Method for predicting gonitis injury based on CT (Computed Tomography) image

Abstract

The invention belongs to the field of medical image mode recognition, and provides a CT image-based gonitis injury prediction method, which comprises the following steps of: 1, acquiring a group of CT images of a healthy contrast and a group of subjects with different degrees of knee joint injury; 2, the X-ray image is preprocessed, a knee joint area is cut out, and an LAB color model is used for converting a gray level image into three channels of L, a and b; step 3, carrying out image enhancement on the image of the L channel by adopting adaptive histogram equalization; step 4, integrating the three channels L, a and b after enhancement; step 5, de-noising the enhanced image; step 6, marking the image of the healthy control as 0, and marking the image of the patient with gonitis as 1, 2, 3 and 4 according to severity; step 7, performing feature extraction by using a discriminant auto-encoder; the method can reduce manual errors, is an effective supplement for clinicians to judge the gonitis, and is expected to realize accurate pre-judgment of the gonitis in areas lacking professional medical personnel.

Description

Technical Field

[0001] The invention relates to the field of medical image pattern recognition, and in particular to a method for predicting knee arthritis damage based on CT images. Background Art

[0002] Knee arthritis, as a common chronic joint disease, affects the quality of life of millions of people around the world. According to epidemiological data, the incidence of knee arthritis is particularly significant in the elderly population, and it is estimated that about 5% to 10% of the elderly worldwide are troubled by it. In China, with the acceleration of the aging process of the population, the prevalence of knee arthritis is also increasing, which has brought a huge economic burden to the social medical system and families. Patients often face persistent pain, joint stiffness and dysfunction, which can limit their daily activities, such as walking, going up and down stairs, and even simple self-care. In addition, the quality of life of patients is significantly reduced, and they often need to rely on family members for daily care, which increases the economic burden and psychological pressure on the family.

[0003] Current methods of evaluating knee arthritis include clinical symptom assessment, imaging examinations, and laboratory tests. Clinical evaluation usually relies on the patient's complaints and physical examination results, such as knee pain, swelling, and limited mobility. In terms of imaging examinations, X-rays are the most commonly used tool. X-rays can clearly show degenerative changes in bones and joints, such as bone spurs and narrowing of the joint space, and can help evaluate cartilage damage and other soft tissue problems in the joint. At present, clinical diagnosis and treatment expect to track knee patients throughout the entire cycle, assess the severity of the injury, develop more effective treatment plans, and intervene in time to delay the progression of the disease. However, due to the complexity of knee damage and the diversity of symptoms, existing diagnostic methods, such as X-rays and magnetic resonance imaging, mainly focus on changes in bone structure. Although they can reveal structural damage to the joint, they may not be able to fully assess the actual damage to joint function.

[0004] Therefore, the present invention proposes a machine learning algorithm based on CT images, which can achieve accurate prediction of knee joint injury grade. This intelligent and highly accurate prediction method helps to predict the patient's accurate prognosis, formulate a personalized rehabilitation plan for them, and monitor the changes of the disease in the long term, thereby improving the treatment effect and the patient's quality of life. Summary of the invention

[0005] In view of the above problems in the prior art, the present application proposes a method for predicting knee arthritis damage based on CT images. To achieve the above purpose, the present invention adopts the following technical solutions:

[0006] Step 1: Obtain CT images of a group of healthy controls and a group of subjects with different degrees of knee joint damage;

[0007] Step 2: Preprocess the X-ray image, crop the knee joint area and convert the grayscale image into three channels of L, a and b using the LAB color model;

[0008] Step 3, the image of the L channel is enhanced by using adaptive histogram equalization; Step 4, the enhanced L, a, and b channels are integrated;

[0009] Step 5: De-noise the enhanced image. Here, the fast non-local mean denoising algorithm is used for denoising, as follows:

[0010] Assume that the original noisy image is v, and the denoised image is The gray value of the pixel in is obtained as follows:

[0011]

[0012] Among them, w(x,y) represents the similarity between pixel points x and y, which reflects the distance between the rectangular neighborhoods centered on x and y.

[0013]

[0014] Where Z(x) represents the normalization coefficient and h represents the Gaussian smoothing parameter.

