Laser millimeter wave therapeutic instrument for treating diabetic foot and control method thereof

Abstract

The invention relates to the field of medical instruments, in particular to a laser millimeter wave therapeutic instrument for treating diabetic feet and a control method of the laser millimeter wave therapeutic instrument. The method comprises the following steps: acquiring an image of a diabetic foot part, and obtaining a Wagner grade of the diabetic foot; preliminarily determining an illumination intensity grade according to Wagner grading of the diabetic foot; adjusting the preliminarily determined illumination intensity according to the age of the patient so as to obtain the preliminarily adjusted illumination intensity; further adjusting the illumination intensity after the preliminary adjustment according to the disease duration of the patient so as to obtain the target illumination intensity of the laser therapeutic instrument, including: if the disease duration does not exceed 3 years, reducing the illumination intensity after the preliminary adjustment by 10%; if the illness duration exceeds 5 years, the illumination intensity after preliminary adjustment is increased by 10%; and during treatment, the laser irradiation intensity of the laser therapeutic instrument is adjusted to the target illumination intensity. By adopting the method, the treatment effect of the laser millimeter wave therapeutic instrument on the diabetic foot can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of medical devices. More specifically, this invention relates to a laser millimeter-wave therapy device for treating diabetic foot and its control method. Background Technology

[0002] Diabetic foot is one of the most serious complications of diabetes, posing a significant threat to patients' health and quality of life. Laser millimeter wave therapy, as an effective physical therapy method, improves local blood circulation and promotes tissue repair and regeneration by irradiating the affected area with lasers of specific intensities. The treatment effect largely depends on the precise setting of treatment parameters, especially the intensity of laser irradiation and the duration of each treatment session. In clinical practice, the setting of these parameters often relies on the physician's personal experience, lacking unified and objective standards. This easily leads to a "one-size-fits-all" treatment model, making it difficult to personalize adjustments based on the patient's specific condition, thus affecting the final treatment outcome.

[0003] To address these issues, existing technologies are beginning to incorporate artificial intelligence to assist in treatment planning. For example, deep learning-based neural network models can be used to analyze images of diabetic foot lesions and automatically identify and determine their clinical grading (e.g., Wagner classification). Treatment parameters for laser millimeter-wave therapy can then be set based on the Wagner classification of the diabetic foot lesion. However, relying solely on a single pathological grading output by the neural network model to determine treatment parameters ignores the individual differences between patients. For instance, factors such as the patient's age and medical history significantly influence tissue repair capabilities and the degree of response to treatment. Therefore, relying solely on the disease grading derived from image recognition to formulate treatment plans results in a relatively singular decision-making dimension, failing to achieve truly precise and personalized treatment and hindering further improvements in treatment outcomes.

[0004] In summary, the existing treatment methods for diabetic foot using laser millimeter wave therapy devices have technical problems such as poor applicability and poor treatment effects. Summary of the Invention

[0005] To address the technical problems of poor applicability and poor treatment effect in existing treatment plans for diabetic foot using laser millimeter wave therapy devices, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a control method for a laser millimeter-wave therapy device for treating diabetic foot, comprising: Images of diabetic foot areas are acquired, and Wagner classifications of diabetic foot are obtained by combining them with a pre-defined neural network model. The light intensity level is initially determined based on the Wagner classification of diabetic foot, including: if the Wagner classification is less than or equal to 1, a low light intensity level is selected; if the Wagner classification is 2, a medium light intensity level is selected; if the Wagner classification is greater than or equal to 3, a high light intensity level is selected. The initially determined light intensity is adjusted based on the patient's age to obtain the preliminarily adjusted light intensity, including: if the patient is older than 40 years but younger than 60 years, the current light intensity level is multiplied by 80%; if the patient is older than 60 years, the current light intensity level is multiplied by 60%; if the patient is younger than 40 years, the current light intensity level is multiplied by 100%. The initial light intensity is further adjusted based on the duration of the patient's illness to obtain the target light intensity for the laser therapy device. This includes: if the duration of the illness is less than 3 years, the initial light intensity is reduced by 10%; if the duration of the illness is more than 5 years, the initial light intensity is increased by 10%. During treatment, the laser irradiation intensity of the laser therapy device is adjusted to the target light intensity.

