Treatment method and system based on diabetic foot

Abstract

The invention relates to the technical field of image processing, in particular to a treatment method and system based on diabetic foot, and the system comprises a millimeter wave therapeutic instrument which is used for treating the diabetic foot, and comprises the steps: collecting a foot image of a patient, and preprocessing the foot image to obtain a target image; and inputting the target image into a CNN network classification model to obtain a first probability value that the patient area in the target image belongs to various grade labels, the various grade labels including a first grade label, a second grade label, a third grade label and a fourth grade label. According to the method, the first probability value that the patient area belongs to the various grade labels is calculated through the CNN network classification model, the second probability value that the patient area belongs to the various grade labels is calculated through the similarity between the target image and the historical foot image, the results of the first probability value and the second probability value are combined, and different weights are given to the results. And the accuracy of judging the illness level of the patient is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology. More specifically, the present invention relates to a treatment method and system based on diabetic foot. Background Art

[0002] Diabetic foot ulcers (DFUs) are a major complication of diabetes and can put patients at risk of death if not properly treated. However, DFUs require doctors to carefully diagnose the affected area, and require long-term treatment and expensive treatment and nursing costs. Therefore, if deep learning can be used to automatically classify DFUs, the diagnosis and treatment efficiency of DFU patients will be greatly improved. By developing automatic labelers, foot images can be automatically segmented and classified without the help of clinicians, and automatic detection, recognition, and segmentation of ulcers can be developed with the help of these classifiers; at the same time, various software tools (such as mobile applications for ulcer recognition) can be combined to facilitate self-diagnosis by users. In addition, this framework may be helpful for classifying other skin injuries, such as wound classification, infections such as chickenpox or herpes zoster, or other skin lesions such as moles, freckles, papules, etc.

[0003] However, most of the current methods for identifying the severity of diabetic foot disease use deep learning methods to identify the severity of diabetic foot disease. However, if the data used to train the model is mislabeled, the model's learning process will be affected, resulting in inaccurate prediction results. Or if the amount of data for a certain type of sample (for example, severe diabetic foot disease) is small, the deep learning model may not be able to effectively learn the characteristics of this category, resulting in prediction bias. Therefore, relying solely on deep learning methods to identify the severity of diabetic foot disease in patients will not result in accurate recognition results. Summary of the invention

[0004] The present invention provides a treatment method and system based on diabetic foot, aiming to solve the problem that the related art only relies on deep learning methods to judge and identify the severity of patients' diabetic foot, and the identification results are not accurate enough.

[0005] In a first aspect, the present invention provides a treatment method based on diabetic foot, including a millimeter wave therapeutic device for treating the diabetic foot, including: collecting a patient's foot image, preprocessing the foot image, and obtaining a target image; inputting the target image into a CNN network classification model to obtain a first probability value that the diseased area in the target image belongs to each type of grade label, wherein each type of grade label includes a first-level serious illness label, a second-level serious illness label, a third-level serious illness label, and a fourth-level serious illness label; according to the similarity between the target image and a historical foot image, calculating a second probability value that the diseased area in the target image belongs to each type of grade label, wherein the similarity is negatively correlated with the difference between the average grayscale value of all pixels in the target image and the historical foot image; weighted summing the first probability value and the second probability value of the target image in the same type of grade label to obtain a final probability value of the type of grade label, determining the patient's disease grade according to the final probability value, and treating the patient based on the patient's disease grade.

[0006] Furthermore, the foot image is preprocessed to obtain a target image, including: graying the foot image, performing edge detection on the grayed foot image, extracting a diseased area in the foot image, and using the image of the diseased area as the target image.

[0007] Furthermore, the second probability value that the diseased area in the target image belongs to each type of grade label is calculated, including: calculating the second probability value that the target image belongs to the i-th grade label, including: calculating a first average grayscale value of all pixels in the target image, and a second average grayscale value of all pixels in all historical foot images belonging to the i-th grade label; calculating the difference between the first average and the second average, and taking the normalized difference as the second probability value; traversing the above steps to obtain the second probability value that the target image belongs to each type of grade label.

[0008] Furthermore, calculating the second probability value that the diseased area in the target image belongs to each category of grade labels also includes: calculating the second probability value that the target image belongs to the i-th category of grade labels, including: calculating the sum of the difference in grayscale values ​​of all pixels in the target image and the pixel points at the same position in the historical foot image of the i-th category of grade labels, and taking the normalized sum of the differences as the second probability value; traversing the above steps to obtain the second probability value that the target image belongs to each category of grade labels.