[0015] Step 6: Mark the images of healthy controls as 0, and mark the images of patients with knee arthritis as 1, 2, 3, and 4 according to the severity (1-4 represents the severity, 1 is the mildest injury, and 4 is the most severe injury);

[0016] Step 7: Use the discriminative autoencoder to extract features; the input image is sampled as 224×224, and the columns are concatenated into a column vector. Here the sample set is {x i}, i = 0, 1, 2, ..., 4, the output label is {y i},i=0,1,2,...,4. The loss function of the discriminative autoencoder model consists of three parts:

[0017] L=L AE +L distance +L weight

[0018] in,

[0019]

[0020] And the encoder function in the autoencoder model is f enc , the decoder function is f dec , the implicit variable is z i, N is the number of samples, I is the number of categories, θ is the number of model hyperparameters in the autoencoder, is the weight of the encoder hidden layer.

[0021]

[0022] L distance Represents the distance between different categories of discriminative encoders expected by categories |μ i -μ j |Maximum, μ i The autoencoder model learns the mean of the latent features in the low-dimensional space. is the variance, m (i) is the number of training samples in the category.

[0023]

[0024] L weight represents the L2 regularization parameter, and K is the number of hidden layers in the autoencoder.

[0025] Step 8: Output the classification results. After learning the parameters corresponding to each category using the above discriminative autoencoder model, a logistic regression model is connected after the autoencoder model of each category to determine whether the input latent variable is of the current type. The specific model is as follows:

[0026]

[0027] Among them, f(z i ) is the linear transformation of latent variables, γ(z i ) is the output value of the Sigmod function. In the discriminative self-encoding model, the output {γ 1 (z1),γ 2 (z2),...,γ N (z N )}. Take γ(z i ) is the final output and converted into labels 0, 1, 2, 3, 4. The present invention uses classification accuracy, recall rate, precision rate, and F1 score to measure the classification effect; the calculation method is as follows:

[0028]

[0029] BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0031] Figure 1The present invention is a flow chart of a method for predicting knee joint injury grade based on CT;

[0032] Figure 2 A schematic diagram of knee joint imaging injury levels used in the present invention;

[0033] Figure 3 This is a schematic diagram of the prediction model proposed by the present invention;

[0034] Figure 4 This is the receiver operating characteristic curve of the model proposed in the present invention for predicting five grades of knee joints. DETAILED DESCRIPTION

[0035] In order to make the purpose and technical implementation of the present invention clearer, embodiments of the present invention are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present invention.

[0036] In the description of the present invention, unless otherwise specified and limited, it should be noted that terms such as "damage", "image", "encoder", "decoder", "connection", "model", "optimization", "training set" and "test set" should be understood in a broad sense. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0037] like Figure 1 This application provides a CT-based knee injury grade prediction method, which uses 500 CT data of subjects for testing, including 145 healthy controls, 130 patients with injury grade I, 75 patients with injury grade 2, 70 patients with injury grade 3, and 80 patients with injury grade 4. The test includes the following steps:

[0038] Step 1: Obtain CT images of a group of healthy controls (HC) and a group of subjects with different degrees of knee joint damage, a total of 500 cases;

[0039] Step 2: Preprocess the X-ray image, crop the knee joint area and convert the grayscale image into three channels of L, a and b using the LAB color model;

[0040] Step 3, the L channel image is enhanced by using adaptive histogram equalization;

[0041] Step 4: Integrate the enhanced L, a, and b channels;

[0042] Step 5: De-noise the enhanced image. Here, the fast non-local mean denoising algorithm is used for denoising, as follows:

[0043] Assume that the original noisy image is v, and the denoised image is The gray value of the pixel in is obtained as follows:

[0044]

[0045] Among them, w(x,y) represents the similarity between pixel points x and y, which reflects the distance between the rectangular neighborhoods centered on x and y.

[0046]

[0047] Where Z(x) represents the normalization coefficient and h represents the Gaussian smoothing parameter.

[0048] Step 6: Figure 2 , the images of healthy controls were marked as 0, and the images of patients with knee arthritis were marked as 1, 2, 3, 4 according to the severity (1-4 represents the severity, 1 is the mildest injury, 4 is the most severe injury);

[0049] Step 7: Use discriminative autoencoder to extract features; Figure 3 , the input image is sampled to 224×224, and the columns are concatenated into a column vector. Here the sample set is {x i}, i = 0, 1, 2, ..., 4, the output label is {y i},i=0,1,2,...,4. The loss function of the discriminative autoencoder model consists of three parts:

[0050] L=L AE +L distance +L weight

[0051] in,

[0052]

[0053] And the encoder function in the autoencoder model is f enc , the decoder function is f dec , the implicit variable is z i , N is the number of samples, I is the number of categories, θ is the number of model hyperparameters in the autoencoder, is the weight of the encoder hidden layer.