[0007] Preferably, the low-level light intensity ranges from 10-20 mW / cm². 2 The range of the medium-level light intensity is 20-40 mW / cm². 2 The range of the high-level light intensity is 40-80 mW / cm². 2 .

[0008] Preferably, it further includes: setting the single treatment duration of the laser therapy device according to the Wagner classification of diabetic foot, including: if the Wagner classification is level 1, the single treatment duration is set to 5 minutes; if the Wagner classification is level 2 or 3, the single treatment duration is set to 10 minutes; if the Wagner classification is level 4 or above, the single treatment duration is set to 15 minutes.

[0009] Preferably, obtaining the Wagner classification of diabetic foot lesions by combining a preset neural network model includes: The images of the ulcer core region and the annular periulcer region were extracted from the images of the diabetic foot. The annular periulcer region is the ring-shaped area surrounding the ulcer core region. The image of the ulcer core region is input into a pre-defined neural network model to obtain an initial score vector. and the initial probability distribution vector The elements in the initial probability distribution vector represent the initial probabilities of each Wagner classification, and the elements in the initial score vector represent the initial scores of each Wagner classification; and the classification uncertainty index is calculated. The calculation expression is: ; In the formula, The initial probability for the i-th Wagner classification; The vascular vitality index of the peri-wound tissue was calculated based on the image of the annular peri-wound area. The vascular vitality index was positively correlated with the average intensity of the red channel of all pixels in the peri-wound area, and negatively correlated with the average intensity of the green channel and the average intensity of the blue channel of all pixels in the peri-wound area. Calculate the context dynamic correction factor F used to correct the initial score vector. The context dynamic correction factor is positively correlated with the grading uncertainty index and negatively correlated with the vascular vitality index of the peri-traumatic tissue. Construct the correction direction vector The elements of the correction direction vector are all no greater than 1 and are monotonically non-decreasing from left to right; The initial score vector is corrected according to the context dynamic correction factor F to obtain a new score vector. The new score vector The new score used to characterize each Wagner level is calculated using the following expression: ; New score vector The input is fed into the Softmax function of the output layer of the preset neural network model to obtain the final probability distribution vector after global context information correction; and the Wagner grade corresponding to the largest element in the final probability distribution vector is selected as the Wagner grade of diabetic foot.

[0010] Preferably, extracting the ulcer core area image and the annular surrounding wound area image from the image of the diabetic foot site includes: An image segmentation algorithm is used to locate and segment the ulcer core region in the image, and a morphological dilation operation is performed on the binary mask of the segmented ulcer core region to generate the first region; The image of the annular surrounding wound region is obtained by performing a difference operation between the first region and the ulcer core region.

[0011] Preferably, the image segmentation algorithm may employ a U-Net network.

[0012] Preferably, the formula for calculating the vascular vitality index of the peri-injury tissue is: ; In the formula, This indicates the vascular vitality index of the tissues surrounding the wound. , and These represent the average intensity of the red channel, the average intensity of the green channel, and the average intensity of the blue channel for all pixels within the enclosed area, respectively. Let e ​​denote the natural logarithm, and let e denote the natural constant.

[0013] Preferably, the calculation expression for the context dynamic correction factor F is: ; In the formula, Indicates a graded uncertainty index. This represents an exponential function with base e. This indicates the vascular vitality index of the tissues surrounding the wound. This indicates the lower limit of the vascular vitality index.

[0014] Preferably, the preset neural network model uses a cross-entropy loss function during training.

[0015] In a second aspect, the present invention provides a laser millimeter-wave therapy device for treating diabetic foot, comprising a laser emitting system, a millimeter-wave emitting system, a laser probe connected to the laser emitting system, a millimeter-wave probe connected to the millimeter-wave emitting system, a control system, and an image acquisition device. The control system is connected to the image acquisition device and to the laser emitting system and the millimeter-wave emitting system to execute the control method of the laser millimeter-wave therapy device for treating diabetic foot of the present invention.