[0009] Furthermore, the second probability value of the target image belonging to the i-th class label is calculated, and the calculation formula is: ; Indicates the second probability value of the target image belonging to the i-th class label, Indicates the target image The gray value of a pixel, represents the number of all historical foot images belonging to the i-th class label The average gray value of pixels, N represents the total number of pixels in the target image. Represents the standard normalization function.

[0010] Further, weighted summing of the first probability value and the second probability value of the target image in the same class level label includes: assigning different weights to the first probability value and the second probability value, wherein the weight of the first probability value is , the weight of the second probability value is , The empirical value is 0.4.

[0011] Furthermore, determining the disease level according to the final probability value includes: obtaining the final probability value of the target image belonging to each type of grade label, and sorting the final probability values ​​of each type of grade label by size, and taking the grade label corresponding to the largest final probability value as the disease level of the patient's foot.

[0012] Furthermore, the patient is treated based on his / her illness level, including: adjusting the treatment time of the millimeter wave therapy device according to the patient's illness level, wherein the treatment time is positively correlated with the illness level; in response to the patient's foot illness level being a level one serious illness label, a level two serious illness label, a level three serious illness label or a level four serious illness label, the treatment time of the millimeter wave therapy device is increased by the corresponding preset time lengths in sequence.

[0013] Furthermore, the preset duration is 10 minutes. In a second aspect, the present invention further provides a treatment system for diabetic foot, comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to implement any of the above treatment methods for diabetic foot.

[0014] Beneficial effects: (I) The CNN network classification model is used to calculate the first probability value of the patient area belonging to each type of grade label. The second probability value of the patient area belonging to each type of grade label is calculated through the similarity between the target image and the historical foot image. The results of the two are combined and assigned different weights, which improves the accuracy of judging the severity of the patient's illness.

[0015] (ii) After determining the severity of the patient's illness, the treatment time of the millimeter wave therapy device can be adjusted according to the severity of the patient's illness, thereby increasing the diversity of treatment and enabling better treatment of the patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Several embodiments of the present invention are shown in an exemplary rather than restrictive manner by reading the following detailed description with reference to the accompanying drawings, and the same or corresponding reference numerals represent the same or corresponding parts, wherein: Figure 1 It is a flowchart schematically showing the acquisition of the patient's critical illness level according to an embodiment of the present invention. Specific embodiments

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] In one embodiment, diabetic foot is one of the common and serious complications of diabetes, referring to foot problems caused by long-term poor blood glucose control, resulting in peripheral nerve damage, blood circulation disorders, and weakened immune function. These problems usually manifest as foot ulcers, infections, necrosis, etc., and may lead to amputation in severe cases. The symptoms of diabetic foot vary from person to person, and common manifestations include: Numbness and tingling: Due to nerve damage, patients often feel numbness, tingling, burning, or pain in the feet, especially at night. Foot ulcers and wounds: Patients may have skin breaks or ulcers on the feet, and the ulcers usually appear on the soles, toes, or heels. Due to nerve damage, patients may not feel much pain. Swelling and redness: The feet may swell due to infection or poor blood circulation, especially when infected, the feet may become red, swollen, and hot.

[0020] In one embodiment, for patients with diabetic foot, millimeter waves can improve local blood circulation, reduce blood viscosity, promote wound healing, and are very helpful for the treatment of skin ulcers and peripheral neuritis. However, currently, when treating patients with diabetic foot, it is generally through the doctor's manual judgment of the severity of the patient's illness. There are small visual differences between the symptoms of some patients, and it is easy to have errors and inaccurate judgment accuracy through manual judgment. Therefore, the present invention solves the above problems through the following steps.

[0021] S101: Collect the foot image of the patient, preprocess the foot image, and obtain a target image.

[0022] In one embodiment, the patient's foot image is captured by a camera, and in one embodiment, the camera may be a video camera, etc. Preprocessing the foot image includes: graying the foot image, detecting the grayed foot image using an edge detection algorithm, extracting the diseased area in the foot image, and using the image of the diseased area as the target image.

