A model optimization method and device for primary ciliary dyskinesia

Abstract

The present invention relates to a model optimization method and device for primary ciliary dyskinesia, comprising: obtaining a chest CT examination image of a target patient; removing artifacts in the chest CT examination image to obtain an artifact-removed chest CT examination image; calibrating the position of abnormal tissue in the artifact-removed chest CT examination image to form a training sample; inputting the training sample into a neural network for training to obtain a recognition model; and using the recognition model to determine the position of abnormal tissue in the chest CT examination image of the target patient. The present invention can improve image quality by removing artifacts in chest CT images, ensure that the features of abnormal tissues are clearer, and thus improve the accuracy of the recognition model.

Description

Technical Field

[0001] The present invention relates to the technical field of model optimization, and in particular to a model optimization and method for primary ciliary dyskinesia. Background Art

[0002] Primary Ciliary Dyskinesia (PCD) is a genetic disease characterized by abnormal cilia function, which leads to impaired function of multiple organs such as the respiratory tract, ear, nose and throat, and reproductive system. Cilia are tiny hair-like structures that can effectively propel fluids and particles to maintain self-cleaning and normal physiological functions in the body. PCD patients often show symptoms such as recurrent respiratory infections, sinusitis, ear infections, and fertility problems. The pathogenesis of this disease is complex and usually involves mutations in multiple genes, resulting in impaired structure and movement of cilia.

[0003] In order to diagnose PCD early, clinical diagnosis is often made by analyzing abnormal tissues on the patient's chest CT scan. However, traditional detection methods often rely on manual analysis, which is not only time-consuming and labor-intensive, but also subjective and easily leads to insufficient diagnostic accuracy. Summary of the invention

[0004] To solve the above problems, an object of the embodiments of the present invention is to provide a model optimization method and device for primary ciliary dyskinesia.

[0005] A recognition model optimization method for primary ciliary dyskinesia, comprising:

[0006] Step 1: Obtain chest CT images of target patients;

[0007] Step 2: removing artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image;

[0008] Step 3: calibrating the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample;

[0009] Step 4: Input the training sample into the neural network for training to obtain a recognition model;

[0010] Step 5: Use the recognition model to determine the location of abnormal tissue in the chest CT examination image of the target patient.

[0011] Preferably, the step 2: removing artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image comprises:

[0012] Step 2.1: Divide the chest CT examination image into a plurality of pixel blocks;

[0013] Step 2.2: Calculate the pixel mean and pixel variance within each pixel block;

[0014] Step 2.3: Extract the background area in the chest CT examination image;

[0015] Step 2.4: Determine an adaptation threshold based on the background area;

[0016] Step 2.5: Determine an artifact removal function based on the pixel mean, pixel variance, and adaptation threshold;

[0017] Step 2.6: Using the artifact removal function to remove artifacts from the chest CT examination image, obtain an artifact-removed chest CT examination image.

[0018] Preferably, in step 2.2, calculating the pixel mean and pixel variance in each pixel block includes:

[0019] Using the formula:

[0020]

[0021] Calculate the pixel mean and pixel variance within each pixel block; where D s represents pixel variance, E s represents the pixel mean, S(x,y) represents the pixel value of the pixel block at the (x,y) position, and the range of x is [x1, x2-1], the range of y is [y1, y2-1], and M represents the length of the pixel block N is the width of the pixel block

[0022] Preferably, the step 2.3: extracting the background area in the chest CT examination image includes:

[0023] Step 2.3.1: Calculate the gradient values ​​in the horizontal and vertical directions of the chest CT examination image;

[0024] Step 2.3.2: Determine the spatial gradient change value according to the gradient values ​​in the horizontal and vertical directions;

[0025] Step 2.3.3: The area with a value smaller than the spatial gradient change is regarded as the background area.

[0026] Preferably, in the step 2.3.2, it includes:

[0027] Using the formula:

[0028] G Y (x,y)=S(x,y)*H Y (x,y)

[0029] GX (x,y)=S(x,y)*H X (x,y)

[0030]

[0031] G(x,y)=|G X (x,y)+|G Y (x,y)|

[0032] Determine the spatial gradient change value; where H X (x,y) represents the horizontal response matrix, H Y (x,y) represents the response matrix in the vertical direction, G X (x, y) represents the gradient value in the horizontal direction, G Y (x, y) represents the gradient value in the vertical direction, and G(x, y) represents the spatial gradient change value.