[0016] The beneficial effects of this invention are as follows: Compared with existing technologies that rely on doctors' subjective experience or use fixed parameters for laser treatment, this invention proposes a scientific, precise, and personalized laser therapy device control method. This method uses a neural network to objectively classify diabetic foot lesions and dynamically adjusts the light intensity by comprehensively considering three core factors: disease severity (Wagner classification), patient physiological condition (age), and medical history (duration of illness). This allows for customized treatment plans for each patient, avoiding poor treatment outcomes or safety risks caused by improper parameter settings, and significantly improving the scientific rigor, targeted approach, and effectiveness of the treatment.

[0017] Furthermore, by quantifying the health status of the peri-wound tissue as global contextual information, the model effectively addresses ambiguous cases that are difficult to distinguish based solely on local ulcer features. When the neural network model becomes confused by local information, it can introduce external, clinically relevant contextual cues for decision support, making the final grading result closer to the comprehensive judgment of experienced clinicians, thereby significantly improving the diagnostic consistency and accuracy of the neural network model in complex situations. In addition, a dynamic correction mechanism is established, making the model more resistant to local feature distortions caused by factors such as changes in lighting and shooting angles. Even if the visual features of the ulcer core area are misleading due to noise, healthy peri-wound tissue information can prevent the model from jumping to erroneous higher grades. Conversely, a poor peri-wound tissue condition can provide early warning for seemingly minor ulcers, enhancing the overall robustness of the model. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart illustrating a laser millimeter-wave therapy device control method for treating diabetic foot according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structure of a laser millimeter-wave therapy device for treating diabetic foot according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the structure of a control system according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0021] Example of a control method for a laser millimeter-wave therapy device for treating diabetic foot: like Figure 1 As shown, the control method for a laser millimeter-wave therapy device for treating diabetic foot of the present invention includes: S101. Collect images of the diabetic foot area and obtain the Wagner classification of the diabetic foot by combining them with a preset neural network model; The preset neural network model in this embodiment uses the cross-entropy loss function during training.

[0022] S102. Preliminary determination of the light intensity level of the laser millimeter wave therapy device, specifically: based on the Wagner classification of diabetic foot, the light intensity level is preliminarily determined, including: if the Wagner classification is less than or equal to level 1, a low light intensity level is selected; if the Wagner classification is level 2, a medium light intensity level is selected; if the Wagner classification is greater than or equal to level 3, a high light intensity level is selected. In this embodiment, the value range of the low-level light intensity is 10-20 mW / cm². 2 The range of the medium-level light intensity is 20-40 mW / cm². 2 The range of the high-level light intensity is 40-80 mW / cm². 2 .

[0023] S103. Adjust the initially determined light intensity, specifically by adjusting the initially determined light intensity according to the patient's age to obtain the initially adjusted light intensity, including: if the patient is older than 40 years but younger than 60 years, multiply the current level of light intensity by 80%; if the patient is older than 60 years, multiply the current level of light intensity by 60%; if the patient is younger than 40 years, multiply the current level of light intensity by 100%. S104. Obtain the target light intensity of the laser therapy device, specifically by further adjusting the initially adjusted light intensity based on the patient's duration of illness, thereby obtaining the target light intensity of the laser therapy device. This includes: if the duration of illness is no more than 3 years, reducing the initially adjusted light intensity by 10%; if the duration of illness is more than 5 years, increasing the initially adjusted light intensity by 10%. S105. During treatment, adjust the laser irradiation intensity of the laser therapy device to the target light intensity.

[0024] In one embodiment, the method further includes setting the single treatment duration of the laser therapy device according to the Wagner classification of diabetic foot, including: if the Wagner classification is level 1, the single treatment duration is set to 5 minutes; if the Wagner classification is level 2 or 3, the single treatment duration is set to 10 minutes; if the Wagner classification is level 4 or above, the single treatment duration is set to 15 minutes.