[0023] It should be noted that the reason for grayscale processing of foot images is that foot images usually contain rich color information, but for most image processing and lesion area detection tasks, color information is not the most important. After converting the image to a grayscale image, the complexity of the calculation can be effectively reduced, because the grayscale image only contains brightness information, not color information. In this way, the structural and morphological characteristics of the image can be focused on, reducing the dependence on color and tone, and the lesion areas in medical images, especially in foot diseases (such as diabetic foot ulcers, skin infections, etc.), often show brightness changes or texture features different from normal skin. The brightness changes (high and low grayscale values) of grayscale images can directly reflect the abnormality of these lesion areas, while color information (such as red, green, and blue channels) may be interfered with by the color of healthy areas. Therefore, grayscale processing helps to highlight the contrast of the lesion area more clearly, making the edges and lesion areas easier to identify. In this embodiment, the diseased area in the foot image can be obtained by using the Canny edge detection algorithm, and the Canny edge detection accurately detects the edge in the image through a series of steps (smoothing, gradient calculation, non-maximum suppression, double thresholds and edge connection). Its advantages are: it can effectively remove noise; accurately locate edges; and has a strong ability to capture edge details. In other embodiments, the diseased area can also be extracted from the foot image by using threshold segmentation and region growing methods, which will not be described in detail here.

[0024] S101: Obtain a first probability value that the diseased area in the target image belongs to each level label.

[0025] In one embodiment, the target image is input into a CNN network classification model to obtain a first probability value of the patient area in the target image belonging to each type of grade label. The process of constructing the CNN network classification model is as follows: any number of foot images are selected from a hospital database, divided into training sets in proportion, and the foot images are preprocessed, wherein the preprocessing includes: resizing: all images are adjusted to a uniform size (such as 224x224 or 256x256) to meet the requirements of the CNN input layer. Grayscale adjustment, converting the foot image into a grayscale image. Standardization: normalizing the pixel value to a certain range (for example: 0-1), or performing z-score standardization (mean is 0, standard deviation is 1), and manually assigning a corresponding grade label to each image, wherein the grade label includes a first-level severe disease label, a second-level severe disease label, a third-level severe disease label, and a fourth-level severe disease label. Then the CNN network classification model is trained according to the training set. After the training is completed, the input feature of the CNN network classification model is the foot image, and the output feature of the CNN network classification model is the grade label corresponding to the foot image.

[0026] It should be noted that the first-level serious illness label, the second-level serious illness label, the third-level serious illness label, and the fourth-level serious illness label are arranged from low to high according to the severity of the illness.

[0027] S103: Calculate the second probability value that the diseased area in the target image belongs to each class label.

[0028] In one embodiment, based on the similarity between the target image and the historical foot image, a second probability value of the diseased area in the target image belonging to each grade label is calculated, and the similarity is negatively correlated with the difference between the average grayscale values ​​of all pixels in the target image and the historical foot image.

[0029] In one embodiment, calculating the second probability value that the diseased area in the target image belongs to each category of grade labels includes: calculating the second probability value that the target image belongs to the i-th category of grade labels, including: calculating a first average grayscale value of all pixels in the target image, and a second average grayscale value of all pixels in all historical foot images belonging to the i-th category of grade labels; calculating the difference between the first average and the second average, and taking the normalized difference as the second probability value; traversing the above steps to obtain the second probability value that the target image belongs to each category of grade labels.

[0030] In another embodiment, calculating the second probability value that the diseased area in the target image belongs to each type of grade label also includes: calculating the second probability value that the target image belongs to the i-th grade label, including: calculating the sum of the difference between the grayscale values ​​of all pixels in the target image and the historical foot image of the i-th grade label at the same position, and taking the normalized sum of the difference as the second probability value; traversing the above steps to obtain the second probability value that the target image belongs to each type of grade label. The calculation formula for the second probability value in this embodiment is: ; Indicates the second probability value of the target image belonging to the i-th class label, Indicates the target image The gray value of a pixel, represents the number of all historical foot images belonging to the i-th class label The average gray value of pixels, N represents the total number of pixels in the target image. Represents the standard normalization function. So far, the similarity between the target image and each level label can be obtained according to the similarity between the pixel points at the same position in the historical foot image in each level label. The higher the similarity between the target image and a certain level, the greater the second probability value that the target image belongs to that level.

