissue regions; For the The gray value loss degree of each pixel; The larger the grayscale variance of each pixel and the greater the continuity of the grayscale variance, the The more likely a pixel is an edge point between tissues, the smaller the gray value loss will be. When the grayscale variance of each pixel is smaller and the continuity of grayscale variance is smaller, The more likely a pixel is to be a pixel inside the tissue area, the greater the loss of its gray value.

[0029] So far, the gray value loss degree of each pixel in the remote consultation image is obtained.

[0030] S003. Obtain a gray value change threshold of each pixel point according to the gray value loss degree of each pixel point.

[0031] It should be noted that, since the grayscale values ​​of various tissue regions vary greatly, for a pixel point in each tissue region, the change range of its grayscale value cannot exceed the grayscale value characteristic of the tissue region to which it belongs.

[0032] In the embodiment of the present invention, a grayscale histogram of a remote consultation image is obtained, and the grayscale histogram of the remote consultation image is used The algorithm fits a mixed Gaussian model and obtains multiple sub-Gaussian distribution models and the standard deviation of each sub-Gaussian model. It should be noted that each tissue area corresponds to a sub-Gaussian model. Therefore, for the pixel point of any tissue area, the change range of its grayscale value shall not exceed the standard deviation of the sub-Gaussian model corresponding to the tissue area.

[0033] It should be noted that the grayscale value information of the pixels inside the tissue area can be lost, and the degree of loss of the grayscale value of the pixels is large. The grayscale value information of the edge points between the tissue areas cannot be lost, and the degree of loss of the grayscale value of the pixels is small. Therefore, the grayscale value change threshold of each pixel is obtained by combining the degree of loss of the grayscale value of each pixel and the standard deviation of the sub-Gaussian distribution model corresponding to each pixel.

[0034] In an embodiment of the present invention, the gray value change threshold of each pixel is obtained: In the formula, For the The degree of gray value loss of each pixel; For the The standard deviation of the sub-Gaussian distribution model corresponding to each pixel; For the The threshold for changing the gray value of each pixel point is larger for the pixels inside the tissue area, and smaller for the edge points between tissue areas.

[0035] At this point, the grayscale value change threshold of each pixel point is obtained according to the grayscale value loss degree of each pixel point.

[0036] S004. Change the gray value of each pixel in the remote consultation image by changing the threshold value according to the gray value of each pixel.

[0037] It should be noted that the grayscale value change threshold of each pixel is the maximum grayscale value that each pixel can increase or decrease. However, if the grayscale value change threshold is used directly to change the grayscale values ​​of all pixels in the remote consultation image, it may cause the loss of detail information in the remote consultation image. Therefore, the grayscale value of each pixel is changed according to the grayscale value change threshold of each pixel and the difference between the grayscale value of each pixel and the grayscale value of the surrounding pixels. When the grayscale value difference between the pixel and the surrounding pixels is greater than the grayscale value change threshold, it is considered that the loss of the grayscale value of the pixel is too large and may destroy the original image features. Therefore, the grayscale value of the pixel is not changed. When the grayscale value difference between the pixel and the surrounding pixels is less than the grayscale value change threshold, it is considered that the loss of the grayscale value of the pixel is within an acceptable range and the grayscale value of the pixel is changed according to the grayscale value difference.

[0038] In an embodiment of the present invention, the difference between the grayscale value of each pixel and the grayscale mean of the pixels in its eight neighborhoods is obtained, recorded as the difference value, and the difference value is compared with the grayscale value change threshold of the pixel. When the difference value is greater than the grayscale value change threshold, the grayscale value of the pixel does not change. When the difference value is less than the grayscale value change threshold and the grayscale value of the pixel is less than the grayscale mean of the pixels in the eight neighborhoods, the grayscale value of the pixel is increased according to the difference value. When the difference value is greater than the grayscale value change threshold and the grayscale value of the pixel is greater than the grayscale mean of the pixels in the eight neighborhoods, the grayscale value of the pixel is reduced according to the difference value.

[0039] So far, the grayscale value of each pixel in the remote consultation image has been changed.

[0040] S005. Compress and transmit remote consultation images.

[0041] The frequencies of the gray values ​​after the changes are counted to construct the Huffman tree, and the remote consultation images are compressed. The compressed remote consultation images are then transmitted remotely to facilitate remote experts to diagnose the patient's condition.

[0042] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent triage system for emergency department based on machine learning, characterized in that: The system implements the following steps: Collect remote consultation images; Obtain the grayscale variance of each pixel; obtain the grayscale variance continuity of each pixel according to the grayscale variance of each pixel; obtain the grayscale value loss degree of each pixel according to the grayscale variance of each pixel and the grayscale variance continuity; Obtain a grayscale histogram of the remote consultation image; obtain a mixed Gaussian model based on the grayscale histogram of the remote consultation image; obtain the standard deviation of the sub-Gaussian distribution model corresponding to each pixel point based on the mixed Gaussian model; obtain the grayscale value change threshold of each pixel point based on the grayscale value loss degree of each pixel point and the standard deviation of the corresponding sub-Gaussian distribution model; Obtain the difference value of each pixel; change the threshold value according to the difference value of each pixel and the gray value, and change the gray value of each pixel; According to the changed grayscale value of each pixel, the remote consultation image is compressed and transmitted.