[0025] In one embodiment, obtaining the Wagner classification of diabetic foot using a preset neural network model includes: S201. Extract the ulcer core area image and the annular periulcer area image from the image of the diabetic foot site. The annular periulcer area is the annular area surrounding the ulcer core area. S202. Input the image of the ulcer core area into the preset neural network model to obtain the initial score vector. and the initial probability distribution vector The elements in the initial probability distribution vector represent the initial probabilities of each Wagner classification, and the elements in the initial score vector represent the initial scores of each Wagner classification; and the classification uncertainty index is calculated. The calculation expression is: ; In the formula, The initial probability for the i-th Wagner classification; When calculating the graded uncertainty index using the graded uncertainty index expression of this embodiment, if a certain If it approaches 1, then the rest The value will approach 0, at which point the calculated classification uncertainty index will be very small, indicating that the preset neural network model is highly certain about the classification result; conversely, if each When the values ​​are relatively close, the calculated level uncertainty index will be very large, indicating that the preset neural network model has difficulty choosing between two or more leveling results. Therefore, the calculation expression used in this embodiment can calculate the level uncertainty index more accurately.

[0026] The magnitude of the grading uncertainty index directly reflects the difficulty and ambiguity of the model in classifying the local features of the current ulcer.

[0027] Initial probability distribution vector The expression is: .

[0028] S203. Calculate the vascular vitality index of the wound tissue based on the image of the annular wound area. The vascular vitality index is positively correlated with the average red channel intensity of all pixels in the wound area, and negatively correlated with the average green channel intensity and the average blue channel intensity of all pixels in the wound area. Clinically, healthy skin with good blood supply typically appears rosy and bright, while ischemic or infected tissue with poor blood supply tends to be dull, purplish, or pale. This indicates that when the peri-wound area has good blood supply and healthy tissue, its red channel component is significantly higher than its blue and green components, and the color distribution is relatively even. Conversely, when tissue ischemia, congestion, or early infection occurs, the color becomes dull and purplish, the red channel component decreases, and the blue and green components increase. Therefore, by setting the vascular vitality index positively correlated with the average red channel intensity of all pixels in the peri-wound area and negatively correlated with the average green and blue channel intensity of all pixels in the peri-wound area, the vascular vitality index of the peri-wound tissue can be calculated relatively accurately.

[0029] S204. Calculate the context dynamic correction factor F used to correct the initial score vector. The context dynamic correction factor is positively correlated with the graded uncertainty index and negatively correlated with the vascular vitality index of the peri-traumatic tissue. The magnitude of the context dynamic correction factor reflects the degree of correction applied to the output of the neural network model. The larger the context dynamic correction factor, the greater the degree of correction applied to the output of the neural network model.

[0030] Firstly, when the local features of the ulcer are clear and the model grading certainty is high (i.e., grading uncertainty is low), excessive intervention in the model's prediction results is unnecessary. However, when the features are ambiguous and the model uncertainty is high, the vascular vitality index of the surrounding tissue should be used more to correct the model's output. Secondly, when the vascular vitality index of the surrounding tissue is high, it indicates that the surrounding tissue is very healthy. Identification of the Wagner grade of diabetic foot based solely on the characteristics of the ulcer core area is sufficient, and almost no correction to the model's output is required. Conversely, when the vascular vitality index of the surrounding tissue is low, it indicates poor health. Even if the ulcer core area appears to have a low Wagner grade, if the surrounding skin is purplish and cold, it indicates severe ischemia and a very critical overall condition, potentially developing into large-area gangrene at any time. Grading should not be based solely on the small ulcer; the poor condition of the surrounding tissue must be considered with greater weight, escalating the diagnosis. Therefore, strong correction of the model's output based on the health status of the surrounding tissue is necessary. In summary, the context dynamic correction factor F can be calculated relatively accurately by making it positively correlated with the grading uncertainty index and negatively correlated with the vascular vitality index of peri-traumatic tissues.