[0031] In one embodiment, the disease level of the target image can be determined based on the second probability value that the target image belongs to the class level. For example, the second probability value of the target image belonging to the first-level serious disease label is calculated as , the second probability value of the target image belonging to the first-level serious disease label is calculated as , the second probability value of the target image belonging to the third-level severe disease label is calculated as , the second probability value of the target image belonging to the fourth level serious disease label is calculated as . The order from largest to smallest is , , , , and select the level label with the second largest probability value as the patient's disease level, that is, The corresponding first-level serious illness label is the patient's illness level.

[0032] S104: Perform a weighted summation of the first probability value and the second probability value of the target image in the same class level label to obtain a final probability value of the class level label.

[0033] In one embodiment, the formula for weighted sum of the first probability value and the second probability value is: , where Indicates the final probability value that the target image belongs to the i-th class label, Indicates the first probability value of the target image belonging to the i-th class label, Indicates the second probability value of the target image belonging to the i-th class label, represents the weight of the first probability value calculated by the CNN network classification model, Represents the weight of the second probability value.

[0034] In one embodiment, the first probability value calculated by the CNN network classification model can be directly set, and the weight of the first probability value is set to 0.4, and the weight of the second probability value is set to 0.6. The reason is that during the calculation process, the second probability value is calculated based on the difference in pixels at the same position between the target image and the historical foot image, so the final calculated result is relatively accurate, and therefore a larger weight is given to the second probability value.

[0035] In other embodiments, the weight of the first probability value can be adjusted at any time according to the classification accuracy of the CNN network classification model. If the classification accuracy of the CNN network classification model is high, the weight of the first probability value calculated by the CNN network classification model can be increased. Otherwise, the weight of the first probability value calculated by the CNN network classification model can be reduced.

[0036] S105: Determine the disease level according to the final probability value, and treat the patient based on the disease level.

[0037] In one embodiment, determining the disease level includes: obtaining final probability values ​​of the target image belonging to each level label, sorting the final probability values ​​of each level label by size, and taking the level label corresponding to the largest final probability value as the disease level of the patient's foot.

[0038] For example, the final probability value of the target image belonging to the first-level severe disease label is calculated as , the final probability value of the target image belonging to the first-level severe disease label is calculated as , the final probability value of the target image belonging to the third-level severe disease label is calculated as , the final probability value of the target image belonging to the fourth level of serious illness label is calculated as . The order from largest to smallest is , from which we can select The corresponding four-level severity label is used as the patient's illness level.

[0039] In one embodiment, after obtaining the disease level of a patient, treatment is carried out based on the patient's disease level. Specifically, the treatment time of the millimeter wave therapeutic apparatus can be adjusted according to the patient's disease level. The higher the patient's disease level, the longer the treatment time of the millimeter wave therapeutic apparatus, and vice versa, the shorter the treatment time of the millimeter wave therapeutic apparatus.

[0040] In one embodiment, when the disease level of the patient's foot is a first-level severe disease label, a second-level severe disease label, a third-level severe disease label, or a fourth-level severe disease label, the treatment time of the millimeter wave therapeutic apparatus is increased by a corresponding preset duration in sequence, where the preset duration is 10 minutes.

[0041] Exemplarily, when the disease level of the patient's foot is a first-level severe disease level, the treatment time is the basic treatment time. Among them, the basic treatment time can be considered set, such as 20 minutes or 30 minutes, etc. When the disease level of the patient's foot is a second-level severe disease level, the treatment time for the second-level severe disease level is increased by 10 minutes on the basis of the basic treatment time. When the disease level of the patient's foot is a third-level severe disease level, it is increased by 10 minutes on the treatment time for the second-level severe disease level, and so on.

[0042] In other embodiments, adjusting the treatment time of the millimeter wave therapeutic apparatus according to the patient's disease level further includes: correcting the basic treatment time according to the patient's severe disease level, and the correction formula is: , where represents the final treatment time of the i-th type of level label, represents the correction coefficient of the i-th type of level label, represents the initial treatment time. Among them, the correction coefficient for the first-level severe disease label is 0.3, the correction coefficient for the second-level severe disease label is 0.5, the correction coefficient for the third-level severe disease label is 0.6, and the correction coefficient for the fourth-level severe disease label is 0.8. That is to say, the higher the patient's severe disease level, the larger the correction coefficient corresponding to this level, and the longer the treatment time for the patient.