[0054]

[0055] Ldistance Represents the distance between different categories of discriminative encoders expected by categories |μ i -μ j |Maximum, μ i The autoencoder model learns the mean of the latent features in the low-dimensional space. is the variance, m (i) is the number of training samples in the category.

[0056]

[0057] L weight represents the L2 regularization parameter, and K is the number of hidden layers in the autoencoder.

[0058] Step 8: Output the classification results. Figure 4 After learning the parameters corresponding to each category using the above discriminative autoencoder model, a logistic regression model is connected after the autoencoder model of each category to determine whether the input latent variable is of the current type. The specific model is as follows:

[0059]

[0060] Among them, f(z i ) is the linear transformation of latent variables, γ(z i ) is the output value of the Sigmod function. In the discriminative self-encoding model, the output {γ 1 (z1),γ 2 (z2),...,γ N (z N )}. Take γ(z i ) is the final output and converted into labels 0, 1, 2, 3, 4. The present invention uses classification accuracy, recall rate, precision rate, and F1 score to measure the classification effect; the calculation method is as follows:

[0061]

[0062] The test results are shown in the following table:

[0063]

[0064]

[0065] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A method for predicting knee arthritis damage based on CT images, characterized in that: The steps include: Step 1: Obtain CT images of a group of healthy controls and a group of subjects with different degrees of knee joint damage; Step 2: Preprocess the CT image, crop the knee joint area and convert the grayscale image into three channels of L, a and b using the LAB color model; Step 3, the L channel image is enhanced by using adaptive histogram equalization; Step 4: Integrate the enhanced L, a, and b channels; Step 5, denoising the enhanced image; The denoising algorithm is as follows: Assume that the original noisy image is v, and the denoised image is The gray value of the pixel in is obtained as follows: Among them, w(x,y) represents the similarity between pixel points x and y, which reflects the distance between the rectangular neighborhoods centered on x and y. Where Z(x) represents the normalization coefficient, and h represents the Gaussian smoothing parameter; Step 6: Mark the images of healthy controls as 0, and mark the images of patients with knee arthritis as 1, 2, 3, and 4 according to the severity; Step seven, use the discriminative autoencoder to extract features; after using the discriminative autoencoder model to learn the parameters corresponding to each category, connect the logistic regression model after the autoencoder model of each category to determine whether the input latent variable is of the current type.

2. The method for predicting knee arthritis damage based on CT images according to claim 1, characterized in that: The input image is sampled as 224×224, and the columns are concatenated into a column vector; the sample set is {x i }, i = 0, 1, 2, ..., 4, the output label is {y i },i=0,1,2,...,4; The loss function of the discriminative autoencoder model consists of three parts, L=L AE +L dis tan ce +L weight And the encoder function in the autoencoder model is f enc , the decoder function is f dec , the implicit variable is z i , N is the number of samples, I is the number of categories, θ is the number of model hyperparameters in the autoencoder, is the weight of the encoder hidden layer; L distance Represents the distance between different categories of discriminative encoders expected by categories |μ i -μ j |Maximum, μ i The autoencoder model learns the mean of the latent features in the low-dimensional space. is the variance, m (i) is the number of training samples in the category; L weight represents the L2 regularization parameter, and K is the number of hidden layers in the autoencoder.

3. The method for predicting knee arthritis damage based on CT images according to claim 1, characterized in that: The discriminative autoencoder model is as follows: in, is the linear transformation of latent variables, Output value of the Sigmod function; in the discriminative self-encoding model, the output {γ 1 (z1),γ 2 (z2),...,γ N (z N )};Pick The largest category is taken as the final output and converted to labels 0, 1, 2, 3, 4.

4. The method for predicting knee arthritis damage based on CT images according to claim 3, characterized in that: The classification accuracy, recall rate, precision rate and F1 score are used to measure the classification effect; the calculation method is as follows:

Images