[0031] S205, Constructing the correction direction vector The elements of the correction direction vector are all no greater than 1 and are monotonically non-decreasing from left to right; In this embodiment, the correction direction vector The values ​​can be [0, 0.1, 0.3, 0.7, 1.0, 1.0]. By setting each element of the correction direction vector to be monotonically non-decreasing from left to right, it can be ensured that when the peri-aortic tissue condition is poor, the corrected Wagner grade tends to be a larger Wagner grade, thereby ensuring the accuracy of the final Wagner grade prediction result.

[0032] S206. Correct the initial score vector according to the context dynamic correction factor F to obtain a new score vector. Score vector The new score used to characterize each Wagner level is calculated using the following expression: ; S207, Change the new score vector The input is fed into the Softmax function of the output layer of the preset neural network model to obtain the final probability distribution vector after global context information correction; and the Wagner grade corresponding to the largest element in the final probability distribution vector is selected as the Wagner grade of diabetic foot.

[0033] The training process of the preset neural network model in this embodiment includes: (1) Collect data and preprocess it. This includes: Collect images of foot lesions, especially ulcers of diabetic foot. Data can be obtained from publicly available medical image databases (e.g., medical datasets related to diabetic foot) or collected in collaboration with hospitals and research institutions. Each image should have a corresponding Wagner classification label. For each image, label its diabetic foot with a Wagner classification (0 to 5).

[0034] After labeling the images, they are standardized, for example, by adjusting the image size to a uniform size (such as 224x224 or 256x256) to fit the neural network input.

[0035] (2) Training the neural network model, specifically including: Data loading: Load the training and validation sets. The training set is used to train the model, and the validation set is used to evaluate the model's performance in real time and prevent overfitting.

[0036] Optimizer selection: Commonly used optimizers include Adam and SGD. The Adam optimizer typically performs well in image classification tasks and can adaptively adjust the learning rate.

[0037] Training and validation: Use the training set to train the model and the validation set to monitor its performance.

[0038] The model is trained in multiple epochs, and its performance on the validation set is evaluated after each epoch.

[0039] Adjusting hyperparameters: This includes learning rate, batch size, number of training epochs, etc., and the hyperparameters are adjusted based on the performance of the validation set during training.

[0040] Existing automatic grading methods based on convolutional neural networks have a significant limitation: when extracting and analyzing features, their receptive field or attention mechanism is primarily focused on the ulcer itself, resulting in an "isolated" and localized analysis process. This design completely ignores global contextual information that is crucial for Wagner grading, such as the color, luster, swelling degree of the surrounding skin tissue, and blood supply to the extremities. Clinically, experienced experts make comprehensive judgments by evaluating these contextual clues. Therefore, when wounds of different grades are highly similar in local visual features—for example, a severely inflammatory Wagner grade 1 ulcer versus a Wagner grade 2 ulcer in the early stages of infection—existing models are prone to confusion based solely on information about the wound itself. This leads to inconsistencies between the grading results and the judgments of clinical experts, limiting their reliability in complex clinical scenarios.

[0041] The method in this embodiment effectively addresses ambiguous cases that are difficult to distinguish based solely on local ulcer features by quantifying the health status of the peri-wound tissue as global contextual information. When the model becomes confused by local information, it can introduce external, clinically relevant contextual cues for decision support, making the final grading result closer to the comprehensive judgment of experienced clinical experts, thereby significantly improving the diagnostic consistency and accuracy of the neural network model in complex situations. Furthermore, a dynamic correction mechanism is established, making the model more resistant to distortions in local features caused by factors such as changes in lighting and shooting angles. Even if the visual features of the ulcer core area are misleading due to noise, healthy peri-wound tissue information can prevent the model from jumping to erroneous higher grades. Conversely, a poor peri-wound tissue condition can provide early warnings for seemingly minor ulcers, enhancing the overall robustness of the model.

[0042] In one embodiment, extracting an image of the ulcer core region and an image of the annular surrounding wound region from an image of the diabetic foot site includes: S301. The image segmentation algorithm is used to locate and segment the ulcer core region in the image, and the morphological dilation operation is performed on the binary mask of the segmented ulcer core region to generate the first region. In this embodiment, the image segmentation algorithm may employ a U-Net network.