[0033] Preferably, the step 2.4: determining the adaptation threshold based on the background area includes:

[0034] The adaptation threshold is determined based on the mean and variance of the background area; wherein the adaptation threshold calculation formula is:

[0035]

[0036] Where k represents the adaptation threshold, D s ′ represents the variance of the background area, E s ′ represents the mean value of the background area.

[0037] Preferably, in step 2.5, the artifact removal function is:

[0038]

[0039] Among them, S * (x, y) represents the pixel value of the chest CT examination image after artifacts are removed.

[0040] Preferably, the step 4: inputting the training samples into a neural network for training to obtain a recognition model comprises:

[0041] Step 4.1: normalize the training samples to obtain normalized training samples;

[0042] Step 4.2: Divide the training samples into training set and validation set;

[0043] Step 4.3: Use the training set to optimize and train the U-Net neural network to obtain the initial recognition model;

[0044] Step 4.4: Use the validation set to evaluate the initial recognition model. When the evaluation value is within the set range, output the recognition model.

[0045] The present invention also provides a model optimization device for primary ciliary dyskinesia, comprising:

[0046] An image acquisition module, used to acquire chest CT examination images of target patients;

[0047] An artifact denoising module is used to remove artifacts from a chest CT examination image to obtain an artifact-removed chest CT examination image;

[0048] A position calibration module, used to calibrate the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample;

[0049] A training module, used for inputting the training samples into a neural network for training to obtain a recognition model;

[0050] The location determination module is used to determine the location of abnormal tissue in the chest CT examination image of the target patient using the recognition model.

[0051] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps in the above-mentioned method for optimizing an identification model for primary ciliary dyskinesia.

[0052] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0053] The present invention relates to a model optimization method for primary ciliary dyskinesia. Compared with the prior art, the present invention can improve image quality by removing artifacts in chest CT images, ensure that the features of abnormal tissues are clearer, and thus improve the accuracy of the recognition model.

[0054] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] 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.

[0056] Figure 1 A flow chart of a recognition model optimization method for primary ciliary dyskinesia provided by the present invention;

[0057] Figure 2 A comparison diagram of artifact removal effects is provided for the present invention; wherein, a1, b1, c1 and d1 are magnified details of the original chest CT examination image, a2, b2, c2 and d2 are effect diagrams of a traditional artifact removal algorithm, and a3, b3, c3 and d3 are effect diagrams of the artifact removal algorithm provided by the present invention. DETAILED DESCRIPTION

[0058] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0059] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0060] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0061] See also Figure 1 , a model optimization method for primary ciliary dyskinesia, comprising:

[0062] Step 1: Obtain chest CT images of target patients;

[0063] Step 2: removing artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image;

[0064] Artifacts may be confused with lesions, leading to misdiagnosis or missed diagnosis. Removing artifacts allows doctors to see the true anatomical structure and abnormal manifestations more clearly, thereby improving the accuracy of subsequent diagnosis.

[0065] Furthermore, the step 2 comprises:

[0066] Step 2.1: Divide the chest CT examination image into a plurality of pixel blocks;

[0067] Step 2.2: Calculate the pixel mean and pixel variance within each pixel block;

[0068] In step 2.2, the formula is used:

[0069]

[0070] Calculate the pixel mean and pixel variance within each pixel block; where D s represents pixel variance, E s represents the pixel mean, S(x,y) represents the pixel value of the pixel block at the (x,y) position, and the range of x is [x1, x2-1], the range of y is [y1, y2-1], and M represents the length of the pixel block N is the width of the pixel block

[0071] Step 2.3: Extract the background area in the chest CT examination image;

[0072] Among them, step 2.3 includes:

[0073] Step 2.3.1: Calculate the gradient values ​​in the horizontal and vertical directions of the chest CT examination image;

[0074] Step 2.3.2: Determine the spatial gradient change value according to the gradient values ​​in the horizontal and vertical directions;

[0075] In step 2.3.2, the formula is used:

[0076] G Y (x,y)=S(x,y)*H Y (x,y)

[0077] G X (x,y)=S(x,y)*H X (x,y)

[0078]

[0079] G(x,y)=|G X (x,y)+|G Y (x,y)|

[0080] Determine the spatial gradient change value; where H X (x,y) represents the horizontal response matrix, H Y (x,y) represents the response matrix in the vertical direction, G X (x, y) represents the gradient value in the horizontal direction, G Y (x, y) represents the gradient value in the vertical direction, and G(x, y) represents the spatial gradient change value.

[0081] Step 2.3.3: The area with a value smaller than the spatial gradient change is regarded as the background area.