2. According to claim 1, an intelligent triage system for emergency department based on machine learning is characterized in that: The specific steps of obtaining the grayscale variance of each pixel are as follows: The default sliding window size is , traverse any pixel point in the remote consultation image, and the grayscale variance of the pixel point in the corresponding sliding window is taken as the grayscale variance of the pixel point.

3. According to claim 1, an intelligent triage system for emergency department based on machine learning is characterized in that: The grayscale variance continuity of each pixel point is obtained according to the grayscale variance of each pixel point, and the specific steps include the following: Traverse any pixel point in the remote consultation image, obtain the grayscale variance of other pixel points in the eight neighborhoods of the pixel point, and sort the grayscale variances in descending order to obtain a sequence, obtain the pixel points corresponding to the first two grayscale variances from the sequence for analysis, record them as the first pixel point and the second pixel point, establish a rectangular coordinate system with the pixel point as the center point, connect the first pixel point and the second pixel point with the origin of the rectangular coordinate system to form two straight lines, obtain the angle between each straight line and the vertical axis of the rectangular coordinate system, record them as the first angle and the second angle, obtain the difference between the first angle and the second angle, record them as the first difference, and obtain the first difference and The difference between them is taken as the grayscale variance continuity of the pixel.

4. According to claim 1, an intelligent triage system for emergency department based on machine learning is characterized in that: The grayscale value loss degree of each pixel point is obtained according to the grayscale variance and grayscale variance continuity of each pixel point, and the specific steps include the following: In the formula, For the Grayscale variance of pixels; For the The first angle of pixels; For the The second angle of pixels; Representative Continuity of grayscale variance of pixels; For the The degree of gray value loss of each pixel; Represents an exponential function with a natural constant as base.

5. The intelligent triage system for emergency department based on machine learning according to claim 1, characterized in that: The specific steps of obtaining the grayscale histogram of the remote consultation image are as follows: The grayscale value of the pixel in the remote consultation image is used as the horizontal coordinate, and the number of pixels corresponding to the grayscale value is used as the vertical coordinate to draw a grayscale histogram of the remote consultation image.

6. The intelligent triage system for emergency department based on machine learning according to claim 1, characterized in that: The method of obtaining a mixed Gaussian model according to the grayscale histogram of the remote consultation image and obtaining the standard deviation of the sub-Gaussian distribution model corresponding to each pixel point according to the mixed Gaussian model includes the following specific steps: Use the grayscale histogram of the remote consultation image The algorithm fits the mixed Gaussian model, obtains the mixed Gaussian model, and obtains the standard deviation of each sub-Gaussian model. The standard deviation of the sub-Gaussian model to which each pixel belongs is the standard deviation of the sub-Gaussian distribution model corresponding to each pixel.

7. The intelligent emergency department triage system based on machine learning according to claim 1, characterized in that: The gray value change threshold of each pixel point is obtained according to the gray value loss degree of each pixel point and the standard deviation of the corresponding sub-Gaussian distribution model, and the specific steps include the following: In the formula, For the The degree of gray value loss of each pixel; For the The standard deviation of the sub-Gaussian distribution model corresponding to each pixel; For the The gray value of each pixel changes the threshold.

8. The intelligent triage system for emergency department based on machine learning according to claim 1, characterized in that: The specific steps of obtaining the difference value of each pixel point are as follows: Get the difference between the gray value of each pixel and the gray mean of other pixels in the eight neighborhoods, and record it as the difference value.

9. The intelligent triage system for emergency department based on machine learning according to claim 1, characterized in that: The specific steps of changing the threshold value according to the difference value of each pixel and the gray value and changing the gray value of each pixel are as follows: Traverse any pixel in the remote consultation image. When the difference value of the pixel is greater than the grayscale value change threshold, the grayscale value of the pixel does not change. When the difference value of the pixel is less than the grayscale value change threshold and the grayscale value of the pixel is less than the grayscale mean of other pixels in the eight neighborhoods, the grayscale value of the pixel is increased according to the difference value. When the difference value of the pixel is greater than the grayscale value change threshold and the grayscale value of the pixel is greater than the grayscale mean of other pixels in the eight neighborhoods, the grayscale value of the pixel is reduced according to the difference value.

10. The intelligent triage system for emergency department based on machine learning according to claim 1, characterized in that: The specific steps of compressing and transmitting the remote consultation image according to the changed grayscale value of each pixel are as follows: The frequencies of the changed gray values ​​are counted to construct a Huffman tree, the remote consultation images are compressed, and the compressed remote consultation images are transmitted remotely.

Images