[0043] S302. Perform a difference operation between the first region and the ulcer core region to obtain an image of the annular surrounding wound region.

[0044] In one embodiment, the expression for calculating the vascular vitality index of the peri-traumatic tissue is: ; In the formula, This indicates the vascular vitality index of the tissues surrounding the wound. , and These represent the average intensity of the red channel, the average intensity of the green channel, and the average intensity of the blue channel for all pixels within the enclosed area, respectively. Let e ​​denote the natural logarithm, and let e denote the natural constant.

[0045] The training process of the U-Net network in this embodiment includes: 1. Obtain the dataset, including: Obtain the input image and its corresponding label image. The input image is a foot image of a diabetic foot, which can be in common medical image formats, such as color, grayscale, or infrared images. The label image is a binarized image corresponding to the input image, where ulcer areas are marked with 1 and the background with 0.

[0046] The input and label images should be the same size, and the dataset should ideally be organized and labeled to ensure that each pair of images corresponds one-to-one.

[0047] 2. Data preprocessing, including: Data preprocessing includes image standardization and data augmentation.

[0048] Standardization includes normalizing the pixel values ​​of the input image to [0, 1], or standardizing based on the mean and standard deviation of the image.

[0049] Data augmentation includes using methods such as rotation, flipping, cropping, scaling, and color changes to increase data diversity, prevent overfitting, and enhance the model's generalization ability.

[0050] 3. Construct the U-Net model. The core structure of the constructed U-Net model includes convolutional layers, pooling layers, upsampling layers (deconvolution), and skip connections.

[0051] 4. Train the U-Net model using the input image and label image. When training U-Net, the cross-entropy loss function is selected, and the Adam optimizer can be selected.

[0052] 5. Use the validation set to evaluate the performance of the U-Net model.

[0053] In one embodiment, the expression for calculating the context dynamic correction factor F is: ; In the formula, Indicates a graded uncertainty index. This represents an exponential function with base e. This indicates the vascular vitality index of the tissues surrounding the wound. This indicates the lower limit of the vascular vitality index.

[0054] Example of a laser millimeter-wave therapy device for treating diabetic foot: This invention also provides a laser millimeter-wave therapy device for treating diabetic foot. For example... Figure 2 As shown, the laser millimeter-wave therapy device for treating diabetic foot includes a laser emitting system, a millimeter-wave emitting system, a laser probe connected to the laser emitting system, a millimeter-wave probe connected to the millimeter-wave emitting system, a control system, and an image acquisition device. The control system is connected to the image acquisition device and to the laser emitting system and the millimeter-wave emitting system to execute the laser millimeter-wave therapy device control method for treating diabetic foot described in the above embodiments.

[0055] like Figure 3 As shown, the control system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement a laser millimeter wave therapy device control method for treating diabetic foot according to the first aspect of the present invention.

[0056] The control system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces. Their settings and functions are known in the art and will not be described in detail here.

[0057] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A control method for a laser millimeter-wave therapy device for treating diabetic foot, characterized in that, include: Images of diabetic foot areas are acquired, and Wagner classifications of diabetic foot are obtained by combining them with a pre-defined neural network model. The light intensity level is initially determined based on the Wagner classification of diabetic foot, including: if the Wagner classification is less than or equal to 1, a low light intensity level is selected; if the Wagner classification is 2, a medium light intensity level is selected; if the Wagner classification is greater than or equal to 3, a high light intensity level is selected. The initially determined light intensity is adjusted based on the patient's age to obtain the preliminarily adjusted light intensity, including: if the patient is older than 40 years but younger than 60 years, the current light intensity level is multiplied by 80%; if the patient is older than 60 years, the current light intensity level is multiplied by 60%; if the patient is younger than 40 years, the current light intensity level is multiplied by 100%. The initial light intensity is further adjusted based on the duration of the patient's illness to obtain the target light intensity for the laser therapy device. This includes: if the duration of the illness is less than 3 years, the initial light intensity is reduced by 10%; if the duration of the illness is more than 5 years, the initial light intensity is increased by 10%. During treatment, the laser irradiation intensity of the laser therapy device is adjusted to the target light intensity.

2. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 1, characterized in that, The range of the low-level light intensity is 10-20 mW / cm². 2 The range of the medium-level light intensity is 20-40 mW / cm². 2 The range of the high-level light intensity is 40-80 mW / cm². 2 .

3. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 1, characterized in that, Also includes: The duration of a single laser therapy session is set according to the Wagner classification of diabetic foot, including: 5 minutes for Wagner classification 1, 10 minutes for Wagner classification 2 or 3, and 15 minutes for Wagner classification 4 or above.

4. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 1, characterized in that, The method of obtaining the Wagner classification of diabetic foot by combining a preset neural network model includes: The images of the ulcer core region and the annular periulcer region were extracted from the images of the diabetic foot. The annular periulcer region is the ring-shaped area surrounding the ulcer core region. The image of the ulcer core region is input into a pre-defined neural network model to obtain an initial score vector. and the initial probability distribution vector The elements in the initial probability distribution vector represent the initial probabilities of each Wagner classification, and the elements in the initial score vector represent the initial scores of each Wagner classification; and the classification uncertainty index is calculated. The calculation expression is: ； In the formula, The initial probability for the i-th Wagner classification; The vascular vitality index of the peri-wound tissue was calculated based on the image of the annular peri-wound area. The vascular vitality index was positively correlated with the average intensity of the red channel of all pixels in the peri-wound area, and negatively correlated with the average intensity of the green channel and the average intensity of the blue channel of all pixels in the peri-wound area. Calculate the context dynamic correction factor F used to correct the initial score vector. The context dynamic correction factor is positively correlated with the grading uncertainty index and negatively correlated with the vascular vitality index of the peri-traumatic tissue. Construct the correction direction vector The elements of the correction direction vector are all no greater than 1 and are monotonically non-decreasing from left to right; The initial score vector is corrected according to the context dynamic correction factor F to obtain a new score vector. The new score vector The new score used to characterize each Wagner level is calculated using the following expression: ； New score vector The input is fed into the Softmax function of the output layer of the preset neural network model to obtain the final probability distribution vector after global context information correction; and the Wagner grade corresponding to the largest element in the final probability distribution vector is selected as the Wagner grade of diabetic foot.

5. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 4, characterized in that, Images of the ulcer core region and the surrounding annular wound region were extracted from images of the diabetic foot site, including: An image segmentation algorithm is used to locate and segment the ulcer core region in the image, and a morphological dilation operation is performed on the binary mask of the segmented ulcer core region to generate the first region; The image of the annular surrounding wound region is obtained by performing a difference operation between the first region and the ulcer core region.

6. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 5, characterized in that, The image segmentation algorithm can use the U-Net network.

7. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 4, characterized in that, The formula for calculating the vascular vitality index of the peri-traumatic tissue is as follows: ； In the formula, This indicates the vascular vitality index of the tissues surrounding the wound. , and These represent the average intensity of the red channel, the average intensity of the green channel, and the average intensity of the blue channel for all pixels within the enclosed area, respectively. Let e ​​denote the natural logarithm, and let e denote the natural constant.

8. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in claim 4, characterized in that, The expression for calculating the context dynamic correction factor F is: ； In the formula, Indicates a graded uncertainty index. This represents an exponential function with base e. This indicates the vascular vitality index of the tissues surrounding the wound. This indicates the lower limit of the vascular vitality index.

9. The control method for a laser millimeter-wave therapy device for treating diabetic foot as described in any one of claims 1 to 8, characterized in that, The preset neural network model uses the cross-entropy loss function during training.

10. A laser millimeter-wave therapy device for treating diabetic foot, characterized in that, The device includes a laser emitting system, a millimeter-wave emitting system, a laser probe connected to the laser emitting system, a millimeter-wave probe connected to the millimeter-wave emitting system, a control system, and an image acquisition device. The control system is connected to the image acquisition device and to the laser emitting system and the millimeter-wave emitting system to perform the laser millimeter-wave therapy device control method for treating diabetic foot as described in any one of claims 1 to 9.

Images