[0043] According to the above steps, the first probability value of the diseased area belonging to each type of level label is calculated through the CNN network classification model, and the second probability value of the diseased area belonging to each type of level label is calculated through the similarity between the target image and the historical foot image. The results of the two are combined and different weights are assigned to them, improving the accuracy in judging the patient's severe disease level.

[0044] The present invention also provides a treatment method and system for diabetic foot. The system includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, a treatment method and system for diabetic foot according to the first aspect of the present invention are implemented.

[0045] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0046] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0047] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0048] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0049] The above-mentioned embodiments only express several implementation methods of the present invention, and the description is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent application. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

Claims

1. A method for treating diabetic foot, characterized in that: include: Acquiring a foot image of a patient, and preprocessing the foot image to obtain a target image; Input the target image into the CNN network classification model to obtain a first probability value of the patient area in the target image belonging to each level label, wherein each level label includes a first level serious illness label, a second level serious illness label, a third level serious illness label, and a fourth level serious illness label; Calculate a second probability value of the diseased area in the target image belonging to each level label according to the similarity between the target image and the historical foot image, wherein the similarity is negatively correlated with the difference between the average grayscale value of all pixels in the target image and the average grayscale value of all pixels in the historical foot image; The first probability value and the second probability value of the target image in the same class level label are weightedly summed to obtain a final probability value of the class level label, the patient's disease level is determined according to the final probability value, and the patient is treated based on the disease level.

2. The method for treating diabetic foot according to claim 1, characterized in that: Preprocessing the foot image to obtain a target image includes: The foot image is grayed, and edge detection is performed on the grayed foot image to extract the diseased area in the foot image, and the image of the diseased area is used as the target image.

3. The method for treating diabetic foot according to claim 1, characterized in that: Calculate the second probability value of the patient area in the target image belonging to each level label, including: Calculating the second probability value of the target image belonging to the i-th class label includes: Calculate a first average value of the grayscale values ​​of all pixels in the target image and a second average value of the grayscale values ​​of all pixels in the historical foot images belonging to the i-th class level label; Calculating a difference between the first average value and the second average value, and using the normalized difference as a second probability value; The above steps are repeated to obtain the second probability value that the target image belongs to each level label.

4. The method for treating diabetic foot according to claim 1, characterized in that: Calculating the second probability value of the patient area in the target image belonging to each level label also includes: Calculating the second probability value of the target image belonging to the i-th class label includes: Calculate the sum of the differences between the grayscale values ​​of all pixels in the target image and the pixels at the same position in the historical foot image of the i-th grade label, and use the normalized sum of the differences as the second probability value; The above steps are repeated to obtain the second probability value that the target image belongs to each level label.

5. The method for treating diabetic foot according to claim 4, characterized in that: Calculate the second probability value of the target image belonging to the i-th class label. The calculation formula is: ; Indicates the second probability value of the target image belonging to the i-th class label, Indicates the target image The gray value of a pixel, represents the number of all historical foot images belonging to the i-th class label The average gray value of pixels, N represents the total number of pixels in the target image. Represents the standard normalization function.

6. The method for treating diabetic foot according to claim 1, characterized in that: The weighted summation of the first probability value and the second probability value of the target image in the same class level label includes: Different weights are assigned to the first probability value and the second probability value, wherein the weight of the first probability value is , the weight of the second probability value is , The empirical value is 0.

4.

7. The method for treating diabetic foot according to claim 1, characterized in that: Determining the disease level according to the final probability value includes: The final probability values ​​of the target image belonging to each level label are obtained, and the final probability values ​​of each level label are sorted by size, and the level label corresponding to the largest final probability value is used as the disease level of the patient's foot.

8. The method for treating diabetic foot according to claim 1, characterized in that: Treatment is based on the severity of the disease and may include: Adjusting the treatment time of the millimeter wave therapeutic device according to the patient's disease level, wherein the treatment time is positively correlated with the disease level; In response to the patient's foot disease level being a level one serious disease label, a level two serious disease label, a level three serious disease label or a level four serious disease label, the treatment time of the millimeter wave therapy device is increased by the corresponding preset time lengths in sequence.

9. The method for treating diabetic foot according to claim 8, characterized in that: The preset duration is 10 minutes.

10. A treatment system for diabetic foot, comprising a processor and a memory, characterized in that: The memory stores a computer program, and the processor executes the computer program to implement the treatment method based on diabetic foot as described in any one of claims 1-9.

Images