[0082] Step 2.4: Determine an adaptation threshold based on the background area;

[0083] In step 2.4, the adaptation threshold is determined based on the mean and variance of the background area; wherein the adaptation threshold calculation formula is:

[0084]

[0085] Where k represents the adaptation threshold, D s ′ represents the variance of the background area, E s ′ represents the mean value of the background area.

[0086] Step 2.5: Determine an artifact removal function based on the pixel mean, pixel variance, and adaptation threshold;

[0087] In step 2.5, the artifact removal function is:

[0088]

[0089] Among them, S * (x, y) represents the pixel value of the chest CT examination image after artifacts are removed.

[0090] Step 2.6: Using the artifact removal function to remove artifacts from the chest CT examination image, obtain an artifact-removed chest CT examination image.

[0091] The present invention utilizes statistical features (such as mean and variance) to better identify the difference between artifacts and normal tissues, thereby improving the accuracy of removing artifacts. Figure 2 As shown, the artifact removal algorithm in this application is significantly better than the conventional denoising method.

[0092] Step 3: calibrating the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample;

[0093] Step 4: Input the training sample into the neural network for training to obtain a recognition model;

[0094] Furthermore, step 4 includes:

[0095] Step 4.1: normalize the training samples to obtain normalized training samples;

[0096] Step 4.2: Divide the training samples into training set and validation set;

[0097] Step 4.3: Use the training set to optimize and train the U-Net neural network to obtain the initial recognition model;

[0098] Step 4.4: Use the validation set to evaluate the initial recognition model. When the evaluation value is within the set range, output the recognition model.

[0099] Step 5: Use the recognition model to determine the location of abnormal tissue in the chest CT examination image of the target patient.

[0100] The present invention can improve image quality by removing artifacts from chest CT images, ensure that the features of abnormal tissue are clearer, and then input the training samples into the neural network, which can fully utilize the powerful ability of deep learning algorithms in image recognition, automatically extract highly complex features, and improve the ability to detect abnormal tissue. It can provide a reliable basis for the early diagnosis of primary ciliary dyskinesia, help doctors make clinical decisions more accurately, and improve the treatment effect of patients.

[0101] The present invention also provides a model optimization device for primary ciliary dyskinesia, comprising:

[0102] An image acquisition module, used to acquire chest CT examination images of target patients;

[0103] An artifact denoising module is used to remove artifacts from a chest CT examination image to obtain an artifact-removed chest CT examination image;

[0104] A position calibration module, used to calibrate the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample;

[0105] A training module, used for inputting the training samples into a neural network for training to obtain a recognition model;

[0106] The location determination module is used to determine the location of abnormal tissue in the chest CT examination image of the target patient using the recognition model.

[0107] Compared with the prior art, the beneficial effects of the model optimization device for primary ciliary dyskinesia provided by the present invention are the same as the beneficial effects of the model optimization method for primary ciliary dyskinesia described in the above technical solution, which will not be elaborated here.

[0108] The present invention also provides an electronic device, comprising a bus, a transceiver, a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory and the processor are connected via the bus, and wherein the computer program, when executed by the processor, implements the steps in the above-mentioned method for optimizing an identification model for primary ciliary dyskinesia. Compared with the prior art, the beneficial effects of the electronic device provided by the present invention are the same as the beneficial effects of the method for optimizing a model for primary ciliary dyskinesia described in the above-mentioned technical solution, and are not elaborated herein.

[0109] The present invention also provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps in the above-mentioned model optimization for primary ciliary dyskinesia are implemented. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by the present invention are the same as the beneficial effects of the method for model optimization for primary ciliary dyskinesia described in the above-mentioned technical solution, which will not be repeated here.

[0110] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technical solution that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A recognition model optimization method for primary ciliary dyskinesia, characterized in that: include: Step 1: Obtain chest CT images of target patients; Step 2: removing artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image; Step 3: calibrating the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample; Step 4: Input the training sample into the neural network for training to obtain a recognition model; Step 5: using the recognition model to determine the location of abnormal tissue in the chest CT examination image of the target patient; The step 2: removing artifacts from the chest CT examination image to obtain the chest CT examination image after the artifacts are removed, comprises: Step 2.1: Divide the chest CT examination image into a plurality of pixel blocks; Step 2.2: Calculate the pixel mean and pixel variance within each pixel block; Step 2.3: Extract the background area in the chest CT examination image; Step 2.4: Determine an adaptation threshold based on the background area; Step 2.5: Determine an artifact removal function based on the pixel mean, pixel variance, and adaptation threshold; Step 2.6: using the artifact removal function to remove artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image; In step 2.2, calculating the pixel mean and pixel variance in each pixel block includes: Using the formula: Calculate the pixel mean and pixel variance within each pixel block; where D s represents pixel variance, E s represents the pixel mean, S(x,y) represents the pixel value of the pixel block at the (x,y) position, and the range of x is [x1, x 2-1 ], the range of y is [y1,y 2-1 ], M represents the length of the pixel block N is the width of the pixel block 2. The identification model optimization method for primary ciliary dyskinesia according to claim 1, characterized in that: The step 2.3: extracting the background area in the chest CT examination image includes: Step 2.3.1: Calculate the gradient values ​​in the horizontal and vertical directions of the chest CT examination image; Step 2.3.2: Determine the spatial gradient change value according to the gradient values ​​in the horizontal and vertical directions; Step 2.3.3: The area with a value smaller than the spatial gradient change is regarded as the background area.

3. The identification model optimization method for primary ciliary dyskinesia according to claim 2, characterized in that: In the step 2.3.2, it includes: Using the formula: G Y (x,y)=S(x,y)*H Y (x,y) G X (x,y)=S(x,y)*H X (x,y) G(x,y)=|G X (x,y)|+|G Y (x,y)| Determine the spatial gradient change value; where H X (x,y) represents the horizontal response matrix, H Y (x,y) represents the response matrix in the vertical direction, G X (x, y) represents the gradient value in the horizontal direction, G Y (x, y) represents the gradient value in the vertical direction, and G(x, y) represents the spatial gradient change value.

4. The identification model optimization method for primary ciliary dyskinesia according to claim 3, characterized in that: The step 2.4: determining the adaptation threshold based on the background area comprises: The adaptation threshold is determined based on the mean and variance of the background area; wherein the adaptation threshold calculation formula is: Where k represents the adaptation threshold, D s ′ represents the variance of the background area, E s ′ represents the mean value of the background area.

5. The identification model optimization method for primary ciliary dyskinesia according to claim 4, characterized in that: In step 2.5, the artifact removal function is: Among them, S * (x, y) represents the pixel value of the chest CT examination image after artifacts are removed.

6. The identification model optimization method for primary ciliary dyskinesia according to claim 5, characterized in that: The step 4: inputting the training sample into a neural network for training to obtain a recognition model includes: Step 4.1: normalize the training samples to obtain normalized training samples; Step 4.2: Divide the training samples into training set and validation set; Step 4.3: Use the training set to optimize and train the U-Net neural network to obtain the initial recognition model; Step 4.4: Use the validation set to evaluate the initial recognition model. When the evaluation value is within the set range, output the recognition model.

7. A model optimization device for primary ciliary dyskinesia, characterized in that: include: An image acquisition module, used to acquire chest CT examination images of target patients; An artifact denoising module is used to remove artifacts from a chest CT examination image to obtain an artifact-removed chest CT examination image; A position calibration module, used to calibrate the position of abnormal tissue in the chest CT examination image after artifact removal to form a training sample; A training module, used for inputting the training samples into a neural network for training to obtain a recognition model; A location determination module, used to determine the location of abnormal tissue in the chest CT examination image of the target patient using the recognition model; Removing artifacts from a chest CT examination image to obtain an artifact-removed chest CT examination image includes: Dividing the chest CT examination image into a plurality of pixel blocks; Calculate the pixel mean and pixel variance within each pixel block; Extract the background area in the chest CT examination image; determining an adaptation threshold based on the background area; Determine an artifact removal function based on the pixel mean, pixel variance, and adaptive threshold; Using the artifact removal function to remove artifacts from the chest CT examination image to obtain an artifact-removed chest CT examination image; Calculate the pixel mean and pixel variance within each pixel block, including: Using the formula: Calculate the pixel mean and pixel variance within each pixel block; where D s represents pixel variance, E s represents the pixel mean, S(x,y) represents the pixel value of the pixel block at the (x,y) position, and the range of x is [x1, x 2-1 ], the range of y is [y1,y 2-1 ], M represents the length of the pixel block N is the width of the pixel block 8. An electronic device, comprising a bus, a transceiver, a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the transceiver, the memory, and the processor are connected via the bus, wherein: When the computer program is executed by the processor, the steps of the identification model optimization method for primary ciliary dyskinesia according to any one of claims 1 to 6 are implemented.

Images